Acta Energetica Electrical Power Engineering Quarterly no. 03/2011

act

nergetica

03/2011

number 8 /year 3

Power Engineering Quarterly

3XEOLVKHU ENERGA SA

3DWURQDJH

3ROLWHFKQLND *GDñVND

ENERGA SA

$FDGHPLF &RQVXOWDQWV -DQXV] %LDïHN 0LHF]\VïDZ %UG\Ă 0LURVïDZ &]DSLHZVNL $QWRQL 'PRZVNL 0LFKDï 'XG]LDN ,VWYDQ (UOLFK $QGU]HM *UDF]\N 3LRWU .DFHMNR 7DGHXV] .DF]RUHN 0DULDQ .DěPLHUNRZVNL -DQ .LFLñVNL .ZDQJ < /HH =ELJQLHZ /XERĂQ\ -DQ 0DFKRZVNL 2P 0DOLN -RYLFD 0LODQRYLF -DQ 3RSF]\N =ELJQLHZ 6]F]HUED 0DUFLQ 6]SDN * .XPDU 9HQD\DJDPRRUWK\ -DFHN :DñNRZLF]

5HYLHZHUV 6WDQLVïDZ &]DSS $QGU]HM *UDF]\N 3LRWU .DFHMNR -DQ .LFLñVNL =ELJQLHZ /XERĂQ\ -DQ 0DFKRZVNL -ö]HI 3DVND -DQ 3RSF]\N 'HVLUH 'DXSKLQ 5DVRORPDPSLRQRQD 6\OZHVWHU 5REDN 0DULDQ 6RELHUDMVNL 3DZHï 6RZD =ELJQLHZ 6]F]HUED $UWXU :LOF]\ñVNL

(GLWRU LQ &KLHI =ELJQLHZ /XERĂQ\ 9LFH (GLWRU LQ &KLHI 5DIDï +\U]\ñVNL &RS\ (GLWRUV .DWDU]\QD ¿HOD]HN %HUQDUG -DFNVRQ 7RSLF (GLWRUV -DFHN .OXF]QLN -HU]\ 0DMHZVNL 6HEDVWLDQ 1RMHN 6WDWLVWLFDO (GLWRU 3DZHï 6]DZïRZVNL (GLWRULDO DVVLVWDQW -DNXE 6NRQLHF]Q\ 3URRIUHDGLQJ 0LURVïDZ :öMFLN *UDSKLF GHVLJQ 0LURVïDZ 0LïRJURG]NL 7\SHVHWWLQJ 5\V]DUG .XěPD 7UDQVODWLRQ 6NULYDQHN 6S ] R R 3ULQW *UDğ[ &HQWUXP 3ROLJUDğL 'LVSDWFK SUHSDUDWLRQ (1(5*$ 2EVïXJD L 6SU]HGDĝ 6S ] R R (GLWRULDO 6WDII 2IğFH $FWD (QHUJHWLFD XO 0LNRïDMD 5HMD *GDñVN 32/$1' WHO ID[ H PDLO UHGDNFMD#DFWDHQHUJHWLFD RUJ www.actaenergetica.org

(OHFWURQLF 0HGLD $QQD )LEDN &RS\ (GLWRU

3DZHï %DQDV]DN 7HFKQLFDO (GLWRU

,QIRUPDWLRQ DERXW WKH 7KH SDSHU HGLWLRQ RI $FWD (QHUJHWLFD LV WKH RULJLQDO YHUVLRQ RI WKH MRXUQDO RULJLQDO YHUVLRQ 7KH MRXUQDO LV DOVR DYDLODEOH RQ WKH ZHEVLWH ZZZ DFWDHQHUJHWLFD RUJ ,661

featuring 4

TRAVELLING WAVE FAULT LOCATION IN HV LINES Krzysztof Glik, Ryszard Kowalik Désiré Dauphin Rasolomampionona

14

ALGORITHMS TO ENSURE RELIABILITY OF POWER SYSTEM OBSERVABILITY Irina I. Golub

22

INVESTIGATION OF THE IMPACT OF SWITCHING A HEAVILY LOADED TRANSMISSION LINE ON OPERATION OF A POWER PLANT WITH TURBINE-GENERATOR UNITS Andrzej Kąkol, Bogdan Sobczak Robert Trębski

30

SYSTEM FOR DYNAMIC MODEL IDENTIFICATION BASED ON REAL POWER SYSTEM MEASUREMENTS Jacek Klucznik, Krzysztof Dobrzyński Zbigniew Lubośny, Robert Trębski

38

EXPERIMENTAL INVESTIGATIONS ON THE USEFULNESS OF AIR SOLAR COLLECTORS SUPPORTING THE DRYING PROCESSES Jerzy Majewski

46

NEW ALGORITHM FOR CONTROLLING TRANSFORMERS THAT SUPPLY DISTRIBUTION NETWORK Robert Małkowski, Zbigniew Szczerba

52

IMPACT OF UPFC OPERATIONAL MODES ON POWER SYSTEM STATE ESTIMATION Tomasz Okoń, Kazimierz Wilkosz

60

SETTING METHODS OF THE IMPEDANCE TYPE POWER SWING BLOCKING FUNCTIONS APPLIED IN DISTANCE PROTECTIONS Adam Smolarczyk

68

TOWARDS SMART GIRD – PILOT PROJECT “SMART PENINSULA” Grzegorz Widelski, Sławomir Noske

Electric power system is a group of interconnected devices whose task is the production, transmission and distribution of electricity, i.e. supplying electricity with proper quality parameters to consumers in a reliable manner.

The main parameters determining the electricity quality are frequency and voltage. Therefore, the basic processes of control and adjustment implemented in the power system are the following: â&#x20AC;˘ frequency adjustment process, in which frequency is the controlled parameter, and changes in active power load are the nuisance parameter; the purpose of control is to balance active power by changing its generation (production) â&#x20AC;˘ voltage adjustment processes, in which voltages at the system nodes are the controlled parameters, and changes in the reactive power load are the nuisance parameter; the purpose of control is to balance reactive power. In practice this objective is achieved by maintaining the voltage at the power system nodes within acceptable limits.

From a geographical point of view, the power system is a vast system. At the same time, from the point of view of the control system structure, it is a hierarchical and multi-level system. It is characterized by a large number of elements subject to control of various importance for the functioning of the entire system.

The frequency control process is usually coordinated by a central control unit located on the highest level of the control system, and all power plants participate in it, directly or indirectly (including traditional power plants and wind farms). Executive elements in this process are turbine control units.

On the one hand, voltage level control process may be described as a power system coordinated at all levels of control, from power plants to energy reception. On the other hand, voltage adjustment in the power system is of a distributed nature, due to the local impact of changes in voltage. The following system elements are used to adjust voltages: synchronous generators, transformers and autotransformers in extra high, high, medium and low voltage networks; compensating reactors (in high-voltage networks); capacitor banks (in medium-voltage networks) and individual capacitors for certain types of reception; the so-called Flexible AC Transmission Systems (FACTS); high-voltage, direct current (HVDC) systems; asynchronous machinery with double fed asynchronous generators in wind turbines (when a voltage control algorithm is used).

The above show the complexity of control and adjustment processes implemented in the power system.

This complexity will increase with the growing saturation of the systems with the so-called distributed energy sources.

This issue is dedicated to selected issues related to power system control. I invite you to read it.

Zbigniew LuboĹ&#x203A;ny Editor-in-Chief of Acta Energetica

4

Krzysztof Glik, Ryszard Kowalik / Warsaw University of Technology Désiré Dauphin Rasolomampionona / Warsaw University of Technology

Authors / Biographies

Krzysztof Glik Warsaw / Poland

Ryszard Kowalik Warsaw / Poland

Graduated from the Faculty of Electrical Engineering at Warsaw University of Technology (2009).

He is currently a PhD student at the Institute of Electrical Power Engineering, Warsaw University of Technology and works in EDC Poland. His main professional interests concern the travelling wave fault location in HV lines and underwater systems powering oil rigs.

He has worked at the Institute of Electrical Power Engineering, Warsaw University of Technology since 1989. He is a co-author of modern digital protection and microprocessor technology laboratories launched at the Division of Electrical Power Protection and Control. His professional interests concern the power of automation equipment, synchronization systems and telecommunications systems.

Désiré Rasolomampionona Warsaw / Poland Graduate of Warsaw University of Technology. He has been working at the Faculty of Electrical Engineering at the Institute of Electrical Power Engineering since 1994. He is currently the head of the Division of Electrical Power Protection and Control.

His research interests are focused mainly on the issues of electrical power control and protection, control of the power system operation and the use of telecommunications and modern information technologies in electrical power engineering.

Travelling Wave Fault Location in Hv Lines

TRAVELLING WAVE FAULT LOCATION IN HV LINES Krzysztof Glik / Warsaw University of Technology Ryszard Kowalik / Warsaw University of Technology Désiré Dauphin Rasolomampionona / Warsaw University of Technology

1. INTRODUCTION Determination of fault location in high-voltage lines is one of the most important issues which the protection services have to deal with.

Determination of fault location may be used for the proper operation of protection equipment or for inspection and repair purposes. In the first case, it is very important to find the fault location quickly, whereas the accuracy may be limited only to determining the area of protection operation. Determining the location for inspection and repair purposes must have a high degree of accuracy. It is carried out through the fault location function implemented in a protection device, interference recorder or by a separate locator.

Accurate determination of the fault location for inspection and repair purposes enables: • faster restoration of the line to operation • preventing permanent faults • verification of protection operation.

Faster restoration of the line to operation is a result of more efficient work of energy services, who, having accurate information about the distance to the fault location, may quickly locate it even in mountainous or forested areas.

Most of the faults occurring in high-voltage lines are transient faults. Precise determination of the location of these faults enables carrying out preventive works (e.g. replacement of insulators, trimming of trees) in order to prevent permanent faults.

The use of information on the designated distance to verify the protection operation is based on confirming that the protection is operational in the relevant zone, in a designated location in the case of fault simulation.

2. COMPARISON OF IMPEDANCE AND WAVE DETERMINATION OF FAULT LOCATION Among the used locators two are of the greatest importance: impedance and wave locators. Impedance locators may be a part of a protection device, interference recorder, or constitute a separate device, similar to wave locators.

Operation of impedance locators is based on the measurement of current and voltage during fault. Due to the use of these two electrical values in determining the fault location, we deal with a measurement that is characterised by errors resulting from multiple factors, such as: • transient components in current • current distortions caused by core saturation in current transformers • pre-charge current in the line immediately before the occurrence of a fault • transition resistance at the fault location • capacitance to earth of the line • magnetic coupling between the channels in dual lines

Abstract The article compares wave and impedance methods used for determining fault location in high voltage lines, and presents the basic issues related to the exami-

nation of wave phenomena, describes the wave measurement methods and particular elements included in the measurement systems of wave locators.

5

6

Krzysztof Glik, Ryszard Kowalik / Warsaw University of Technology Désiré Dauphin Rasolomampionona / Warsaw University of Technology

• inaccuracies in the data concerning the line impedance, particularly inaccurate determination of the zero line impedance due to the change in ground resistance along the line • the phenomenon of current flow at the connection point of a tapped line in branched lines.

Thanks to many years of operation of impedance-based fault locators, there are methods that reduce or eliminate the effect of particular factors on the accuracy of measurement. Nevertheless, the accuracy of determination of fault location using impedance locators is in the range of 1-20 percent. The lower limit of error refers to metallic faults, determined at both ends of the line, while the upper limit occurs in the case of long lines, usually series-compensated. Error in determining fault location using the function in Siemens protection 7SA522 is declared for certain conditions at 2.5 percent of the line length. Such an accuracy is insufficient, given that transmission lines often have a length of hundreds of kilometres in environmentally diverse areas. Locating the particular damage by the operating staff in such conditions may result in too long a break in the transmission of electricity.

Wave fault locators measure time instead of measuring current and voltage. This way, the effect of many of the above-mentioned factors on the measurement error is eliminated.

However, wave fault locators also have their drawbacks. The main factors that affect the error in determining the distance to a fault location in such locators are the following: • small fault angles • faults close to locator installation points • device synchronization error • ill-defined wave propagation velocity in the line • travelling wave detection error.

The term “small fault angles” refers to a situation in which a fault occurs when the instantaneous voltage value is close to zero, which prevents fault detection due to the low value of amplitude of the formed electromagnetic wave. A sudden change in voltage is required for voltage and current wave with a high amplitude to appear in a high-voltage line, which in this case is not possible. This issue can be eliminated by a simultaneous determination of the fault location using a wave and impedance locator, with the latter responsible for locating the fault occurring at a small angle.

The error associated with a fault close to the locator installation point, which causes multiple wave reflections between the locator installation point and fault location, may be eliminated by applying a sufficiently high sampling frequency.

Device synchronization error occurs for fault location determination using measurements on the two ends (type D location). This error is typically ±1 μs, which is associated with uncertainty in determining the distances of ±150 m for a single locator.

Wave propagation velocity in line is one of the values used to calculate the distance to the location. It depends on the line parameters and the path of electromagnetic wave – conductors (no-ground fault) or conductors and ground (ground fault).

Travelling wave detection error is associated with the reduction of amplitude and lengthening of the wave moving in the line. If a fault occurs closer to station A than station B, then due to the greater extension of the wave front reaching station B, detection of the wave in this station occurs later, causing additional error.

Wave locators are characterized by accurate determination of fault location in the range of 150-500 m, regardless of the line length. Such an accuracy applies also to long lines that are series-compensated, multicircuit lines with cable sections, and direct current lines.

High accuracy in determining the distance to the fault location and increase in network reliability, as well as cost savings resulting from the use of wave fault locators, made them widely used in such countries as the U.S., China, South Africa, Scotland and Canada. The national power system uses LAS-type wave locators, produced by the Institute of Power Systems Automation Ltd, a Wroclaw-based Company, and TWS-type locator produced by Qualitrol.

The basic issues associated with the operation of wave fault locators are described below.

Travelling Wave Fault Location in Hv Lines

3. WAVE PHENOMENA Of all the transient states that occur in the power system, wave phenomena in HV lines are characterized by the shortest duration, ranging from microseconds to milliseconds.

Wave phenomena are related to the propagation of electromagnetic waves, resulting from: a fault occurring in power lines, atmospheric discharges or switching operations in the grid. A sudden and significant change in voltage, in at least one location of HV line (Fig. 1) leads to the initiation of an electromagnetic wave, which propagates in opposite directions from that point.

Fig. 1. Propagation of electromagnetic wave due to a fault

An electromagnetic wave can be divided into a voltage wave associated with the phenomena occurring in the electric field, and a current wave associated with the magnetic field. An important feature of such a wave is the movement of specific values of voltage and current with a finite speed along the line.

The use of wave phenomena in determining fault location requires consideration of many theoretical issues, such as: • wave propagation velocity in the line • power line model with distributed parameters • wave attenuation and distortion • wave passage and reflection • diagonalizing transformations • wavelet transform.

The accuracy of determining fault location using the wave phenomena depends on the correct estimation of wave propagation velocity in the particular power line. This velocity depends on the power line parameters, which change with the change of environment temperature, conductor surface contamination or icing. Wave propagation velocity also depends on the path of electromagnetic wave movement, and thus it is determined separately for each line, for ground faults and faults not affected by the ground. Aerial-mode propagation velocity is approximately v = 295 000 km/s, while ground-mode propagation velocity is approximately v = 188 000 km/s. When installing a wave locator the wave propagation velocity is determined forcing the movement of the travelling wave in the power line by switching capacitors or circuit breaker.

Circuit with distributed parameters is characterized mainly by the fact that the signal appearing at the system input requires a certain specified time to appear at its output. These circuits are described by partial differential equations. Voltages and currents in such a circuit are a function of two variables – time t and location x.

Power lines cannot be considered as circuits with concentrated parameters when their length l [m] is commensurate with the length of wave λ = v/f [m] occurring in this line. Transmission lines that operate at a frequency of 50 Hz and are shorter than 6000 km are modelled as circuits with concentrated parameters. However, if the signal frequency increases, e.g. to 100 kHz, a 3-kilometre line should be treated as a circuit with distributed parameters.

Wave attenuation and distortion causes the reduction of the wave amplitude and wavelength as a result of its movement in the line. This is associated with energy loss in resistances of conductors or conductors and ground, loading of insulator capacity and the escape of.

Wave passage and reflection is also the cause of attenuation and distortion of travelling waves at the points of wave impedance change. Wave impedance in the line is determined by the ratio between the amplitude of voltage and current of wave running in this line. Usually its value is in the range of 200 – 400 Ω and is dependent – mainly on the voltage level of the line. When the incoming wave encounters the point of wave impedance change, called a node, a part of the wave energy is reflected from that point, and a part moves further.

7

Krzysztof Glik, Ryszard Kowalik / Warsaw University of Technology Désiré Dauphin Rasolomampionona / Warsaw University of Technology

8

Diagonalizing transformations are used in order to consider three-phase lines as three separate singlephase lines without mutual magnetic couplings. Theoretically, there is an infinite number of diagonalizing transformations, the most common of which is the symmetrical components method. However, in the case of wave phenomenon analysis, such a transformation is not used, which results from the nature of wave phenomena described by instantaneous values of voltages and currents, which cannot be converted to compatible, negative and zero component.

Transformation matrices that consist of elements which are not complex numbers (as in the case of the transformation of symmetrical components) are used.

Wavelet transform is used to analyse non-stationary signals, i.e. signals whose statistical characteristics (mean value, mean square value, correlation function) are time functions (they depend on the choice of baseline).

One of the most important features of wavelet transform is the ability to determine the time at which a high frequency signal occurred, at the same time examining the components of a low frequency signal.

4. MEASUREMENT METHODS Depending on the measuring method used, wave fault locators are divided into five types: A, B, C, D and E. The operation of each type of locator is based on an analysis of the incoming electromagnetic wave caused by the fault. The types of locators are described below.

A-type locators A-type locators perform measurements on one side of the line. The distance to the fault location is calculated by measuring the time between the moment when the first wave, generated at the fault location, reaches the locator, and the second moment when the wave reflected from the fault location reaches the locator. The electromagnetic wave is entirely reflected from the fault location if the occurring fault angle has a resistance less than the wave impedance of the line. The examined network system and course of travelling waves is shown in Fig. 2.

Station A

Station B

Waves moving from the fault location

Fig. 2. The use of an A-type wave locator

The distance to the fault location from station A results from the following dependence: D

t3  t1 v 2

(1)

where: D – distance to fault location [m] t, – time in which the first wave generated at fault location reaches station A [s] t3 – time in which the wave reflected from fault location reaches station A [s] v – wave propagation velocity [m/s].

The error in measuring the distance to the fault location using method A is affected by such factors as short duration of fault arc, the transition resistance, branching and taps in the line and the difficulty in identifying the appropriate wave. These errors are eliminated using method D.

Travelling Wave Fault Location in Hv Lines

B-type locators B-type locators perform measurements on both sides of the line. The wave created at the fault location runs towards stations A and B. The arrival of the first wave of several microseconds to station A activates the timer. The timer is disabled in station A when a signal from the device installed in station B is sent, when the wave running from the fault location is detected in the device. The examined network system and course of travelling waves is shown in Fig. 3.

Station A

Station B

Fig. 3. The use of a B-type wave locator

The calculation of fault location distance is similar to measuring method D, and the calculations must take into account the delay associated with the transmission of the signal from station B to station A, which stops the timer.

C -type locators C -type locators perform measurements on one side of the line. The locator sends a pulse to the line where the interference occurred. The distance to the fault location is calculated using the time difference between the moment of sending the pulse and the time when the device receives the wave reflected from the fault arc. The examined network system and course of travelling waves is shown in Fig. 4.

Station A

Station B

Pulse sent to the line Pulse generator

Fig. 4. The use of a C -type wave locator

The distance to the fault location from station A results from the following dependence: D

t2  t1 v 2

(2)

where: t 1 – time of sending the pulse by the generator [s] t2 – time in which the wave reflected from fault location reaches station A [s] It should be noted that the current use of this type of locators encounters difficulties associated with the correct coupling of the pulse generator with the power line and its high price.

9

Krzysztof Glik, Ryszard Kowalik / Warsaw University of Technology Désiré Dauphin Rasolomampionona / Warsaw University of Technology

10

D-type locators D-type locators perform measurements on both sides of the line. Waves generated at a fault location run towards stations A and B, and reach them within several microseconds. For a correct determination of the fault location, a D-type locator requires the use of two devices synchronized with each other in time (e.g. by means of GPS), installed on two ends of the line. The locator determines the moment in which the wave is coming to stations A and station B, then they are used to calculate the distance from fault location. The examined network system and course of travelling waves is shown in Fig. 5.

GPS Station A

Station B

Fig. 5. The use of a D-type wave locator

The distance to the fault location from station A results from the following dependence: D

L  (t A  t B )  v 2

(3)

where: t1 – time in which the first wave generated at fault location reaches station A [s] t3 – time in which the first wave generated at fault location reaches station B [s] L – line length [m].

The accuracy of D-type locators is not reduced by a short duration of a fault or branching in the line. The subsequent reflections of the wave at the points of wave impedance change do not affect determination of the distance to interference location. The main error in calculating the distance to the fault location is the synchronization error.

It should be noted that D-type locators are resistant to the factors mentioned earlier in this article, which prevent the correct determination of fault location or introduce an additional error in A-type locators.

E-type locators E-type locators perform measurements on one side of the line. For this purpose they use the wave induced by the breaker on the line. In its operation the E-type method is similar to the pulse method used in determining a fault location in cables. The breaker switching the HV line can be treated as three separate pulse generators. Voltages in switched phases have a different amplitude and phase shift, which results from switching each breaker pole in a minimally different time. The time difference between the pulse generated by switching and the pulse reflected from fault location is used to determine the distance to the fault location. The operation of this type of locator is shown in Fig. 6.

Travelling Wave Fault Location in Hv Lines

Station A

Station B

Fig. 6. The use of an E-type wave locator

The distance to the fault location from station A is calculated based on the following dependence: D

t2  t1 v 2

(4)

where: D – distance to fault location [m] t1 – time in which the wave is generated as a result of switching [s] t2 – time in which the reflected wave reaches station A [s] v – wave propagation velocity [m/s].

E-type locators can be used for detection and location of the interrupted line cord. In addition, this method may be used to check whether the electrical length of the operating line corresponds to the line length measured using another method. Such a procedure is based on switching off the line breaker and then measuring the time in which the reflected wave returns to the locator. The known line length is compared with the measured time of the reflected wave movement.

In the latest wave fault locator solutions, fault locations are applied simultaneously in types A, D and the new type E. They use current waves in their operation. The D-type method is usually the basic method of measurement used in wave locators. Methods A and E are added to method D, which, as a result of operating experience, has proved to be reliable and accurate.

5. DESCRIPTION OF ELEMENTS IN MEASUREMENT SYSTEMS OF WAVE LOCATORS Due to the nature of wave phenomena it is worth describing various key elements included in the measurement systems of wave fault locators, i.e.: • current and voltage transformers • digital signal processing systems • satellite navigation systems.

Initially, voltage transformers were used in capturing travelling waves; however, due to unsatisfactory transfer characteristics of these transformers, current transformers are mainly used. Fault location is determined using protective current transformers, which successfully carry signals with a frequency of up to 100 kHz. The most common solution is a system in which a protective current transformer is used as the main transformer, and a current transformer with an open core is used as an intermediary transformer.

Wave fault locators need appropriate systems, which are able to receive and analyse large amounts of data, and distinguish between relevant waveforms coming to the device. A wave fault locator requires the use of a data collection unit with a sampling frequency higher than or equal to 1 MHz, which is far more than in conventional types of protection. The higher the sampling frequency of the input signal, the more accurate the result. On the other hand, an increased number of samples per period increases the processor load and requires more memory for data storage.

Additionally, a key problem is caused by noises in the measured signal.

Time synchronization of locators installed on both ends of the line is done using GPS. The time synchronization error is 1 μs, which corresponds to the error in determining the distance to the fault location of ±150 m for a single

11

12

Krzysztof Glik, Ryszard Kowalik / Warsaw University of Technology Désiré Dauphin Rasolomampionona / Warsaw University of Technology locator. The European satellite navigation system Galileo may reduce this type of error. GPS receivers have an accuracy of several metres, whereas the accuracy error in Galileo will be less than 1 metre in 2012.

6. SUMMARY The use of wave locators in high-voltage lines enables more accurate determination of a fault location in comparison with impedance locators. Operational experience gained from many countries shows the high accuracy of wave locators in the case of various interferences (e.g. high transition resistance) and use for various types of lines (lines compensated in series, long lines, multicircuit lines with cable sections).

REFERENCES 1. Gale P.F., Taylor P. V., Naidoo P. , Hitchin C., Clowes D., Travelling wave fault locator experience on Eskom’s transmission network, Seventh International Conference on Developments in Power System Protection (IEE) April 2001, pp. 327–330.

2. Siemens: 7SA522 distance protection relay for transmission lines. Catalogue 2009.

3. Lee H., Mousa A.M., GPS travelling wave fault locator systems: investigation into the anomalous measurements related to lightning strikes, IEEE Transactions on Power Delivery, volume 11, issue 3, July 1996, pp. 1214–1223.

4. Christopoulos C., Wright A., Electrical Power System Protection, Kluwer Academic Publishers, Dordrecht 1999.

5. Flisowski Z., Technika wysokich napięć, WNT, Warszawa 2005.

6. IEEE Guide for Determining Fault Location on AC Transmission and Distribution Lines, IEEE Std C37.114™- 2004.

7. Samper J.M., Lagunilla J.M., Perez R.B., GPS and Galileo: Dual RF Front-end receiver and Design, Fabrication, And Test (Communication Engineering), McGraw-Hill Professional, 2008.

8. Gale B. Y. Su. P.F., Ge Y.Z., Fault location based on fault induced current tramients, International Conference on New Development in Power System Protection & Local Contral, Beijing China, May 25–28, 1994, pp. 377–381.

9. Redfern M.A., Terry S.C., Robinson F. V.P., The application of distribution system current transformers for high frequency transient based protection, Eighth IEE International Conference on Developments in Power System Protection, volume 1, 5–8 2004, pp. 108–111.

14

Irina I. Golub / Melentiev Energy Systems Institute of Russian Academy of Sciences

Authors / Biographies

Irina I. Golub Irkutsk / Russia Graduated from the Moscow Power Institute as an engineer-electrician. She has been working at Energy Systems Institute (ESI) Irkutsk since 1972. Her scientiďŹ c interests are connected with real-time control problems, especially in the ďŹ eld of observability of electric power systems and allocation of measurements of operating parameters of electric networks. She is a leading researcher, Univ.-Prof., Doctor of technical sciences.

Algorithms to Ensure Reliability of Power System Observability

ALGORITHMS TO ENSURE RELIABILITY OF POWER SYSTEM OBSERVABILITY Irina I. Golub / Melentiev Energy Systems Institute of Russian Academy of Sciences

1. INTRODUCTION A notion of topological observability of electric power system (EPS) that determines the existence of solution to the problem of load flow calculation on the basis of measurements was introduced in [1]. The notion is based on a relationship between the rank of observability matrix and its structure which depends on the network topology, set and allocation of measurements. Virtually simultaneously the first algorithms for analysis of topological observability were suggested in [2] and [3].

The algorithm for analysis of topological observability [4, 5] which is made independently for active and reactive EPS models is based on construction of a measurement tree on the network graph. Each branch of the graph is connected with one of measurements available in the network. Active (reactive) model is observable if it corresponds to a spanning tree of measurements in which a node with fixed voltage phase (measured magnitude) is a root node. With several voltage measurements reactive model of EPS will be observable if the measurements on the network graph can be used to construct subsystems of measurement trees with measured voltages at the root nodes that in the aggregate cover all nodes of the network graph. Similar condition provides observability of an active model that contains several measurements of voltage phases.

Should the observability analysis reveal that the measurement tree is disconnected or includes not all nodes of the network graph or subsystems of measurement trees without measured voltage magnitude (phase) are detected, EPS is unobservable. For the EPS to become observable it is necessary to add measurements that make it possible to connect separate subsystems of measurement trees, add the nodes that originally did not belong to the tree, or specify the root nodes with a measured voltage magnitude (phase).

In order to ensure reliability of observability when some measurement devices and remote terminal units (RTU) fail, and when the network topology changes the additional measurements are needed. One of possible solutions to the first problem was given in [6] where the authors suggested adding the redundant measurements in each independent loop and branch of the scheme to provide reliability of observability. An original approach to solution of the second problem with the method of integer linear programming was proposed in [7].

Topicality of the observability study is confirmed by annually published new algorithms for analysis and synthesis of topological, algebraic and nonlinear observability. An increased interest in observability is connected with the emergence of new technology of synchronous vector measurements, which is called Phasor Measurement Units (PMU).

The measurement system on the basis of PMU including measurements of voltage vector at the node of PMU placement and current vectors in the adjacent tie lines is intended to provide complete observability of EPS, and is rather costly. The second structure of PMU measurements that contains measurements of voltage vector and current vector in tie line is most appropriate for practical purposes.

The research aims to develop the algorithms for determining the minimum number of PMUs whose measurements can guarantee EPS observability in the normal operating conditions (Problem 1), when some tie lines are disconnected (Problem 2) and when some PMUs fail (Problem 3). The indicated problems are solved for two types of PMU, with measurements available in EPS and without them.

Abstract The algorithms for choosing the minimum number of PMUs of two types have been developed. The PMUs are to be placed at nodes and tie lines in power systems to provide topological observability of power systems under normal operating conditions and ensure reliability of topological observability in case of failure of individual

tie lines and PMUs, in the absence and presence of conventional measurements. The developed algorithms are implemented in the MATLAB environment. Their efficiency has been proved by computations for a large number of test IEEE schemes and real power systems.

15

Irina I. Golub / Melentiev Energy Systems Institute of Russian Academy of Sciences

16

An important study related to the choice of an optimal set of PMUs to maintain observability in the normal operating conditions of EPS and when some tie lines are disconnected is work [8]. The authors of [8] employ the binary search method in combination with the conventional algorithm for analysis of topological observability.

In [9] the choice of the minimum number of PMUs, Problem 1, with and without conventional measurements in EPS is made with the method of integer linear programming. This approach was further developed in [10] which suggested ensuring observability when some PMUs fail, and choosing an optimal set of measurements with account taken of the PMU price. One of the first studies to similarly take into account the price of remote terminal units was [7]. In the absence of conventional measurements in EPS the algorithm [10] allows one to determine the optimal set of PMUs to solve Problems 1 and 3. However, in cases where the conventional measurements of power flow and/or zero injections are available it can result in a solution that contains redundant PMUs.

2. ALGORITHMS FOR CHOOSING AN OPTIMAL SET OF PMUS TO PROVIDE EPS OBSERVABILITY In order to choose the optimal set of PMUs when solving Problems 1-3 we use the method of integer linear programming. The method implements a simplex algorithm that, on the basis of some basic solution, generates another basic solution that has a better value of objective function as compared to its initial value. This procedure is repeated until the basic solution that meets the optimality conditions is obtained. In the beginning each of the problems is solved for the case without conventional measurements of flows and zero injections, and then – for the case with them.

With conventional measurements available, unlike algorithms in [9-11] that also employ the procedure of integer linear programming, the initial matrices are corrected without changing their size. The rules for a posteriori analysis of the solution obtained are suggested. They make it possible to detect redundant PMUs in the solution, if there are any. The algorithms also allow one to take into account the information on nodes, at which PMU placement is necessary for technical or economic reasons, and the nodes, at which PMU placement is irrational or impossible, for example at dangling or transit nodes.

The choice of an optimal number of PMUs for the case of unavailable conventional measurements and zero injections. Most simply the minimum number of PMUs in all the three problems is chosen on the assumption that observability should be provided only by PMU measurements.

The first type of PMU. In order to choose the minimum number of PMUs that will make EPS observable under normal operating conditions we solve the problem of integer linear programming (Problem 1) [9]

min f T x x

Ax  g

(1)

where A – n × n asymmetrical adjacency matrix that consists of 0 and 1, n – the number of nodes, f and g – the unit vectors, vector of solution x – a binary integer vector, with its elements equal to 0 or 1.

Solution (1) allows observability of all nodes in the calculated scheme of EPS to be provided at least once.

Replacement of the unit vector g by the vector whose elements are equal to 2 will provide observability of all nodes in the calculated scheme at least twice (Problem 3). From this condition it follows that when some PMU fails power system will remain observable since measurements of each PMU form a group of noncritical measurements.

In order to choose the minimum number of PMUs which will enable EPS to maintain observability when some tie lines are disconnected it is necessary to solve the integer linear programming problem suggested in [7] (Problem 2) min f T x x

MTx  g

(2)

Algorithms to Ensure Reliability of Power System Observability

where M T – an m × n transposed incidence matrix for undirected graph, m – the number of tie lines, n – the number of nodes in the network graph. Elements of the row of matrix M T, that correspond to the nodes of tie line i-j, are equal to unity, the remaining elements of the row are equal to zero, f and g – the unit vectors, vector of solution x – a binary integer vector whose elements equal 0 or 1.

Solution (2) provides observability of nodes in the calculated scheme that does not have PMUs at least twice, except for observability of dangling nodes. After disconnection of some tie lines in the electric network observability is not lost and each of measurements entering in PMU is noncritical, except for the current vector measurement in a dangling tie line.

Reliability of observability at switchings is particularly important for distribution networks whose topology is subject to frequent changes. Problem 2 should be solved only for the distribution network part in which backup supply can be used in case of electric tie disconnection. In part of the network being a tree, provision of observability at disconnection of some tie lines makes no sense.

Requirement for one-time control of all tie lines in the electric network which is implied by formulation (2) in the general case can lead to a redundant solution.

In some cases a posteriori analysis of solution (2) makes it possible to sufficiently simply identify redundant PMUs. After a set of PMUs is chosen we determine the number of PMUs required for each node to provide its observability. If observability of node i with PMU and all nodes that have no PMU and are adjacent to node i is provided by two more measurements, the PMU at node i is considered as redundant.

On the other hand there can be situations where solution (2) may turn out to be nonoptimal and it will be impossible to determine redundant PMUs in it, whereas some other solution, obtained, for example, by forced placement of PMU at one of the nodes, will enable one to find an optimal solution. For the forced placement of PMU at node i the columns of matrix M T that correspond to the nodes adjacent to node i should be zeroed.

A posteriori analysis of solution (2) makes it possible to choose PMUs that should be added to solve Problem 3. Firstly, additional PMUs should be added at all dangling nodes. Secondly, to provide reliability of observability when critical PMU fails causing loss of observability of a node where the PMU is placed, it is necessary to add PMU at a node adjacent to the largest number of nodes with critical PMUs. The second type of PMU. To choose the minimum number of PMUs of the second type which provide observability of EPS under normal operating conditions (Problem 1), it is necessary to solve the problem [11]

min f T x x

Mx  g

(3)

where M – the incidence matrix for undirected graph, the length of unit vectors f and g is equal to the number of tie lines and the number of nodes in the network scheme, respectively.

If elements of vector g in expression (3) are specified equal to 2 it is possible to determine the set of PMUs which will make it possible to maintain observability both when any of the tie lines is disconnected (Problem 2), and when any of PMUs fails (Problem 3). Indeed, it is always possible to allocate measurements so that there will be no doubling measurements of voltage vectors at one and the same node. In this case all measurements will be noncritical, and Problem 2 and Problem 3 will be solved simultaneously. The dangling tie line in Problem 3 should be simulated as a tie with two parallel lines.

The choice of an optimal number of PMUs for the case of available conventional measurements.

EPS has a great number of zero injections and conventional measurements of nodal voltage magnitudes, active and reactive power flows and nodal powers. Nevertheless, many power systems are incompletely observable.

Observability can be provided by both conventional measurements and PMU measurements.

Let us consider possible approaches to solution of Problems 1-3 for choosing the minimum number of PMUs of the first and second types that should be added to zero injections and power flow measurements to make EPS completely observable. In order to take into account the conventional measurements it is necessary to correct matrix elements in expressions (1)-(3) and/or vector g that correspond to these measurements.

Consideration of power flow measurements. The power flow measurement in tie line i-j (Problems 1 and 3) is considered in matrix A by placing unity into row i (j) of the adjacency matrix in the columns corresponding to the nodes adjacent to node j (i).

17

18

Irina I. Golub / Melentiev Energy Systems Institute of Russian Academy of Sciences

In this case the incidence matrix M is corrected by adding unity to the row corresponding to node i (j) in the columns corresponding to the tie lines, in which node j (i) is one of their nodes.

A kind of integration of nodes i and j into one node takes place, the i-th and j-th rows of matrices A and M become identical and represent the rows of the adjacency matrix and the incidence matrix for integration of nodes i and j.

In Problem 2 the power flow measurement in tie line i-j is taken into account by zeroing the row elements of matrix M T and the element of vector g that correspond to this tie line.

Consideration of zero injections. In Problems 1 and 3 matrices A and M can be formed by using two approaches to consideration of zero injections.

In the first approach the zero injections may be replaced by power flow measurements in one of the tie lines adjacent to zero injections. Such a tie line can be either determined on the basis of analysis of the measurement tree or specified arbitrarily.

If the choice of correspondence between the zero injection and the power flow replacing it is unsuccessful, the global optimum is not always found. Replacement of one tie line associated with the injection by the other tie line can improve the solution. However, the necessity to perform such replacement makes this approach irrational. At the same time if there is a dangling tie line among the tie lines adjacent to the injection, the injection should be put in correspondence only with this tie line.

In the second approach it is assumed that the zero injection at node i provides its observability. This property is taken into consideration by setting the elements of vector g, corresponding to the i-th rows of matrices A and M, equal to zero.

In this case the chosen set of PMUs may turn out to be redundant, which will be revealed in a posteriori analysis of the solution obtained.

For example, when solving Problem 1 for node i with PMU of the first type it is checked whether there are adjacent nodes without PMUs and zero injections, whose observability is provided only by PMU at node i. Should there be no such nodes, but there is a zero injection at one of the nodes adjacent to node i, whose observability is provided at least by one PMU, such injection is put in correspondence with node i and PMU at this node is considered as redundant.

PMU at node i with zero injection is also redundant, if observability of this node is provided by no less than two PMUs and if it has no more than one adjacent node j without PMU, whose observability is provided only by PMU at node i. After removal of the redundant PMU at node i the zero injection of this node is used to provide observability of node ј.

Some other scenarios for the emergence of redundant PMUs are also analyzed. In all cases they are detected by using the procedure of searching for a tie line that can be placed in correspondence with the zero injection which is similar to the procedure of searching for the maximum matching on the bichromatic graph [2].

In Problem 2 the zero injections are taken into consideration by correction of matrix M T and the vector g. The following situations are possible here: 1. Zero injection at node i, that is adjacent to dangling node j is used to provide observability of node j, and the row elements of matrix M T and the element of vector g corresponding to tie line i-j are set to zero.

2. All elements of the column of matrix M T that corresponds to node i with the zero injection, and the elements of vector g that correspond to tie lines i-k are set to zero, if the degree of node k is higher than or equal to 2. If the degree of nodes i and k is 2 and the injection at node k is not equal to zero, the element of vector g is not zeroed. Removal of the nodes with degree 2 with zero injection from the scheme, as it was done in [8], makes it possible not to check the last condition.

Application of the software of topological observability analysis to choose an optimal set of measurements for Problem 1. The procedure of integer linear programming is valid for determination of the optimal set of PMUs in power systems that do not have many conventional measurements or use only PMU measurements.

In a partially unobservable EPS the adjacency matrix can be constructed only for part of the network that does not include the observable nodes connected only with observable nodes. After removal of such observable

Algorithms to Ensure Reliability of Power System Observability

nodes and tie lines among the observable nodes it is necessary to take into account the fact that the remaining nodes in the network are observable. This procedure can essentially reduce the scope of optimization problem of the number of additional PMUs. The algorithm of topological observability analysis that makes it possible to determine branches of the measurement tree, observable and unobservable subsystems for the considered network topology and a set of conventional measurements can also be used to choose additional PMUs [2]. For this purpose it is necessary to analyze the subsystems of measurement trees and to choose the nodes allowing integration of the maximum number of subsystems of trees as the sites for PMU placement. The procedure of checking observability for the analysis of corrected measurement tree should be repeated, after each regular PMU is chosen. In order to analyze topological observability in this case it is necessary to sort out the conventional measurements and leave only those at which the sets of nodes of incompletely observable subsystems of the active and reactive models coincide. The term “incompletely observable” to a greater extent concerns observability of the active model because there is no voltage phase measurement in subsystem of the kind. The reactive model in this case can be both observable and incompletely observable which depends on the location of voltage magnitude measurements in it. Further, after each additional PMU is chosen, the procedure of observability analysis can be performed on the basis of the software intended for topological observability analysis of the reactive model [4].

3. CASE STUDY Fig.1 presents a 43-node HV distribution network, for which availability of both zero injections and power flow measurements is simulated. The results of choosing the optimal sets of PMUs of the first and second types when solving Problems 1-3 are given in Table 1. All the calculations were carried out by using the software developed in the MATLAB environment. The solutions obtained were tested for availability of missing and redundant measurements on the basis of the software [4].

Fig. 1. A 43-node network with power flow measurements (x) and zero injections at nodes 20-27, 32; the numbers of tie lines are shown in bold

Fig. 2. An optimal set of measurements of a 43-node network without tie lines only with observable nodes

Fig. 2 illustrates the possibility of choosing the optimal set of PMUs that coincides with the set determined in line 2 of Table 1 (Problem 1 in terms of power flow measurements) for the network with removed observable nodes that are connected only with observable nodes. Grey boxes represent the remaining observable nodes without PMUs; grey and colorless circles − observable and unobservable nodes with PMUs installed.

While placement of conventional power flow measurements at the beginning or end of tie lines makes no difference when solving Problems 1 and 2, the situation when power flow measurements belong to one and the

19

Irina I. Golub / Melentiev Energy Systems Institute of Russian Academy of Sciences

20

same RTU can substantially influence the optimal solution to Problem 3. This is due to the fact that the choice of the optimal PMU set requires that observability be provided both at failure of any PMU and at failure of any RTU. Since the adjacency matrix A does not take into account whether the power flow measurements belong to RTU, the solution to Problem 3 should be additionally analyzed when individual PMUs and individual RTUs fail, for example, by using the software of topological observability analysis.

The optimal set of PMUs of the 1st type that is presented in line 10 of Table1 was obtained on the assumption that the power flow measurements belong to RTUs as follows: RTU 1 − 13-14, 13-24,13-38; RTU 2 − 1532; RTU 3 − 16-13, 16-15, 16-23, 16-31, 16-39; RTU 4 − 40-29; RTU 5 − 41-40; 41-42, RTU 6 − 43-20, 43-22; RTU 7− 42-43; RTU 8 − 23-41. If to integrate the measurements of RTU 5 and RTU 8, RTU 6 and RTU 7 it will be necessary to add PMUs at nodes 7 and 8 to the set of PMUs shown in line 10 of Table 1.

At the same time the set of PMUs shown in line 9 of Table1 corresponds to the integration of RTU 5 and RTU 8, RTU 6 and RTU 7. For PMU of the 2nd type Problem 3 for the case of available conventional measurements was not solved.

Table 1. Optimal sets of PMUs when solving Problems 1-3 for the network in Fig.1 Problems

1st type of PMUs – node numbers (2nd type of PMUs – tie line numbers)

1.

Problem 1, without power flow measurements and zero injections

1, 13, 16, 17, 19–27, 32, 41 – 15 PMU (1, 4–6, 9–13, 10–13, 19, 21, 23, 25–27, 30, 33, 37, 42–44, 46, 47 – 23 PMU)

2.

Problem 1, with power flow measurements

1, 17, 19, 20–27, 32 – 12 PMU (1, 4–6, 9–13, 27, 31, 33, 35, 37, 43, 44 – 16 PMU)

3.

Problem 1, with power flow measurements and zero 1, 5, 11, 18, 26, 28, 34 – 7 PMU (1, 6, 12, 14, 15, 21, 27, 31, 35, 3, 44 – 11 PMU) injections

4.

Problem 2, without power flow measurements and zero injections

2, 6, 12, 13, 16–27, 30, 32, 35, 40, 42 – 21 PMU (1–16, 19, 21–23, 25–33, 36–39, 41–47 – 40 PMU)

5.

Problem 2, with power flow measurements

1, 6, 12, 17–27, 31, 32, 34 – 17 PMU (1–3, 6–8, 13–15, 26–32, 36–41, 44 – 23 PMU)

6.

Problem 2, with power flow measurements and zero 1, 5, 12, 17–19, 21, 25, 26, 31, 34 – 11 PMU (1–15, 27–33, 36, 37, 39, 41, 43, 44 – 28 PMU) injections

7.

Problem 3, without power flow measurements and zero injections

1, 3, 4, 6–10, 12, 13, 16, 17, 20–33, 35–39, 41, 42 – 33 PMU (1–3, 4, 5, 6–8, 9–12, 13–15, 19–23, 25–32, 33, 36, 37, 38, 42, 43, 44–47 – 47 PMU)

8.

Problem 3, with power flow measurements

1, 2, 3, 4, 6, 7, 8, 9, 10, 12, 18, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 32, 33, 35, 36, 37, 38, 39 – 28 PMU

9.

Problem 3, with power flow measurements and zero 1, 2, 6, 12, 18, 24, 25, 27–29, 32, 34, 35, 38, 39 – 15 PMU injections

4. CONCLUSION The algorithms and the MATLAM software have been developed for choosing the optimal number of PMUs to provide topological observability in normal conditions of EPS operation and reliability of EPS observability at failure of individual tie lines and individual PMUs for the cases with and without conventional measurements.

The software has been tested on a great number of schemes and real EPSs.

Algorithms to Ensure Reliability of Power System Observability

REFERENCES 1. Gamm A.Z., Methodological problems in state estimation and identification in electric power systems, Problems of estimation and identification in energy systems, Irkutsk: SEI SO AN SSSR, 1974, pp. 29–51 (in Russian).

2. Gamm A.Z., Golub I.I., Kesselman D. Y., Observability of electric power systems, Elektrichestvo, 1975, no. 9, pp. 1–7 (in Russian).

3. Clements K.A. and Wollenberg B.F., An algorithm for observability determination in power system state estimation, Proc. IEEE PES Summer Meeting, San Francisco, CA, Jul. 1975, paper A75 447–3.

4. Golub I.I., Synthesis of the system of information and measurement support for CDCS of EPS, Energetika i transport, 1989, no. 2, pp. 19–27 (in Russian).

5. Gamm A.Z., Golub I.I., Observability of electric power system, Nauka, 1990, pp. 220 (in Russian).

6. Golub I.I., Consideration of reliability at the synthesis of data acquisition systems, Information support of dispatching control in electric power industry, Novosibirsk, Nauka, 1985, s. 169–175 (in Russian).

7. Makletsov A.M., Rusanov A.I., Fedorov D.A., Optimization of check measurements in EPS, Information support of dispatching control in electric power industry, Novosibirsk, Nauka, 1985, pp. 181–184 (in Russian).

8. Chakrabarti S., Kyrakides E., Optimal placement of phasor measurement units for power system observability, IEEE Trans. Power Syst., 2008, vol. 23, no. 3, pp. 1433–1440.

9. Bei Gou, Generalized integer linear programming formulation for optimal PMU placement, IEEE Trans. Power Syst., 2008, vol. 23, no. 3, pp. 1099–1104.

10. Abbasy N.H., Yfnafy M.I., A unified approach for the optimal PMU location for power system state estimation, IEEE Trans. Power Syst., 2009, vol. 24, no. 2, pp. 806–813.

11. Emami R., Abur A., Robus measurement design by placing synchronized phasor measurements on network branches, IEEE Trans. Power Syst., 2009, vol. 25, no. 1, pp. 38–44.

21

22

Andrzej Kąkol, Bogdan Sobczak / Institute of Power Engineering Research Institute, Gdańsk Division Robert Trębski / PSE Operator SA

Authors / Biographies

Andrzej Kąkol Gdańsk / Poland

Bogdan Sobczak Gdańsk / Poland

Graduate of the Faculty of Electrical and Control Engineering, Gdańsk University of Technology where he studied automation and robotics. Currently works at the Team of System Analysis of the Institute of Power Engineering Research Institute, Gdańsk Division. Particularly interested in dynamics and safety systems.

Graduate of the Faculty of Electronics, Gdańsk University of Technology, where he studied automation. Currently works as Head of the Team of System Analysis within the Automation and System Analysis Department of the Institute of Power Engineering Research Institute, Gdańsk Division.

Particularly interested in dynamics and system stability issues.

Robert Trębski Warsaw/ Poland Graduate of the Faculty of Electrical Engineering, Warsaw University of Technology (1993) and Faculty of Management, University of Warsaw (1996).

Since 1993 employee of PSE SA and its successor PSE Operator SA. Focuses his research on the problems of power engineering system modelling for on-line and off-line applications used at the National Control Centre (KDM) and on the performance analyses of the electrical power system.

Investigation of the Impact of Switching a Heavily Loaded Transmission Line on Operation of a Power Plant with Turbine-generator Units

INVESTIGATION OF THE IMPACT OF SWITCHING A HEAVILY LOADED TRANSMISSION LINE ON OPERATION OF A POWER PLANT WITH TURBINE-GENERATOR UNITS Andrzej Kąkol / Institute of Power Engineering Research Institute, Gdańsk Division Bogdan Sobczak / Institute of Power Engineering Research Institute, Gdańsk Division Robert Trębski / PSE Operator SA

1. INTRODUCTION The Polish Power Grid is a part of a large synchronised international grid of mainland Europe – ENTSO-E (formerly UCTE) – one of the largest systems in the world, with a peak demand exceeding 400 GW. In a synchronous power grid transfers of power between individual operators influence not only the operators who participate in transactions but also other systems. In most cases the affected systems are direct neighbours of those directly involved in the transaction. Flows through the systems which do not participate in a power transmission transaction are known as loop flows. The loop flows are an undesirable side effect and their negative impact depends on their magnitude.

Development of the European energy market, as well as development of renewable power sources, result in a significant increase in power flow – and subsequently loop flows – within the power grids managed by mainland European operators. These are observed in eastern and western neighbours of Germany: Benelux countries, Poland, Czech Republic and Slovakia. Such a situation occurs as the excess power generated by the wind farms in the northern part of Germany and western Denmark flows to the consumers located in southern Germany, as well as Austria and Italy – areas which have considerable pump storage capacities. Loop flows caused by the wind power generation in Germany and Denmark have already reached an intensity which threatens the safe operation of power systems neighbouring Germany. Large loop flows considerably limit transmission capacities and also transfer some active and reactive power losses related to the transmission transaction to the countries which do not directly participate in it.

Except for the construction of new transmission lines (which is ruled out due to the long investment timescale in most European countries) the loop flows can be limited by solutions such as re-dispatching conventional power generation units, changing transmission system topology and introducing phase shifters in the grid (in most cases at cross-system lines). In addition, application of solutions which limit loop flows requires coordination between transmission system operators, as otherwise they might just result in transferring problems into neighbouring systems.

Large loop flows also cause some less obvious side effects. Transmission lines heavily loaded by loop flows are difficult to reconnect after an emergency trip due to the high voltage angle difference, especially in the case of lines outgoing from a power station’s switchyard. At the same time a loop flow caused by events in another system is difficult to suppress rapidly. This creates a threat of a snowball overload failure and emergency trips of an increasing number of lines – in extreme cases leading to a general system failure.

This paper presents results of a study carried out by the Gdańsk Division of the Institute of Power Engineering for PSE Operator SA aimed at analysing the dynamic threat for Krajnik substation and Dolna Odra Power

Abstract Development of the European energy market, combined with development of renewable power generation sources result in increased power transfers within the grid, including undesirable loop flows. If such flows occur near power stations on a single line (e.g. exchange line) they might cause some threats related to potential line trips and reconnections. Trip of a heavily loaded line considerably distorts active and reactive power balance

in the area of the power plant in question. Reconnection of a tripped line with significant voltage angle difference causes strong torque variations on the shaft of a turbinegenerator unit. This paper presents results of a simulation of electromagnetic and electromechanical phenomena which occur if a heavily loaded transfer line connected to a switchyard of a large power station with turbine generators is tripped and then reconnected.

23

24

Andrzej Kąkol, Bogdan Sobczak / Institute of Power Engineering Research Institute, Gdańsk Division Robert Trębski / PSE Operator SA

Plant (EDO) caused by high power transfers from Germany. Apart from the impact of tripping and reconnecting the transmission line to Vierraden, the impact of power flows from Germany on the stability of EDO’s generators was also investigated.

2. SCOPE OF RESEARCH Investigation of potential threats for the Krajnik substation and Dolna Odra Power Station caused by high power flows coming into the Polish grid over the 220 kV line Krajnik – Vierraden included: • Determining the relation between the power flow from Vierraden and dynamic stability of large disturbances on the Power Station’s generators, defined by critical time of a near-to-generator three-phase short-circuit • Determining the relation between the power flow from Vierraden and dynamic stability of small disturbances on the Power Station’s generators, defined by a damping coefficient of post-disturbance oscillations • Determining the relation between the power flow from Vierraden on the course of electric transition phenomena following a line trip and reconnection, in particular in the case of the Vierraden line • Determining the relation between the power flow from Vierraden on torsional vibrations of turbine-generator shafts at the Power Station following line trips and reconnections, in particular those of the Vierraden line.

3. KRAJNIK SUBSTATION Krajnik Substation is a substation of the Dolna Odra Power Station. The substation is equipped with 110 kV, 220 kV and 400 kV switchyards coupled by 220/110 kV and 400/220 kV transformers (two units). Four power generation units are connected to the 220 kV switchyard, while three others are connected to the 400 kV switchyard. A doublecircuit transmission line to Vierraden substation goes out of the 220 kV switching station. Analytical research indicates that in conditions of very high wind power output in Germany, when no countermeasures are in effect, the load on this line (power flow from Germany) might even reach levels around 1000 MW (which is above the permissible capacity of the line). Fig 1 shows such a situation with 544 MW incoming flow on each circuit of the line. This makes the power flowing from Germany higher than the combined output of all EDO units operating at that time.

Each EDO unit operating at the time had a generator rated at 270.6 MVA equipped with electromechanical AC excitation systems, digital voltage regulators and system stabilisers.

Fig 1. Voltages and power flows at Krajnik substation.

Investigation of the Impact of Switching a Heavily Loaded Transmission Line on Operation of a Power Plant with Turbine-generator Units

4. MODELS OF THE GRID AND DOLNA ODRA POWER STATION Two types of dynamic models were developed to investigate the influence of high load on the line in question: 1. System dynamics model – using PSLF software 2. Model of fast-changing phenomena dynamics – using PSCAD software.

The System dynamics model contains accurate models of base load (“systemic”) power stations of the Polish, German, Czech, Slovak, Austrian and Hungarian grids. Power generation in other systems of mainland Europe was approximated (without exact specification of power generation stations, using exemplary generic models of generators, turbines and control systems). Wind power generation was modelled in the eastern German grid (operated by 50 Hertz Transmission). Increase of wind generation output combined with changing dispatch of conventional power plants was shown to cause increased loop flows in the Polish system. 1100 MW load on the Krajnik – Vierraden line was achieved at wind power output exceeding 10,000 MW. Such a situation should be seen as an overload of the 220 kV Krajnik – Vierraden lines and might be caused for instance by a trip of a north-south transmission line in Germany, resulting in increased flow through the Polish grid and exceeding the permissible capacity of the 220 kV Krajnik – Vierraden lines.

The Model of fast-changing phenomena dynamics covers the Dolna Odra Power Station (generators, their voltage regulation systems and system stabilisers, and step-up transformers), 220 kV and 400 kV switching stations of the Krajnik substation and transmission lines connected to that substation. Due to the fact that scope of analysis included torsional vibrations of each turbine-generator’s shaft, the shaft itself was modelled as multi-mass object. Dynamics of the Polish grid is represented by a large equivalent generator, while the German system is simulated by a fixed network to which the 220 kV Vierraden switching station is connected.

The multi-mass shaft model takes into account the shaft’s structure, discriminating several masses of high torsional rigidity, connected with elements (couplings) of lower rigidity. Rigid components represent rotors of individual turbine sections (high-pressure, intermediate-pressure and low-pressure), generator rotor, and optionally also exciter rotor. Typically, a multi-mass shaft model consists of four to six bodies. For example, a five-body shaft has five natural frequencies, the lowest of which is the natural frequency of the entire shaft in reference to the power grid. The remaining four values are used for analysing phenomena which potentially might cause and amplify shaft torsional vibrations such as: interaction with turbine regulation systems, interaction of torsional vibrations between shafts of parallel turbine units, interaction with AC/DC converters’ control systems or subsynchronous resonance with the grid compensated with series-connected capacitor banks. Multimasses models are also used to investigate the influence of disturbances caused by grid switching operations on turbine-generator’s shaft.

The model description shows that the PSLF system model is appropriate to investigate phenomena occurring both in Polish and German grids, in particular those occurring in the 50 Hertz Transmission area of responsibility, while the latter model would only simulate phenomena occurring at the Krajnik substation and within the EDO power generation units.

5. RESULTS ACHIEVED WITH THE SYSTEM DYNAMICS MODEL The full scope of investigation as used for determining the stability of a base load power plant of the Polish power grid was carried out using the systemic model.

Angle stability of major disturbances was investigated by comparing critical times of near-to-generator three-phase short-circuits during trips of EDO’s outgoing transmission lines. Additionally, for a short-circuit cleared after 150 ms, the number and duration of voltage losses below 80% of nominal value was recorded. Assumed criteria did not reveal any significant impact of the load on the Krajnik – Vierraden line on the stability of EDO operation for major disturbances.

Angle stability of minor disturbances was determined with Prony’s method, by determining frequencies and damping coefficients for primary modes in post-disturbance variations of active power generated by EDO generators. No impact of the load on the Krajnik – Vierraden line on the angle stability of EDO generations for minor disturbances was identified.

25

26

Andrzej Kąkol, Bogdan Sobczak / Institute of Power Engineering Research Institute, Gdańsk Division Robert Trębski / PSE Operator SA

The influence of switching operations on the Krajnik – Vierraden line was investigated by simulating the following sequence: • 1 sec – trip of the first circuit of the line • 5 sec – trip of the second circuit of the line • 20 sec – reconnection of the first circuit of the line • 24 sec – reconnection of the second circuit of the line for various values of the line load (power incoming from Germany).

Fig. 2 presents variation of the active power of the G5 generator feeding power to the 220 kV switching station. Fig. 3 shows corresponding variations of 220 kV voltage in Krajnik and Vierraden substations when the power flow from Germany is 1100 MW. At this load on the line a trip of both circuits of the Krajnik – Vierraden line causes a serious disturbance in both Polish and east German grids. At the 220 kV Krajnik station there is a sudden voltage surge and considerable oscillations of output of EDO generators connected to the 220 kV station (with a magnitude exceeding 100 MW). Voltage surge might be even higher if other outgoing lines of the 220 kV Krajnik station are disconnected. It is possible that the maximum voltage for the power plant’s 220 kV switching station – which is 245 kV – will be exceeded.

The simulations have revealed that tripping the Krajnik line at high power flow from Germany causes more serious disturbance within the east German grid. Voltage surge at Vierraden is almost twice as high as in Krajnik. Local excess of reactive power would occur in the area of Vierraden, with a simultaneous reactive power deficit in the remaining area (losses due to suddenly increased load on transmission lines). The final effect would depend on the availability of quickly dispatchable sources of reactive power in Germany. In a model where the wind farms do have such properties, the situation is kept under control. In a model with restricted wind power control capabilities, voltage losses occur and lead to tripping of some generation units. Increased demand for reactive power in Germany affects operation of the Polish Turów Power Plant, which experiences an increased reactive power load on its generators, following a trip of the Krajnik line.

Fig 2. Variations of active power of the G5 generator for simulated trip-reconnection sequence of the 220 kV line Krajnik – Vierraden loaded at 1100 MW.

Investigation of the Impact of Switching a Heavily Loaded Transmission Line on Operation of a Power Plant with Turbine-generator Units

Fig 3. Variations of the 220 kV voltage at Krajnik and Vierraden switching stations (the latter showing larger surge) for simulated trip-reconnection sequence of the 220 kV line Krajnik – Vierraden loaded at 1100 MW.

After the line trip, the voltage angle difference between 220 kV switchyards in Krajnik and Vierraden exceeds 60°. At this value it is not possible to switch the line back into operation. The presented active power variations indicate that reconnection of the first circuit is just as major a disturbance for EDO generators feeding power into a 220 kV switching station as a near to generator three-phase short circuit.

The identified threat for the power plant and Krajnik station, as well as possible serious disturbances in the grid of eastern Germany after a trip of the second circuit of the Krajnik – Vierraden line show that the load on that line should be limited to a value which – in the case of a single circuit trip – does not automatically result in an immediate trip of the other circuit due to overload.

6. RESULTS OBTAINED WITH A MODEL OF FAST-CHANGING PHENOMENA Simulated sequences of trip/reconnection of the Krajnik – Vierraden line reveal that considerable torsional vibrations of the turbine units’ shafts occur upon the trip of the second circuit of the line (ultimate line trip) and after reconnection of the first circuit. The magnitude of the vibrations increases along with the load on the line (power flow incoming from Germany). The highest torque occurs between the generator rotor and the low-pressure turbine shaft. When a transmission line load exceeds 1000 MW, the oscillations are comparable to the torque values obtained during simulations of near-to-generator short-circuits. Fig. 4 presents torsional oscillations between the low-pressure turbine shaft and generator shaft of the G5 unit recorded for a simulated trip/reconnection sequence of the 220 kV Krajnik – Vierraden line.

Fig. 4. Torsional vibrations (expressed in relative units of the generator’s nominal power) between the LP turbine rotor and rotor of the G5 generator for simulated trip-reconnection sequence of the 220 kV lines Krajnik – Vierraden loaded at 1100 MW.

27

28

Andrzej Kąkol, Bogdan Sobczak / Institute of Power Engineering Research Institute, Gdańsk Division Robert Trębski / PSE Operator SA

7. IMPACT OF GRID SWITCHING OPERATIONS ON EDO TURBINE-GENERATOR’S SHAFTS Each line switching operation – either scheduled or unscheduled – (e.g. trip due to short-circuit, followed by an autoor manual reconnect attempt) occurring near the power station results in a sudden change of the generator’s electric moment and induces torsional vibrations on the turbine-generator unit’s shaft. Vibration amplitude depends on the angle between voltage vectors on the circuit breaker and on the electrical distance between the breaker and the generator. Torsional oscillations cumulatively influence material fatigue and shorten shaft lifetime – the magnitude of those effects strongly depends on the amplitude values. The shaft of a turbine-generator unit is designed to withstand up to several dozen strong shocks caused, for example, by near-to-generator short-circuits or unsuccessful synchronisation attempts and a very high number of lighter shocks caused by remote or minor disturbances, e.g. line switching at small angle differences. According to the available sources a disturbance which results in a change of a electrical generator output lower than 50% of its nominal active power does not affect the actual lifetime of a shaft [1, 2, 3, 4].

Torsional oscillations have one more significant feature – they have low natural damping. Therefore, two disturbances occurring over a short period could result in much greater oscillations than any of those disturbances would cause separately. For that reason, if switching operations are carried near a power station, it is required to separate them in time (according to the previously quoted references – by at least 10 seconds).

In the case of the Krajnik station and Dolna Odra Power Plant, a loop flow over the 220 kV Krajnik – Vierraden line results in a high load not only on this particular line, but also on other transmission lines connected to the Krajnik substation. Potential scheduled or unscheduled switching of the lines outgoing from that station will thus be carried out at larger voltage angle differences, resulting in increased torsional oscillations. The simulations have revealed that if the transmission from Vierraden reaches 1100 MW, the 50% active power criterion may be kept also when attempting to switch on the lines outgoing from the 400 kV switchyard.

8. CONCLUSION The investigation carried out with a system dynamics model and a model for dynamics of fast-changing phenomena has shown that dynamic threats for the Dolna Odra Power Plant related with power transmission from Vierraden to Krajnik exceeding 700-800 MW result mainly from a possible loss of that connection – a trip of both circuits following potential emergency trip of a single circuit. In such a case trip of the second circuit causes a serious disturbance on the EDO generators and voltage surges at Krajnik and Vierraden substations.

After a trip of both circuits of the Krajnik – Vierraden line reconnection may only be possible after decreasing the voltage angle difference which would most probably require considerably decreasing power generation output in north-eastern Germany. High power flows from Vierraden increase load on other lines going out from the Krajnik station which impedes switching operations and increases the risk of inducing excessive torsional vibrations on EDO turbine-generator units’ shafts. Such risks, however, only occur when the load on the cross-border line exceeds permissible values.

REFERENCES 1. Kundur P, Power System Stability and Control, McGraw-Hill Professional Publishing, 1994.

2. IEEE Working Group Interim Report: Effects of switching network disturbances on turbine – generator shaft systems, IEEE Transaction on Power Apparatus and Systems, vol. PAS-101, issue 9, September 1982.

3. Opera L, Popescu V., Sattinger, Coordinated synchronism check settings for optimal use of critical transmission network corridors, IEEE 2007 4. Lamrecht D.R., Problems of torsional stresses in shaft lines of turbine generators, CIGRE WG 11.01, Section 3, Recommendations, Electra no. 143, August 1992.

29

30

Jacek Klucznik, Krzysztof Dobrzyński, Zbigniew Lubośny / Gdańsk University of Technology Robert Trębski / PSE Operator SA

Authors / Biographies

Jacek Klucznik Gdańsk / Poland

Krzysztof Dobrzyński Gdańsk / Poland

Graduated with a Master’s degree from the Faculty of Electrical and Control Engineering at Gdańsk University of Technology (1999). Obtained his Doctor’s degree five years later. Currently employed as a lecturer at the Faculty of Electrical and Control Engineering of Gdańsk University of Technology.

Deals with automatic regulation systems for generators and turbines, wind energy and electrical power engineering safety automatic control

Graduated from the Faculty of Electrical Engineering, Warsaw University of Technology (1999).

Employed as an assistant at the Power Engineering Department of the Faculty of Electrical and Control Engineering at Gdańsk University of Technology since 2005. Areas of interest: cooperation of dispersed generation sources with the power system, mathematical modelling, electrical power engineering system control, BMS.

Zbigniew Lubośny Gdańsk / Poland

Robert Trębski Warszawa / Poland

Zbigniew Lubośny graduated from the Faculty of Electrical and Control Engineering at the Gdańsk University of Technology (1985). Obtained his Doctor’s degree in 1991 and habilitated in 1999 at the same university. A Professor of Sciences since 2004. Currently employed as assistant professor at Gdańsk University of Technology. Areas of interest: mathematical modelling, stability of electrical power systems, electrical power system control, use of artificial intelligence in electrical power system control, modelling and control of wind power plants.

Graduated from the Faculty of Electrical Engineering, Warsaw University of Technology (1993) and the Faculty of Management (1996), University of Warsaw. Employed at PSE S.A. from 1993 and later at PSE Operator S.A. Focuses his research on the problems of power engineering system modelling for on-line and off-line applications used at the National Control Centre and on the performance analyses of the electrical power system.

System for Dynamic Model Identification Based on Real Power System Measurements

SYSTEM FOR DYNAMIC MODEL IDENTIFICATION BASED ON REAL POWER SYSTEM MEASUREMENTS Jacek Klucznik / Gdańsk University of Technology Krzysztof Dobrzyński / Gdańsk University of Technology Zbigniew Lubośny / Gdańsk University of Technology Robert Trębski / PSE Operator SA

1. MODELLING AS A TOOL FOR PERFORMANCE ANALYSIS OF ELECTRICAL POWER SYSTEM Mathematical modelling of objects (systems) has been the main tool for system performance analysis, design and extreme testing. Modelling significantly reduces design costs and allows cutting the potential costs of failure or destruction due to an extreme state of equipment. It also decreases exposure of humans, animals and the environment in general.

The mathematical models which represent actual objects are generated in two ways: • mathematical modelling – an analytical approach in which the relationships to define the model are derived from laws of physics, including the known structure of objects and functional interdependencies of their components; • system identification – an experimental approach in which the object model is obtained from the data acquired by measurements of an existing (actual) object using a suitable method of estimating the model parameters.

The mathematical models based on mathematical modelling are generated with the assumption of numerous simplifications. This necessitates verifying the mathematical modelling validity. This can concern both the model structure and parameters. The model is verified by comparing its response (i.e. the response of an independently developed mode) or any other function which characterises the object dynamics to the experimental results.

Note, however, that the identification is not perfectly reliable. The primary difficulties are: • The lack of ideal measurement data. The data usually contains significant magnitudes of noise which can degrade the identification process. Suitable measurement data is not always available.

• The difficulty of determining the proper model structure in non-linear systems.

• The difficulty or impossibility of building stationary models for non-stationary systems (processes).

This paper presents a method for identification of parameters in a relatively complete model. The thesis states that there is a complete model of an actual electrical power system; the model parameters are identified on the basis of actual object measurements.

Transmission system operators utilize multi-machine models of electrical power systems, both for development design and ongoing system maintenance. The electrical power system model which is construed in this manner consists of defined component model types: transmission lines, transformers, loads, generators and models of automated and control systems. The structure of component models is usually strictly defined and is not (and most usually cannot be) modified by the user of the model. The data which defines the components model can be modified.

Abstract This paper presents a concept of a tool for verification of model parameters applicable to dynamic elements of the electric power system. The tool uses the PSLF design software, which is in general use by the transmission system operator. The innovation consists in using an additional application (in MS Windows environment) which controls the work of the design software. This

results in a tandem of a user-friendly interface for control and analysis of obtained results and of the calculation software engine which can simulate operating states of an electrical power system. The optimisation algorithms of the application adapt the simulation response to the actual course recorded for the verified object and thus estimates the validity of the given model.

31

32

Jacek Klucznik, Krzysztof Dobrzyński, Zbigniew Lubośny / Gdańsk University of Technology Robert Trębski / PSE Operator SA

It is obvious that model-based results depend on the input component data. The simulation tests define the requirements for safe operation of an electrical power system, which also translates into certain financial terms. This is why it is critical to correctly determine the data values which define the specific system components in order to ensure that the actual system is represented at the lowest error rate possible.

2. MODEL VERIFICATION METHOD PSE Operator SA uses the PSLF software to analyse the dynamic states of the electrical power system. This software utilizes an IEEE-compliant library of dynamic models in the simulation generation process. This means that the software user can access models with strictly defined structures, apart from a few exceptions1. Regarding the Polish Power System model, its has been established what models should be used for representation of specific elements of the given power unit, which is based on the knowledge of the control and regulating system structures of actual power plant power units. There are four dynamic models available for most power units.

These models are: the synchronous generator model, the model of a turbine with its governor, the model of the excitation system with its voltage regulator (AVR) and the model of a power system stabiliser (PSS).

The problem is to determine the values of parameters which define the given model. It is not always possible obtain a model dataset from the technical manuals of the equipment in question. The data included in the manuals is not always reliable or corresponding to the actual setup. The divergences may arise from changes in the parameter settings of controllers/governors done during operation, which are not always on record. Another source of divergences between the technical specifications and the actual systems they describe may be a change in the object properties due to operational wear. It is necessary for the transmission system operator to have a tool which allows one to determine or verify the parameters used in the model simulation software.

The generally used methods for model parameter estimations are: • The frequency response method. This method uses sinusoidal courses of varying frequencies as the input signals. The actual object is measured which allows producing its frequency characteristics. The frequency characteristics are widely used to asses the effectiveness of damping electromechanical swings by system stabilisers. They are also a valuable source of information on electromechanical oscillations in the system.

• The time response method. This method consists in comparing the model response with the actual object response. The most frequently used test function is the preset generator voltage peak. The method usually involves testing of bi-directional forcing magnitude changes, e.g. discrete preset voltage increase and decrease. This approach is adopted in this paper.

The processes of parameter identification have been covered in many published papers; however, the difference of the method developed by the authors is that identification is performed by using the same model the system operator uses. This means that when the model parameters are sought, the method uses a comparison between the courses measured for the actual object and the runs determined for the model, which are obtained with the same simulation software which will be used by the system operator once the identification process is complete. The concept of the identification process is presented in Fig. 1.

1 The software also allows creating user models, which sometimes are used in analysis.

System for Dynamic Model Identification Based on Real Power System Measurements

PSLF Electrical power system

Measurements

Preparation of measurement data y s(t)

Non-linear mathematical model

Control application Modification of model parameters

Simulation of system operating state

Preparation of simulation results y m(t)

N

V(y m(t)-ys(t)) < Vmin

T Verified model (identified dynamics)

Fig. 1 The structure of the algorithm for power system dynamics identification based on timings (the system response to a disturbance of the operating state), where: ym(t) is the model output signal, ys(t) is the system output signal, V(•) is the scalar function

The software developed by the authors and used for identification or verification of dynamic model parameters benefits from integration of two programs.

The first program is the control application, henceforth referred to as SDMP (Selection of Dynamic Model Parameters). This application has been developed in Visual Basic for MS Windows. The main program window is presented in Fig. 2. The purpose of this program is to provide an interface for the user who identifies the dynamic model data. The application allows selecting the model for which the parameters will be verified. The model selection process has two steps: first, the system model is selected (i.e. the PSLF files *.save and *.dyd), then the user selects the element model (turbine, system stabiliser, voltage regulator) to be verified. The SDMP application requires the user to submit information about the data measured on a real object and used for identification. The program allows reading measured data directly from recorders by supporting the COMTRADE format2.

2 COMTRADE is the world’s generally accepted standard of data recording format. It has been developed by IEEE and defined in the C37111-1991 standard, „IEEE Standard Common Format for Transient Data Exchange (COMTRADE) for Power Systems”; it was updated in 1999 (C37111-1999). This standard is enforced in Poland as PN-EN 60255-24:2004, „Electrical Relays – Part 24: Common Format for Transient Data Exchange (COMTRADE) for Power Systems”, which is the Polish translation of the international standard IEC 60255-24:2001 [3].

33

34

Jacek Klucznik, Krzysztof Dobrzyński, Zbigniew Lubośny / Gdańsk University of Technology Robert Trębski / PSE Operator SA

Fig. 2 Main window of SDMP testing environment

The program allows one to compare courses measured on actual objects to courses derived from three simulation types. Simulations of the following can be performed: • a generator model running running idle (the simulations are performed in the SDMP software by using the model of a generator, excitation system and voltage regulator described by a set of algebraic and differential equations); • single machine infinite bus system (the simulations are performed in the PSLF software, while the singlemachine model is automatically generated from the data on the multi-machine system); • a multi-machine model (the simulations are performed in the PSLF software by using the data files applicable to the Polish Power System.

The SDMP software requires the determination of the disturbance type to which the loaded response applies (or a selected part of this response3). The disturbance type selection to be modelled depends on the model type to be used in the calculations. The specific model types allow modelling of the following disturbances: • generator running idle: reference voltage step change • single-machine system: reference voltage step change reference power step change • multi-machine system: reference voltage step change reference power step change power unit shutdown 3-phase short-circuit on selected bus 3-phase short-circuit on selected line

3 In reality, a single response may include more than one disturbance.

System for Dynamic Model Identification Based on Real Power System Measurements

The second main purpose of the SDMP program is to evaluate the response from the simulation and to compare this response to the response measured on a real object. This process automatically selects the estimated parameters to approximate the model response as much as possible to the real object response. The measurement of the difference between the model and object responses is a defined scalar function. The program uses a fairly common function which is the sum of distance squares between the model response and the object response:

V(X ) 

t Tkoniec



t Tskok

( y m (t )  y o (t )) 2

(1)

where: V(X) is the scalar function; X = {ppp2,...,pK} is the vector of estimated parameters; ym(t), yo (t) are the model response (index m) and the object response (index o) in moment t; Tskok is the initial moment from which the function V(X) is calculated; Tkoniec is the terminal moment to which the function V(X) is calculated.

The initial moment Tskok, from which the value of V(X) is calculated, should correspond to the time at which the operating state disturbance occurs, e.g. an incremental change of the preset voltage. Inclusion of an earlier course part of the measured magnitude is not substantiated.

The terminal moment Tkoniec, to which the value of V(X) is calculated, may affect the quality of parameter estimation. Too high a value of Tkoniec , which causes a long interval of the stabilized state in the course of the function y (t), results in a condition under which the transient course has little effect on the value of the function V(X).

The process of estimating the parameters of the model X for the defined function V(X) consists in optimizing the function, or more precisely, minimizing it. This process can be performed by using the algorithms for searching of a local or global optimum.

The first group of these algorithms are gradient algorithms. These algorithm exhibit a relatively fast action. Their primary limitation (drawback) is the dependence of the result on the initial point. These algorithms find the local optimum. If the target function V(X) is a multi-extrema function, the usefulness of gradient algorithms is very limited.

The other group of algorithms includes the Monte Carlo algorithms and genetic algorithms. The strength and advantage of these algorithms is their ability to search the entire space stretched along the vector of the parameters being estimated. Their drawback is the inability (or a very limited ability at least) to precisely indicate the location of the V(X) extremum.

Both algorithm types feature specific drawbacks, but since they also have advantages which complement each other in a certain way, sometimes both types are used to indicate the area in which the global extremum is. Hence it is determined with a Monte Carlo or genetic algorithm, and further the precise location of the extremum is determined with a gradient algorithm.

The SDMP software user can choose from two optimisation methods: genetic or gradient. It is also possible to use both methods in a sequence, however, due to their varying specifics, the genetic method is applied first. Then the gradient method is applied based on the results obtained by the genetic approach.

3. VERIFICATION OF SDMP OPERATION The operation of the SDMP software has been verified with the simulations of a generator in running idle, a single-machine system and a multi-machine system. This paper presents the selected results of completed analyses.

Fig. 3 presents the results of the estimation of the generator voltage regulator gain in one of the power units at the Bełchatów power plant. The presented results have been obtained by using both methods simultaneously (in sequence), i.e. where the first step consists in determination of the parameter by using the genetic method and then the estimation process switches to the gradient method which accounts for the new value from the first step.

35

36

Jacek Klucznik, Krzysztof Dobrzyński, Zbigniew Lubośny / Gdańsk University of Technology Robert Trębski / PSE Operator SA

An example of an improper approach to the estimation process is shown in Fig. 4. It pertains to the estimation of selected parameters of the power system stabiliser, which are based on the system’s response to the change of the preset power by -10 MW at Power Unit 3 of the Bełchatów power plant. The results obtained by the genetic method in relation to time constants do not significantly depart from the initial values, but they vary to a large extent for the PSS gain ks. This is due to the fact that the parameter values of the power system stabiliser (provided that they do not destabilise the system) affect only the initial section of the response shown in Fig. 4.

The value of the target function being optimised is calculated from the complete response shown. In such case, even a slight but long-lasting deviation of the response causes an error (i.e. the target function change) larger than the one caused by the change of the parameter being estimated (the power system stabiliser parameter).

Fig. 3 Voltage reference step change by -10% in Power Unit 3 at Bełchatów. Estimated parameter: ka of AVR. Variant: generator in idle run

Fig. 4 Power reference step change by -10 MW in Power Unit 3 at Bełchatów. Estimated parameters: tl, t2, t4, t5, ks of PSS. Variant: singlemachine system

System for Dynamic Model Identification Based on Real Power System Measurements

Fig. 5 Voltage reference step change by -5% in Power Unit 3 at Bełchatów. Estimated parameters: tc, tb, ka, ta of voltage regulatorAVR. Variant: multi-machine system

Fig. 5 shows the calculation results obtained for the reference voltage step change by -5% of Generator 3 at the Bełchatów power plant. The calculations have been completed with the genetic method where the following parameters of the generator voltage regulator have been estimated: ka, ta, tc, tb. The figure indicates that the obtained parameters allow to match the generator response well.

4. SUMMARY The paper presents the concept and operating results of the application which enables verification and estimation of the parameters of the dynamic models for elements of power generating units. The developed tool benefits from the advantages of the PSLF software which is the calculation platform allowing obtaining the simulation courses for any electrical power unit operated within the Polish Power System, as well as from the advantages of the peripheral MS Windows application which provides freedom of programming possibilities. The software functionality described here would have not been possible if only one of the presented environments was used. The EPCL language of the PSLF software is too simple to enable production of complex applications – it lacks the capacity of including GUI elements, which are friendly to the user (e.g. drop-down lists, check boxes, etc.), or advanced mathematical functions. On the other hand, development of a professional simulation software as advanced as PSLF is a very complex task.

The resulting application allows for a comfortable and effective verification of the models used by the system operator, which authenticate the results of the Polish Power System analyses. This undoubtedly contributes to improvements of the operating safety of the Polish Power System.

Note that the method proposed by the authors coupling dedicated calculation tools (PSLF, PLANS, DIgSILENT PowerFactory, etc.) with applications customized for the transmission or distribution system operators enables a significant improvement of software functionalities. This opens completely new and more powerful possibilities of using calculation software suites for operating analysis and development of electrical power systems.

REFERENCES 1. Lubośny Z., Dobrzyński K., Klucznik J., Opracowanie i wykonanie środowiska testowego do badań modeli dynamicznych. Wykorzystanie przebiegów rejestracji szybkozmiennych do weryfikacji modeli dynamicznych KSE. Etap I, II i III, EPS RESEARCH, Gdańsk, 2009.

2. PSLF User’s Manual.

3. PN-EN 60255-24:2004, Electrical Relays – Part 24: Common Format for Transient Data Exchange (COMTRADE) for Power Systems.

37

38

Jerzy Majewski / ENERGA SA

Authors / Biographies

Jerzy Majewski Gdańsk / Poland Acts as principal investment expert in ENERGA SA. His research interests include solar energy, cogeneration, biogas production and treatment, ORC systems, the use of biomass in power engineering.

Graduated from the Faculty of Mechanical Engineering at Gdańsk University of Technology (1990), speciality in cars and tractors. From 1990 to 1997 he was a researcher at the Institute of Fluid-Flow Machinery at the Polish Academy of Sciences in Gdańsk. In 1997 he obtained a doctoral degree in technology, speciality in construction and operation of machinery. His doctoral dissertation won the Prime Minister’s Award. His scientific output includes over thirty publications, studies and papers in the field of construction and operation of machinery, tribology, automotive technology and renewable energy.

Experimental Investigations on the Usefulness of Air Solar Collectors Supporting the Drying Processes

EXPERIMENTAL INVESTIGATIONS ON THE USEFULNESS OF AIR SOLAR COLLECTORS SUPPORTING THE DRYING PROCESSES Jerzy Majewski / ENERGA SA

1. INTRODUCTION “The sun is a huge and virtually inexhaustible source of energy. Nature gave the sun to mankind and endowed people with reason so that they learn to use such a source of energy” - that is my motto, which aptly captures the essence of solar energy use.

The sun is the driving force behind virtually all processes occurring in nature. Rivers flow, wind blows and plants grow in the fields thanks to solar energy. Thermal solar collectors and photovoltaic cells are used for direct conversion of energy from solar radiation into useful forms of energy. In Poland, a relatively fast growing sector is the so-called liquid solar collectors, used for heating hot tap water. This is due to the possibility of widespread use of these collectors in households and public buildings.

The situation is totally different in the case of solar collectors for heating air. The author’s information show that collectors of this type are not manufactured in batches in our country, and the existing devices either come from private import, or are constructed by the users themselves.

Practice shows that air solar collectors may be an effective tool to support with solar energy all technological processes requiring heated air in the period of the largest supply of solar radiation, i.e. in the months May to August. These processes include the drying of agricultural products and biomass.

Negligible use of air solar collectors in Poland, and relatively large areas of possible application encouraged the author to construct a prototype of such a collector, and to perform tests enabling determining the energy yield.

2. AVAILABILITY OF SOLAR ENERGY IN POLAND It is true that Poland does not belong to those countries with the highest supply of solar radiation, but we have solar exposure conditions similar to our neighbour - Germany, the European and world leader in the field of photovoltaics and thermal solar collectors.

The annual sums of total radiation (expressed in kWh/m2 per year) are very important for solar power engineering. The annual sum of total radiation shows the actual amount of solar energy reaching the Earth’s surface per year, including transparency of the atmosphere and cloud cover. Fig. 1 shows a map of the annual total radiation in Poland. It shows that the annual total radiation in the area of Poland is about 1000 kWh/m2 with a deviation of 10 percent.

Abstract The article. presents the results of experimental investigations aimed at determining the efficiency and power of innovative air solar collectors constructed by the author.

The article describes the construction of the experimental post and research methodology. The experiments were conducted at various temperatures of supplying air and various flow rates, so as to present the conditions in which the collector will be used in the actual drying installation as exactly as possible.

The results are presented in the form of graphs showing the dependence of a single collector and a set of two collectors connected in series on the changing flow and temperature parameters.

On the basis of the measured parameters, the power of a single collector, and seasonal energy yield of a set of collectors in a typical drying installation were determined using calculations.

39

40

Jerzy Majewski / ENERGA SA

KWh/m2 per year (MJ/m2 per day)

Fig. 1. Map of solar radiation in Poland (source: IMGW)

3. CONSTRUCTION OF INNOVATIVE AIR SOLAR COLLECTOR Innovative air solar collector is made entirely of plastic. Its structure is shown in Fig. 2.

Fig. 2. Innovative air solar collector made of plastic

Experimental Investigations on the Usefulness of Air Solar Collectors Supporting the Drying Processes

The absorber plate is made of black polycarbonate chamber plate. This solution provides contact between the flowing air with the absorber practically on its entire surface, which contributes to a much more efficient heat exchange than in traditional collectors, where the contact area is small.

The transparent warming of the top side of the collector is made of transparent polycarbonate chamber plate. Such a design is much better than traditional glass, mainly due to the several times less weight and more efficient thermal insulation. The bottom side of the absorber is insulated with polyethylene foam and aluminium foil. The collecting tubes of the collector are made of polyvinyl chloride. All elements of the collector are connected using a flexible adhesive based on SMP (Silyl Modified Polymer).

4. PURPOSE AND SCOPE OF TESTS The aim of the tests was to experimentally determine the efficiency (power) of air solar collectors designed by the author in a version insulated with collecting tubes with diameters of 110 and 160 mm.

The tests included measurements of power of individual collectors of both types, and of sets consisting of two collectors connected in series.

The measurements were performed for various air flow rates and various temperatures of supplying air, so as to present the conditions in which the collector will be used in the drying processes as exactly as possible.

5. CONSTRUCTION OF RESEARCH POST The post consists of: • radial blower driven by a three-phase current motor • inverter used for smooth adjustment of motor speed, and thus the mass flow rate of air supplying the tested collector • pipeline connecting the blower to the tested collector • reducer for measuring air flow rate, located in the supplying pipeline • electronic air flow velocity meter Testo 435 • system for heating the inlet air, equipped with two electric heaters (2 x 1.2 kW) • points for measuring the inlet and outlet air temperatures. Those points are equipped with resistance temperature sensors PT-100 enabling temperature measurement with an accuracy of 0.1 deg. C • tested air collector (set of collectors) • solar radiation simulator enabling irradiation of a single tested collector with luminous flux with a radiation intensity of up to 1.5 kW/m2, in the set of two collectors, the flux intensity is up to 1.0 kW/m2 • pyranometer (instrument for measuring solar radiation intensity).

41

42

Jerzy Majewski / ENERGA SA

Photo 1. Solar radiation simulator

Photo 2. Pyranometer for measuring radiation intensity

6. MEASURING METHOD All measurements were performed at a radiation intensity of about 1000 W/m2 The following were recorded simultaneously during the measurement: • air flow velocity in collecting tubes of the collector (used for volumetric and mass calculations of air flow) • temperature of air supplying and leaving the tested collector.

The tests were performed for various temperatures of supplying air, ranging from ambient temperature (about 20 deg. C) to about 60 deg. C. The air flow rate was also changed during the measurements in the range of 0.014 m3/s to a maximum value that could be obtained due to the efficiency of the blower at a specific system load with a single collector or a set of collectors.

The power of the tested collector (set of collectors) is determined based on the measured values.

Experimental Investigations on the Usefulness of Air Solar Collectors Supporting the Drying Processes

7. MEASUREMENTS FOR SINGLE COLLECTOR AND TWO COLLECTORS CONNECTED IN SERIES WITH COLLECTING TUBES WITH A DIAMETER OF 110 MM

Photo 3. Two collectors connected in series with tubes with a diameter of 110 mm

In order to determine the impact of air flow rate on the collector power, measurements were performed for flow rates ranging from 0.014 m3/s to 0.081 m3/s. In this case, the supplying air had a constant temperature of about 22 deg. C. The results are shown in the following charts.

Temperature single collector 22t.C deg.C Pr zyr os increase t tem p. win apoj. k olek tor ze for dlatinlet t dolof22s

60

35

55

30

50

Δt [deg.C] � t [st.C]

outlettwyl. temp. [deg.C] [st.C]

Outlet for aksingle tinlet22s 22 deg.C Tem air p. temperature w yl. dla poj. olek collector tor a dlafort dol. t.C

45 40 35

25 20 15 10

30 0

0,01

0,02

0,03

0,04

0,05

0,06

0,07

0,08

0,09

Flow rate przepływu [m /s] Natężenie [m3/s] 3

Fig. 3. Graph presenting the relation between outlet temperature and air flow rate

0

0,01 0,02 0,03 0,04 0,05 0,06 0,07 0,08 0,09

Flow rate [m /s] u [m 3/s ] Natężenie pr zepływ 3

Fig. 4. Graph presenting the relation between increase in air temperature in the collector and air flow rate

Power [W] Moc [W]

Max. power of poj. a single collector for tinlet 22t.C deg.C Moc m ax. k olek tor a dla t d o l of 22s

1000 800 600 400 0

0,01 0,02 0,03 0,04 0,05 0,06 0,07 0,08 0,09 Flow rate [m3/s] u [m 3/s ] Natężenie pr zepływ

Fig. 5. Graph presenting the relation between the maximum collector power and air flow rate

43

Jerzy Majewski / ENERGA SA

44

Also, tests aimed at determining the effect of the temperature of supplying air on the collector power (efficiency).

Measurements were performed for the temperatures of supplying air of about 20 deg. C to about 60 deg.

C and for two flow rate values: 0.014 m3/s and 0.028 m3/s.

The results are shown in the following charts.

Power of a single collector Moc pojedynczego kolektora

800 700

Power [W] Moc [W]

600

V=0,028 m 3/s

500 400

V=0,014 m 3/s

300 200 100 0 20

30

40

50

Temp. dolot. [st.C] Inlet airpow. temp. [deg.C]

60

70

Fig. 6. Graph presenting the relation between the collector power and the temperature of supplying air

Temperature increase in 2 collectors for tinlet zyr os t tem p. w 2 k olek t. dla t dol 22s t.C ofPr22 deg.C

Outlet air temperature 2 collectors for t.C tinlet 22 deg.C Tem p. w yl. dla 2for k olek t. dla t dol. 22s 70

50

65

Δt� t[deg.C] [st.C]

outlettwyl. temp. [deg.C] [s t.C]

In order to determine the impact of air flow rate on the power of two collectors connected in series, measurements were performed for flow rates ranging from 0.014 m3/s to 0.059 m3/s. In this case, the supplying air had a constant temperature of about 22 deg. C. The results are shown in the following charts.

60 55 50 45 0

0,01

0,02

0,03

0,04

0,05

0,06

0,07

Natężenie pr3zepływ u [m 3/s ] Flow rate [m /s]

45 40 35 30 25 0

0,01

0,02

0,03

0,04

0,05

0,06

0,07

Natężenie pr zepływ u [m 3/s ] Flow rate [m3/s]

Fig. 7. Graph presenting the relation between outlet temperature and air flow rate

Fig. 8. Graph presenting the relation between increase in air temperature in set of collectors and air flow rate

Moc [W] [W] Power

Max. power of 22kcollectors 22 deg.C Moc m ax. olek tor ów for dla tinlet t dol 22soft.C 2000 1800 1600 1400 1200 1000 800 600 400 0

0,01

0,02

0,03

0,04

0,05

3 FlowNatężenie rate [mpr /s] zepływ u [m 3/s ]

0,06

0,07

Fig. 9. Graph presenting the relation between the maximum power of set of collectors and air flow rate

Also, tests aimed at determining the effect of the temperature of supplying air on the power (efficiency) of a set of two collectors connected in series.

Measurements were performed for the temperatures of supplying air of about 20 deg. C to about 60 deg.

C and for two flow rate values: 0.014 m3/s and 0.028 m3/s.

The results are shown in the following chart.

Experimental Investigations on the Usefulness of Air Solar Collectors Supporting the Drying Processes

Power of two collectors Moc dwóch kolektorów

1400

Power [W] Moc [W]

1200 1000

V=0,028 m 3/s

800 600 V=0,014 m 3/s

400 200 0 20

30

40

50

60

70

pow. [deg.C] dolot. [st.C] InletTemp. air temp.

Fig. 10. Graph presenting the relation between the power of set of collectors and the temperature of supplying air

8. CONCLUSIONS • The maximum power of a single collector is 1006.5 W for the variant with collecting tubes of Φ110 mm and 1076. 4 W for the variant with collecting tubes of Φ160 mm, which means an efficiency of 57-61 percent. The increase in air temperature in such conditions is 11. 4 to 12.5 degrees C. • The maximum power of a set of two collectors connected in series is 1775 W for the variant with collecting tubes of Φ110 mm and 1986. 4 W for the variant with collecting tubes of Φ160 mm, which means an efficiency of 50-56 percent. The increase in air temperature in such conditions is about 27 degrees C. • In order to achieve the maximum efficiency of collectors in a drying installation, they should operate in conditions that make it possible to achieve maximum power. It should not be attempted to make the collectors themselves provide the desired air temperature, as then they will operate at a much lower efficiency.

• In order to achieve the maximum power of the collector (set of collectors), a sufficient air flow at the level of 300 m3/h should be provided. This is the flow value for a single collector. In the case of an entire drying installation (12 collectors) this value is 3600 m3/h.

9. BENEFITS • If we assume that the drying process lasts for 75 days, and the average supply of solar radiation at this time is 4.7 kWh/m2 per day, than considering the average collector efficiency of 50 percent, we obtain energy equal to 13.4 GJ. This energy relates to an installation consisting of twelve collectors.

• This means savings amounting to 382 litres of fuel oil with a calorific value of 39 MJ/litre, assuming a 90-percent efficiency of the burner.

REFERENCES 1.Majewski J., Mikielewicz D., Wajs J., Zygmunt J., Experimental Investigations of a Prototype of Solar Collector on a Rig Equipped with Artificial Source of Illumination, Międzynarodowe Sympozjum Heat Transfer and Renewable Sources of Energy HTRSE-2006, Międzyzdroje 2006.

2.Wiśniewski G., Gołębiowski S., Gryciuk M., Kolektory słoneczne – poradnik wykorzystania energii słonecznej, Warszawa 2001.

3.PN-EN 12975-1:2007 Thermal solar systems and components - Solar collectors - Part 1: General Requirements.

4.PN-EN 12975-2:2007 Thermal solar systems and components - Solar collectors - Part 2: Test methods.

45

46

Robert Małkowski / Gdańsk University of Technology Zbigniew Szczerba / Gdańsk University of Technology

Authors / Biographies

Robert Małkowski Gdańsk / Poland

Zbigniew Szczerba Gdańsk / Poland

Graduated from the Faculty of Electrical and Control Engineering at Gdańsk University of Technology (1999). In 2003, he received his PhD. He is currently employed as an assistant professor at the Department of Electrical Engineering of Gdańsk University of Technology. The scope of his scientific interests covers issues related to wind energy, catastrophic failures of power systems, and to adjustment of voltage levels and distribution of reactive power in power systems.

In 1952, obtained his Engineer degree. Four years later, he obtained the MsC degree. In 1963, he graduated from PhD studies in the Faculty of Electrical Engineering at Gdańsk University of Technology. Among others, within the Power Engineering Department, he managed his team which designed numerous types of excitation systems and generator voltage regulators of power ranging from a few hundred kW for the shipbuilding industry to 500 MW. In the peak period, generators controlled by these regulators constituted 75% of the power provided by the national power system.

In 1972, he was transferred to the Institute of Power Systems Automation in Wrocław where he took up the post of Deputy Scientific Director. In 1977, he obtained the assistant professor degree.

He took up the post of manager of the Power Engineering Unit in the faculty of Electrical Engineering at Gdańsk University of Technology. Soon afterwards, he obtained the associate professor degree and was twice chosen to be the dean of this department. In 1987-90, he worked as visiting professor at the University of Technology in Oran (Algeria). Having returned to Poland, he organised the Power Systems Department in the present Faculty of Electrical and Control Engineering. Since 1991, he has had the full professor degree of Gdańsk University of Technology. In 1990-1996, he held the post of Vice-Rector for Academic Affairs. He is an author or co-author of over 50 patents and over 200 scientific works, most of which have been implemented in practice.

New Algorithm for Controlling Transformers that Supply Distribution Network

47

NEW ALGORITHM FOR CONTROLLING TRANSFORMERS THAT SUPPLY DISTRIBUTION NETWORK Robert Małkowski / Gdańsk University of Technology Zbigniew Szczerba / Gdańsk University of Technology

1. INTRODUCTION In the study [1] the authors describe the current condition of control systems for transformers supplying the distribution network. It presents a feasibility analysis for blocking control units in abnormal voltage conditions. It also provides basic theoretical assumptions to be met by the transformer control algorithm, which takes into account the continuouschange in load voltage sensitivity coefficient

dQo . dU

Today we propose a new transformer operation algorithm, which takes into account the earlier suggestions.

The idea of an intelligent control unit is based on determining on a regular basis the derivative value

dQo and dU

tendency to the formation of voltage avalanches in the power system. These values are obtained by recording (sampling in time) and processing the following: • voltage value on both sides of the transformer • reactive power on the MV side • through recording and processing of these values before and immediately after activation of the tap changer, controlled by the control unit in question.

2. NUMBER OF TRANSFORMER CONTROL UNITS IN THE NATIONAL POWER SYSTEM (NPS) The survey results included in the study [2] concerning transformer control units installed in 110 kV/MV stations show that a great number of transformer control units are RNTH-3 control units (the survey included several hundred transformers).

Fig. 1. The statistical population of transformer control units included in the survey

Abstract This article describes an example of a new intelligent algorithm for a transformer control unit. The control system for 110/SN transformers with the proposed operation algorithm, enables the automatic adjustment of

control unit operation algorithm to the current conditions of the power system (e.g. voltage avalanche) and thus may improve energy security.

Robert Małkowski / Gdańsk University of Technology Zbigniew Szczerba / Gdańsk University of Technology

48

Analysis of the documentations used in transformer control units in the NPS leads to the following conclusions: • Operation of control units is based solely on the measurement of voltage. The criterion for operation of all control units is the exceeded set voltage value (with accuracy to the dead zone). Current measurement is used only to determine the value of voltage of current compensation and overload blockade.

• No transformer control units allow impact on (switching on or off) local sources of reactive power.

• The algorithms for operation of modern control units in no way determine the tendency to the formation of voltage avalanches.

E.g. the literature includes a description of a device whose algorithm allows predicting the situations that can lead to blackouts. It is the Collapse Prediction Relay CPR-D offered by a-eberle [3].

However, the algorithm of this device is very complex (Fig. 2).

Fig. 2. Block diagram of operation algorithm for CPR-D

The following are used to determine the need to block or properly adjust the tap-changer in the transformer: • bifurcation theory, combined with elements of neural networks • defining Lyapunov exponents • identification of voltage lowering • damping coefficients.

The main difference between CPR-D and the operation algorithm presented in this article lies in the assumed purpose of operation. In the case of CPR-D the purpose is to identify the state of emergency. The purpose of operation of the proposed algorithm is to prevent the occurrence of the state of emergency caused by voltage avalanche.

3. DESCRIPTION OF OPERATION ALGORITHM Operation algorithm proposed by the authors ensures adjustment of transformer control to the current state of the power system and thus may significantly contribute to improving energy security.

In an example implementation of the control method for transformers powering a 110/MV distribution network, the operation algorithm may proceed as follows: at specified time intervals the following values are measured and recorded on both sides of the transformer, i.e. the upper voltage in transformer UT and the lower voltage in transformer UTd as well as the current value of reactive power on the lower side of the transformer QTd, i.e. on the MV load side. On this basis, the load voltage sensitivity coefficient for reactive power is determined on a regular basis

dQTd and depending on its value, the upper voltage value UTg and the current rate of dU Td

New Algorithm for Controlling Transformers that Supply Distribution Network

changes in the upper voltage

dU Tg dt

a decision is made on whether the current set value of lower voltage UTd,

is to be maintained by switching transformer taps, or whether the adjustment should be stopped (operation of transformers with fixed tap). In the case of a distribution network node equipped with a capacitor bank, when the current value of the upper voltage in the transformer is outside the range determined by the set minimum value of the upper voltage UTgzm and the set maximum value of the upper voltage UTgzM,, the first adjustment attempt is made by switching on or off the subsequent sections of the capacitor bank (Fig. 3).

Measurement: :

UT, I T

Designation: � � ���� � �� �� � � ��� � �� �� � �

T

��� � ���� � ��

N

� � � ��� � ��

T

T

���������� ��

N

N

Switching on capacitor bank .

�

Switching off . capacitor bank

�

������������� �� � N T �

� � � ��� � ��

T

���������� �� T

N

Adjustment acc. to. crit. “Lower voltage”

�

N

������������� ��

N

T �

Fig. 3. Diagram of the algorithm responsible for the use of reactive power sources, e.g. capacitor banks (the process of switching on and off capacitor banks has not been presented in detail)

In a situation where the current value of the upper voltage UTg is within the range defined by < UTgzm; UTgzM >, the current set value of the lower voltage UTd is maintained by switching the transformer taps. In a situation where the value of the upper voltage UTg is beyond the acceptable set range and can not be adjusted by means of capacitor banks, the transformer transmission is controlled in the direction improving the voltage stability at deficiency of or excess reactive power, accordingly.

49

Robert Małkowski / Gdańsk University of Technology Zbigniew Szczerba / Gdańsk University of Technology

50

�� � �

�

�� � �

��

T

�� � � ��

�

T

�� � �

N

��

Adjustment acc. to crit. “Tap”

M-

N

Adjustment acc. to crit. “Lower voltage”

�� � � �� �� � �

T

�� � � ��

�

T

�� � � ��

� � � > � � ���

M-

N .

Adjustment acc. to crit. “Lower voltage”

T

N

N � � �� = � � ���

.

Adjustment acc. to crit. “Tap”

T

�� � � �� �� � �

�� � � ��

�

N

�� � � ��

M-

� � � < � � ���

Adjustment acc. to crit. “Lower voltage”

T

T

N

N

� � �� = � � ���

Adjustment acc. to crit. “Tap” Adjustment acc. to crit. “Lower voltage” �

Fig. 4. Diagram of the adaptive algorithm for 110 kV/MV transformer control unit. The case of undervoltage

• In the case of deficiency of reactive power the adjustment is made as follows (Fig. 4): if

dQo  0 – a fixed transmission or previously set lower voltage UTdz is maintained the fixed transmission dU dU Tg exceeds the set negative is maintained when the current rate of changes in the upper voltage dt rate of changes in the upper voltage

dUTg dt

M

; otherwise the current set lower voltage UTdz is

maintained by switching the transformer taps • if

dQo  0 – a fixed transmission is maintained or the voltage on the lower side of the transformer is dU

reduced to the minimum acceptable set lower voltage UTdzm and then the set value UTdzm is maintained • the fixed transmission is maintained when the current rate of changes in the upper voltage the set negative rate of changes in the upper voltage changes in the upper voltage

dU Tg dt

M

dU Tg dt

dUTg dt

dU Tg dt

exceeds

, whereas when the current rate of M

does not exceed the set negative rate of changes in the upper voltage

, and the current value of the lower voltage UTd is higher than the set minimum value of the

lower voltage UTdzm – the lower voltage is adjusted to the set minimum value through the transformer tap switch, if

dQo  0 – the previously set voltage on the lower side of the transformer UTdz is maintained, dU

New Algorithm for Controlling Transformers that Supply Distribution Network

51

or this voltage is increased to the maximum acceptable set lower voltage UTdzM, and then the set value UTdzM is maintained • the lower voltage is adjusted to the set maximum value when the current rate of changes in the upper voltage

dU Tg dt

does not exceed the set negative rate of changes in the upper voltage

dU Tg dt

, and the M

current value of the lower voltage UTd is less than the set maximum value of the lower voltage UTdM. The proposed algorithm can be implemented automatically by sampling the appropriate values in time, based on entering the voltage value on both sides of the transformer and reactive power on the MV load side on a regular basis to the memory at specified intervals, adjustable within the limits, e.g. from several to tens of seconds. If the set voltage value on the MV side is maintained, a tap switch signal is sent (identifying the switch moment), and the values of voltage and power before switching and after obtaining a new status set after tap switching are entered to the memory. Recording and processing of the values of voltage and power measured on a regular basis, and before and after activating the tap switch, can detect a tendency to the formation of voltage avalanche in the power system and identify situations in which the adjustment block is useful or harmful.

4. SUMMARY Due to the variability of load power in time, it is not possible to determine (based on periodic measurements) the situations in which a permanent lock or a permanent change in the control unit operation algorithm should be used.

The 110/MV transformer control system with the proposed operation algorithm presented in this article enables the automatic adjustment of the algorithm, the proposed action will provide the algorithm for automatic adjustment of control unit operation algorithm to the current conditions of the power system and will thus improve energy security.

REFERENCES 1. Małkowski R., Szczerba Z., Wpływ struktury, algorytmów działania oraz nastawień układów regulatorów transforma torów 110/SN na możliwość powstania i przebieg awarii napięciowej, APE ’09.

2. Survey sent to distribution companies, 2008.

3. a-eberle, CPR-D Collapse Prediction Relay, http://www.a-eberle.de, 2009.

4. Patent application no. P.391598, title: The method of controlling transformers that supply a distribution network, June 2010.

5. Małkowski R., Szczerba Z., Wpływ struktury, algorytmów działania oraz nastawień układów regulatorów transformatorów 110/SN na możliwość powstania i przebieg awarii napięciowej, materiały konferencji APE ’09, Jurata, June 2009.

52

Tomasz Okoń / Wrocław University of Technology Kazimierz Wilkosz / Wrocław University of Technology

Authors / Biographies

Tomasz Okoń Wrocław / Poland

Kazimierz Wilkosz Wrocław / Poland

Graduated with a Master’s degree from the Faculty of Electrical Engineering, Wrocław University of Technology. Currently a PhD student at the Faculty.

A student member of IEEE (Institute of Electrical and Electronics Engineers) from 2006. His area of interest is modelling of electric power systems with specific focus on power system state estimation.

Graduated with a Master’s degree, received Ph.D. and D.Sc. degrees at Wrocław University of Technology. Following PhD studies at the Institute of Electrical Power Engineering, began working at the Institute. Currently employed as associate professor. A member of SEP, CIGRE, and the scientific secretary at the Power Systems Section with the Committee on Electrical Engineering of Polish Academy of Sciences. Also a member of scientific committees at many national and foreign conferences.

A reviewer of papers submitted to scientific journals (e.g. IEEE Transactions on Power Delivery, The International Journal for Computation and Mathematics in Electrical and Electronic Engineering) and conferences (e.g. PSCC, ICHQP, EPQU).

His scientific interest and educational activities focus on the analysis of electrical power systems and power engineering applications of IT.

Impact of UPFC Operational Modes on Power System State Estimation

IMPACT OF UPFC OPERATIONAL MODES ON POWER SYSTEM STATE ESTIMATION Tomasz Okoń / Wrocław University of Technology Kazimierz Wilkosz / Wrocław University of Technology

1. INTRODUCTION Power system state estimation is a significant part of real-time power system modelling [1, 2]. Based on the redundant set of measurement data collected with use of remote data transmission systems, the estimation enables determining the most reliable assessment of the state vector for a power system (i.e. the voltages in system nodes). The state estimator is required to provide a reliable estimation procedure, a precise determination of the state vector and a short calculation time, especially in emergency situations.

This paper focuses on estimation of the state of a power system in which one of the FACTS (Flexible AC Transmission System) devices is operated, namely the UPFC (Unified Power Flow Controller) device [3]. FACTS devices are becoming an essential control element of modern power systems. These devices enable smooth control of power systems. Their important feature is short reaction time.

UPFC is the most advanced device in the FACTS device family. This device is considered in the paper. UPFC combines the functionalities of STATCOM (STATic COMpensator) and of SSSC (Static Synchronous Series Compensator). The following modes of its operation are investigated [4]: 1) operation with full device functionality, 2) operation as SSSC (the SSSC mode, 3) operation as STATCOM (the STATCOM mode), 4) the device is off.

The purpose of this paper is to present the results of an analysis of state estimation features, when this estimation is performed on a model of the power system with UPFC , and the information about the UPFC operational modes is not used in the computational procedure. The features of the state estimation are compared for the conditioning of the estimation computational process, accuracy of the estimation results and the number of iterations after which the result is obtained.

Shunt

Series transformer

�� � � ���

DC connector �������������

�������������� ��������� Konwerter Converter 22

Converter 1 Konwerter

transformer ��������������

����

Fig. 1 UPFC configuration diagram

Abstract The paper deals with the state estimation of a power system in which a UPFC device is installed. UPFC can operate in several modes: full use of its functions, only SSSC functions, only STATCOM functions and its switching off. The purpose of this paper is to present the results of analysis of state estimation which does not use

the information about the actual operational mode of UPFC. The features of the state estimation performed in the indicated conditions are compared for the estimation conditioning of the calculation process, the estimation accuracy and the number of iterations after which the estimation results are obtained.

53

Tomasz Okoń / Wrocław University of Technology Kazimierz Wilkosz / Wrocław University of Technology

54

2. UPFC DEVICE Fig. 1 presents a configuration diagram of the UPFC device. The device includes two converters built on GTO thyristors and coupled by a DC link (a DC connection) [4]. When the device operates in the standard configuration, converter no. 1 supplies reactive power through the shunt transformer and thus can maintain the preset voltage level at the connection point; it can also supply or consume active power, depending on the demand of converter no. 2. Converter no. 2, which supplies voltage with regulated modulus and phase angle into the power system, gives effect equivalent to introduction of an additional impedance which modifies the series impedance of the power line at the terminal of which the UPFC operates.

UPFC Device Model Fig. 2 shows the model of the UPFC device [3]. UPFC is modelled by the controlled voltage sources with the following source voltages: U vR  U vR  e j�vR , U cR  U cR  e j�cR ; the impedances, z vR and z cR are connected in series to these sources. The primary function of the serial source is to control the impedance between the nodes i and j. The primary function of the source in the shunt branch is to control the voltage at the node i. Operation of UPFC results in control of the voltage at the node i and of the active and reactive power flows between the nodes i and j. The mentioned flows can be controlled independently. i

Ii

I vR Ui

zcR

I cR

 U cR 

j





* * z v R Re U vR  I vR  U cR  I cR  0



Ij

Uj

U vR 

Fig. 2 Model of UPFC

3. MODELS OF DEVICES WHICH FUNCTIONS CAN BE PERFORMED BY UPFC Model of SSSC Fig. 3 shows the model of the SSSC device represented by the controlled voltage source with the source voltage U cR  U cR  e j�cR , which is connected in series to the power line, and by the impedance z cR connected in series with the source [3, 5]. SSSC supplies voltage with a regulated modulus into the system; the voltage is phase-shifted by 90° in relation to the current in the branch. Hence in the case of SSSC it is possible to control the equivalent reactance between the nodes i and j. This reactance can be higher or lower than the longitudinal reactance of the power line. SSSC does not enable independent control of reactive and active power as UPFC does.

Model of STATCOM Fig. 4 shows the model of the STATCOM device represented by the controlled voltage source, which is shunt-connected to the power line and has the source voltage U vR  U vR  e j�vR , and by the impedance z vR [3] connected in series with the source. STATCOM can only supply reactive power into the node i. If the resistance RvR is omitted, the device does not generate and also does not absorb any active power.

Ii

i

I cR

 U cR 

zcR



j

i

Ii

Ij



I vR

* Re U cR  I cR 0

Uj

Ui

 U vR 

�

Rys. 3. Model of SSSC

z vR  RvR  jX vR





* Re U vR  I vR 0

�

Rys. 4. �Model of STATCOM

�

Impact of UPFC Operational Modes on Power System State Estimation

55

4. ESTIMATION OF POWER SYSTEM STATE WITH WEIGHTED LEAST SQUARES METHOD When estimating the power system state by using the weighted least squares method, the following objective function is minimised [1, 6]: J x  



1 m   z  hi (x) 2 i 1 i



2

Rii 

1 T  z  h(x)  R 1  z  h(x) 2

(1)

where: z is the m-dimensional vector of measurements; h(x) is the vector of non-linear functions among the measured quantities and the -dimensional state vector x; R is the diagonal matrix with elements Rii=σii2; σii2 is the variance of the i-th measurement data item.

Minimisation of the function (1) leads to iterative solution of normal equations:

 



 



G x k  x k 1  x k  H T x k  R1 z h(x k )



(2)

 

 

h x k

is where k is the number of iterations, G x k  HTx k  R 1 H xk  is the gain matrix and H x k  k  x a Jacobian matrix.

The node powers and power flows in the considered power system are defined by the following formulas: *

(3)

Pi  jQi  U i  Y row i  U



 



*

2

Pij  jQij   y sij  y ij , y ij  U i , U j  U i



T

(4)

where: Ρi Q i are the node active and reactive powers, respectively, in the i-th node; Yrow i is the i-th row T of the matrix of node admittances; Y row i  y i1 , y i 2 ,..., y in ; U  U 1 , U 2 ,..., U n Ρij, Qii are the flows of active and reactive powers, respectively, in the branch between the i-th and j-th node (the branch i-j) at the node i: yij is the longitudinal admittance of the branch i-j; ysij is the shunt admittance of the branch i-j at the node i (type model).

If UPFC is operated in the power system, in the power system state estimation, the following relationships are taken into account [4, 7]:



 



*

*







*

2

*

*

Sij  U i  I cR  I vR   y vR  y cR , y cR , y cR , y vR  U i , U i U j , U i  U cR , U i U vR



*

*





*

*

2

S ji  U j  I CR  y CR ,  y CR ,  y CR  U j U i ,

U j , U j  U CR



T



T

(6)

and the condition: Pbb  e S VR  S CR   0   *

(7)









*

*

where: S CR  U CR  I CR   y CR , y CR , y CR  U CR  U i , *





S VR  U VR  I VR   y VR ,







y VR  U VR  U i ,

UVR

,

2

T

U CR U j , U CR

2

, T

1 y CR  Z cR1, , yVR  Z vR .

(5)

56

Tomasz Okoń / Wrocław University of Technology Kazimierz Wilkosz / Wrocław University of Technology

5. CONSIDERED PARAMETERS OF STATE ESTIMATION Conditioning of the calculation process The measure of conditioning (i.e. the measure of numerical instability [8]) of the calculation process for the power system state estimation is the conditioning index of the gain matrix G. In this paper the index is defined as follows: cond G    M  m , where: λm, λM are respectively the minimum and the maximum modulus of eigenvalues of the matrix G. The higher the index cond (G) is, the worse the conditioning of the problem is.

Accuracy of state estimation results In order to assess the accuracy of the state estimation results, this paper uses the state variables estimation error index, SEE (State Estimation Error), which is determined as follows: SEE = trace(G-1(x)), and the measured quantities estimation error index EE (EE = trace(H×G-1(x)×HT)). The lower the SEE and EE indexes, the more accurate the results of the power system state estimation are. Note that the matrix G-1(x) is actually the matrix of covariances of the state variables and the matrix H×G-1(x)×HT is the matrix of covariances of the measured quantities [9]. The SEE and EE indexes are the sums of the variances of the state variables and of the measured quantities.

Number of iterations The number of iterations is an essential factor which characterizes the time of calculation. It is related to the convergence of the calculation process.

6. ASSUMPTION FOR INVESTIGATIONS The investigations the results of which are presented in this paper, have been completed for the following assumptions: 1. The IEEE 14-bus test system is considered.

2. 10 power flows are generated for each operational mode of UPFC. The power flow is characterised by the n

index FL, which is defined as follows: FL   Pi i 1

m

 Pi , where: Ρ is the active power injection at the node i,  i i 1

Pi+ is a positive active power injection at the node i.

3. Each power flow serves as the basis for generation of 10 measurement data sets. Each measurement data item is burdened with a random error which has the standard distribution with zero expected value and the standard deviation σ, where: •   1/3  0,001  0,0025 FS  0,02  M  for the measurement data of active power; •   1/3   0,001 0,005 FS  0,02 M  for the measurement data of reactive power; •   1/3 0,0005  0,0025FS  0,003 M  for the measurement data of the voltage modulus, FS is the measurement range and Μ is the measured value [10, 11].

4. For the specified operational modes of UPFC, the power flows are calculated with the use of the UPFC, SSSC and STATCOM models or without any of them in the case, when UPFC is off.

5. The estimation calculations are conducted with the use of the state estimator, in which always a model of power system with UPFC is utilized. The state vector is considered in the Cartesian coordinate system.

6. UPFC is installed at the node 5 in the branch between the nodes 5 and 4 of the test system. The connection point of UPFC and of the power line which is between the nodes 4 and 5 is designated as the node 15. The purpose of UPFC is to increase the equivalent reactance of the branch between the nodes 4 and 5 to 0.1 p.u. and to maintain the voltage at the node 5 at 1 p.u.

7. UPFC in the SSSC mode increases the equivalent reactance of the branch between the nodes 4 and 5 to 0.1 p.u.

8. UPFC in the STATCOM mode maintains the voltage at the node 5 at 1.0 p.u.

Impact of UPFC Operational Modes on Power System State Estimation

9. One considers measurement data redundancy characterised by the coefficient rd, which is equal to 1.30 (43 measurements) and 3.12 (103 measurements). The coefficient rd is defined as rd = m/(2n-1).

10. For each value of rd, 500 sets of measurement locations are generated.

7. RESULTS OF INVESTIGATIONS The results of calculations are shown in Figs. 5 to 8 and in tables 1 and 2. In the carried out analysis, one considers the relative changes in the investigated parameters of the power system state estimation. These changes are calculated with the formula: δχ = (X – Xav)/Xsr, where: X is the considered parameter; Xav is the average value of the parameter X for all the conditions (including the UPFC operational modes) of the investigations the state estimation. The relative change δχ is expressed in percentage values.

The analysis of the relative changes of the investigated parameters indicated that the highest deviations from the mean value of the given parameter generally occur at extreme system loads. Given the absolute value, the deviations achieve as much as dozens percent. This can be observed in relation to the index cond(G). The mentioned deviation values are lower at the average system load.

The differences between the changes of δx for different operational modes of UPFC, are usually lower when system loads are not distinguished. In such a case, these differences do not exceed ca. 5%, but when for example the relative change δcond(G) (δcond) is investigated for FL = 0.108 and rd = 3.12, the observed difference amounts to 37.04%. The analysis of δSEE without distinguishing the system loads forms the basis for an unambigua)

b)

U

PF

C

is

U

PF

of f

C

is

of f

Fig. 5 The relative change of the estimation process conditioning index for different UPFC operational modes when (a) rd = 1.30, (b) rd = 3.12

a)

b)

U

U

PF

C

is

PF

of f

C

is

of f

Fig. 6 The relative change of the SEE ( δSEE) index for different UPFC operational modes when (a) rd = 1.30, (b) rd = 3.12

57

Tomasz Okoń / Wrocław University of Technology Kazimierz Wilkosz / Wrocław University of Technology

58

ous conclusion that the most favourable features of the state estimation for the SEE parameter occur when UPFC operates with full utilization of its functionality. This conclusion cannot be drawn on the base of the analysis of δcond(G) (δcond), δΕΕ, δl it. It can also be indicated that the state estimation for the STATCOM mode is most favourable for the change of δl it when the system loads are not discriminated.

8. FINAL NOTES The UPFC device can operate in different modes. It can perform its designed functions in full, only the STATCOM functions, only the SSSC functions or it can be off. This paper considers the power system state estimation with the use of the estimator in which a model of UPFC is taken into account but information on actual operational mode of UPFC is inaccessible. The analyses allow concluding that it cannot be unambiguously stated for which operational mode of UPFC the considered state estimation has the most favourable features. Only certain values of the system load index can be identified at which the state estimation is most favourable to the UPFC operational mode in which all functions of the system are fully utilized. Regarding other system loads, for at least one of the UPFC operational modes, features of the state estimation become more favourable. This fact can be explained by the high flexibility of the UPFC model utilized by the investigated estimator. The question arises whether consideration of different UPFC operational modes during the state estimation would improve its features or not. a)

b)

Fig. 7 The relative change of the EE ( δEE) index for different UPFC operational modes when (a) rd = 1.30, (b) rd = 3.12

a)

b)

Fig. 8 The relative change of the number of iterations in the state estimation process ( δl it) for different UPFC operational modes when (a) rd = 1.30, (b) rd = 3.12

Impact of UPFC Operational Modes on Power System State Estimation

59

Tab. 1. The minimum, maximum and mean relative change values of the considered parameters of the power system state estimation when rd = 1.30 δcond(G)

δSEE

δEE

δl it

Operational mode

min.

max.

mean

min.

max.

mean

min.

max.

mean

min.

max.

mean

UPFC is off

-46.6

35. 4

0.62

-50.0

53.1

2.65

-44.0

51.2

2.42

-1.95

10.7

0.77

STATCOM

-61.0

63.6

2.82

-57. 4

57.7

2.56

-52.1

52.9

1.45

-19.3

12.2

-0.82

SSSC

-47.9

30.9

-2.54

-52.0

45.0

-2. 47

-45.5

44.8

-1. 47

-1.97

9. 45

0.53

UPFC

-58.7

51. 4

-0.89

-58.1

48. 4

-2.75

-51.8

45.1

-2. 40

-14.5

11. 4

-0. 48

Tab. 2. Wartości minimalne, maksymalne oraz średnie względnych zmian wyróżnianych parametrów estymacji stanu systemu elektroenergetycznego, gdy rd = 3,12 Tryb pracy

δcond(G)

δSEE

δEE

δl it

min.

max.

mean

min.

max.

mean

min.

max.

mean

min.

max.

mean

UPFC is iff

-33.9

14.5

-2.63

-41.1

36. 4

1.67

-32.5

24.8

-0.02

0. 41

1.61

1.24

STATCOM

-48.1

41.7

2.01

-47. 4

40.9

2.12

-39.9

35.6

1.33

-24.0

1. 46

-1.23

SSSC

-33. 4

17.3

-1. 43

-42.7

31.2

-1.82

-33.0

23.0

-1.00

0.51

1. 41

1.23

UPFC

-44.7

33.9

2.05

-48.0

33.0

-1.97

-39.1

30. 4

-0.31

-24.0

1.36

-1.23

REFERENCES 1. Monticelli A., Electric Power System State Estimation, Proceedings ofthe IEEE, vol. 88, no. 2, February 2000, pp.

262–282.

2. Wu F. F., Moslehi K., Bose A., Power System Control Centers: Past, Present, and Future. Proceedings of the IEEE, vol.

93, no. ll, November 2005, pp. 1890–1908.

3. Acha E., Fuerte-Esuivel CR., Ambriz-Perez H., Angeles-Camach C, FACTS: Modelling and Simulation in Power Networks. Chichester, John Wiley & Sons, 2004.

4. Mehraban A.S., Edris A., Schauder CD., Provanzana J.H., Installation, Commissioning, and Operation of the World’s First UPFC on the AEP System. International Conference on Power System Technology, 1998, vol. 1, pp. 323–327 5. Zhang, X.P: Advanced Modeling of the Multicontrol Functional Static Synchronous Series Compensator (SSSC) in Newton Power Flow, IEEE Trans, on Power Systems, vol. 18, no. 4, November 2003, pp. 1410–1416.

6. Schweppe FC, Wildes J., Power System Static-State Estimation. Part I–III. Trans, on Power Apparatus and Systems, vol. 89, no. 1, January 1970, pp. 120–135.

7. Zamora E.A., Fuerte-Esquivel CR., Static State Estimation of Power Systems Containing Series and Shunt FACTS Controllers. 15th PSCC, Liege, 22–26 August 2005.

8. Gu JW., Clements K.A., Krumpholz G.R., Davis R/V., The Solution of Ill-Conditioned Power System State Estimation Problems Via the Method of Peters and Wilkinson. IEEE Trans, on Power Apparatus and Systems, vol. 102, no. 10, October 1983, pp. 3473–3480.

9. Larson R.E., Tinney W.F, Peschon J., State Estimation in Power Systems. Part I: Theory and Feasibility. IEEE Trans, on Power Apparatus and Systems, vol. 89, no. 3, March 1970, pp. 345–352.

10. Dopazo J.F, Klitin O.A., Stagg GW., Van Slyck L.S., State Calculation of Power Systems From Line Flow Measurements. IEEE Trans, on Power Apparatus and Systems, vol. 89, no. 7, September 1970, pp. 1698–1708.

11. Dopazo J.F, Klitin O.A., Van Slyck L.S., State Calculation of Power Systems from Line Flow Measurements, Part II. IEEE Trans, on Power Apparatus and Systems, vol. 91, no. 1, January 1972, pp. 145–151.

60

Adam Smolarczyk / Warsaw University of Technology

Authors / Biographies

Adam Smolarczyk Warsaw / Poland Graduated from the Faculty of Electrical Engineering, Warsaw University of Technology, obtaining his Master of Science Engineer degree (1995), after which he received his Ph.D. (1999). In December 1999 he started working at the Institute of Electrical Power Engineering. He is currently an assistant professor at the Faculty of Electrical Engineering, Warsaw University of Technology.

He is an author and co-author of research reports on digital power relays. His research interests are related to digital power protection automatics and the modelling of phenomena occurring in power systems.

Setting Methods of The Impedance Type Power Swing Blocking Functions Applied in Distance Protections

SETTING METHODS OF THE IMPEDANCE TYPE POWER SWING BLOCKING FUNCTIONS APPLIED IN DISTANCE PROTECTIONS Adam Smolarczyk / Warsaw University of Technology

This article was prepared under research work financed with funds for research in 2008-2010 as research project No. N N511 358234.

1. INTRODUCTION Power swings are phenomena occurring quite often in the power system after the elimination of faults.

During power swings the measured impedance seen by the distance protection of the particular line decreases to a value corresponding to its measurement zones, which can lead to unnecessary disconnection of the line.

Such disconnection can significantly weaken the transmission network, causing an overload of other lines and, consequently, their disconnection.

In order to avoid unnecessary operation of distance protection during power swings, they are equipped with power swing blocking elements (power swing blocking functions). Power swing blocking functions are part of the protection system associated with power swings in power systems [1, 2].

The types of used power swing blocking functions and power swing detection methods used in them are described in the literature [1, 3]. The description in article [4] shows the changes in measured impedance during power swings on the impedance plane in relation to distance function zones. Fig. 1 shows examples of measured impedance occurring during asynchronous and synchronous swings in distance protection zones. As can be seen, during asynchronous swings (Fig. 1a) impedance Z(t) passes on a curve through the outer zone B and the inner zone F of the blocking function and zones 1 and 2 of the distance protection. During synchronous swings (Fig. 1b) the impedance Z(t) reaches the inside of zone 1 through the outer zone B and the inner zone F of the blocking function, turns round and exits on the same side of the outer zone B of the blocking function.

Z(t) – type power swing blocking functions distinguish between the change in measured impedance caused by the power swing and the change in impedance caused by fault, and in the event of power swings they block the selected operation zones of distance protection. The distinction between fault swings is based on measuring the speed at which the measured impedance passes between the outer zone B of the blocking function and its inner zone F (which is usually the starting zone of the distance protection function). When this time (Δt in Fig. 1) is longer than the set value, a signal is transmitted to block the selected distance protection zones. During faults, the speed of impedance change is high and the passage time is shorter than the set value. The blockade of measuring zones of the distance protection is not initiated.

Abstract The article describes the methods of setting the Z(t) type impedance power swing blocking functions.

Blocking functions of this type are commonly used for blocking the operation of distance protection during power swings. The methods for setting the basic parameters

are described, which include: resistance and reactance reaches, setting the timer associated with stimulation of the blocking function (power swing detection time) and setting the timer associated with blocking of the operative power swing blocking function (reset time).

61

Adam Smolarczyk / Warsaw University of Technology

62 a)

b) X

X F

B

F

2 1

B

2 t

� �� �

1 t

� �� �

R

R

Fig. 1. Example changes in measured impedance on the impedance plane during swings: a) asynchronous, b) synchronous

The described principle of operation of power swing blocking function, based on measuring the speed of transition of impedance vector between two characteristics, was used in many previously manufactured electro-mechanical and electronic (static) protections. This principle is also used (of course, taking into account additional criteria) in current digital protection solutions.

2. SETTING THE IMPEDANCE REACHES AND TIME ELEMENTS OF Z(t) – TYPE BLOCKING FUNCTIONS A major problem with setting the Z(t) – type power swing blocking functions is setting the timer needed to distinguish between power swings and faults. This setting is not easily calculated and depends largely on the system where the blocking function is to be installed. It may be necessary to perform extensive stability tests to determine the fastest power swings needed to properly determine the blocking function settings for the particular system. The value of slip between the two systems is a function of accelerating motion and the system inertia. The distance relay operation algorithm can not determine the slip value analytically because of the complexity of the power system. However, when analysing the stability of the system and analysing the changes in angles as a function of time, it is possible to estimate the average slip value in the particular system.

Such an approach may be appropriate for the systems whose slip frequencies do not change significantly when the analysed system switches to asynchronous operation. However, in many systems where slip frequency increases significantly after the first and subsequent asynchronous turn, the set fixed impedance interval between the outer and inner characteristics of the power swing blocking function (distance between the characteristics B and F in Fig. 1) and the set time of transition of measured impedance between these characteristics – may be inadequate to block the distance function zones properly.

The following should be done in order to set the Z(t) type power swing blocking function: • Setting the resistance reach of the blocking function outer zone outside the operation zones of distance protection, which will be blocked during power swings. Setting the resistance reach of the outer blocking function zone as close to the area of maximum line load as possible (after taking into consideration the appropriate safety margin). Resistance reaches of both zones (outer and inner zones) should be set with an appropriate safety margin in relation to the area of maximum load flow and the resistance range of the distance function zone with the greatest reach (zone that is to be blocked). Normally, a 20-percent safety margin ensures the proper operation of the blocking function [6]. The literature [7] recommends that the load flow resistance (Rload = Zload x cosφload) is at least 20 percent higher than the zone with the highest resistance range (Rp), i.e. Rload ≥ 1.2Rp . To prevent the distance function operation during the motion changes in the line load, the reach of the blocking function outer zone should be 10÷20 percent higher than the reach of the distance function zone with the highest resistance reach.

Setting Methods of The Impedance Type Power Swing Blocking Functions Applied in Distance Protections

• In the case of typical impedance characteristics of the Z(t) – type blocking functions, the reactance reaches of the “top” and “bottom” of the blocking function outer characteristics are not critical. It is recommended [6] that these settings are close to the maximum reaches of settings for the particular relay, or that are 2÷3 times greater than the reactance reaches of the most further distance function zone set. According to [6] the "top" and "bottom" of the blocking function outer characteristics should be set to the same values. In addition, the resistance and reactance settings of the blocking function outside zone are to be set to no less than 130 percent (typically 135÷150 percent) of the setting for resistance and reactance zones of the inner blocking function.

• Usually zone 1 of the distance protection and the zone (usually 2) that is used in cooperation between the connection and the protection on the other end of the line [6] are blocked against power surges.

Zone 1 is not blocked during power swings – according to [7] – when it has a short resistance reach and when the impedance enters its area (large power angles near 180 deg) it is obvious that the system will lose its stability. Blocking the higher (than zone 2) distance function zones is not required if the duration of the particular zone is longer than the expected time during which the measured impedance occurs in the area of the zone during power swing. The currently used methods for selecting the blocked zones are described in detail in the literature [1, 2, 5].

• Based on the resistance reaches of the outer and inner zones of the blocking function, it is possible to calculate the setting of time ΔT for passing of measured impedance between the outer and inner zones of the blocking function, after which it is activated. The following data is assumed for this purpose: reactance appropriate on the left side of the relay point Xa and reactance appropriate on the right side of the relay point Xb (Fig. 2a). Angles δ1 and δ0 are angles of powers, for which the resistance range of inner and outer zones of the blocking function occur, as shown in Fig. 2b. In addition, the calculation of setting for time ΔT assumes the maximum slip frequency Δƒmax that may occur in the system. Of course, the best method of determining the maximum slip frequency in the particular section of the power system in transition is the stability test using simulation programs. However, if it is not possible, usually the maximum slip frequency between 4 and 7 Hz [5, 6] is assumed. In some sources 8 Hz, 10 Hz, or even 15 Hz [1] are assumed as the maximum slip frequency that may occur during swings.

In order to calculate the setting ΔT the relevant calculations are presented later in this article. a)

b)

� Inner zone of the ���������� blocking function ��������������

��

� ��

� � �

��

Outer zone of the ���������� blocking function ��������������

��

� ��



�



�����

� �� ��

� ��

�

�

Fig. 2. Method of determining time Δt of transition between the inner and outer zones of the power swing blocking function: a) the analysed two-machine system, b) presentation of equivalent reactances of the two-machine system and zones of the power swing blocking function on the impedance plane.

63

Adam Smolarczyk / Warsaw University of Technology

64

To calculate time Δt of transition of impedance vector between the inner and outer zones of the blocking function (Fig. 2b) the known dependence on slip Δω should be used:

d� rad � rad   �  2�  f dt t

(1)

where Δδrad = (δ1 – δ0) means the difference between the power angles in Fig. 2b in [rad], Δω – means the slip in [rad/s], Δƒ– means the slip frequency in [Hz], Δt – means the time of transition of impedance vector between the outer and inner zone of the blocking function [s].

After simple transformations and conversion of angles from radians to degrees (marked as “deg” in the article), the formula for the time Δt of transition of measured impedance vector between the outer and inner zone of the blocking function is as follows:

t 

� deg 360  f



� 1� 0 360  f

(2)

where angles Δδdeg , δ1 , δ0 are expressed in [deg], and the slip frequency Δƒ in [Hz].

In order to calculate the setting for time ΔT of the power swing blocking function for the maximum slip frequency Δƒmax the following formula should be used:

T 

� 1� 0 360  f max

(3)

where: time ΔT is expressed in [s]; angles δ1 , δ0 are expressed in [deg], Δfmax – the maximum slip frequency is expressed in [Hz].

For time ΔT to be expressed in multiplicities of the nominal frequency period (f = 50 Hz) the right side of equation (3) should be divided by the duration of one period (20 ms for ƒn = 50 Hz) or multiplied by the converse of the period, i.e. frequency ƒn (in the English literature time ΔT is usually given in multiplicities of the period for frequency ƒn).

The setting for time ΔT the power swing blocking function should be “reasonable”. The selected setting should allow (be high enough) the power swing blocking function algorithm to take the correct decision (by blocking the distance function zones during power swings and not blocking the zones during faults). The literature [6] recommends setting time ΔT of the power swing blocking function in the range of 1.5 to 2.5 times the nominal frequency period (30÷50 ms for ƒn = 50 Hz) to ensure that the blocking function is not activated in situations not associated with power swings. The text [7] proposes the setting for time ΔT ranging from 20 to 40 ms.

Many manufacturers of protection equipment sets permanent resistance intervals between the inner and outer zones of the blocking function, and the user can not change this setting. In this case, to calculate time Δt of transition of impedance vector between the zones (inner and outer zone) for a specific slip frequency Δƒ, angles δ 1 , δ0 in formula (3) should be calculated. The following formula can be used for this purpose:

Rs (t )  X a  X b 

kE

k E2  2k E cos�  1

sin �

(4)

In this formula X = (Xa + Xb) means equivalent reactance of the entire system.

Formula (4) shows that resistance Rs measured at the relay point P (Fig. 2a) does not depend on the location of the relay point and the ratio kX = Xa / Xb . The resistance measured by the relay depends on the ratio of electromotive forces kE = Ea / Eb and the current value of the load angle δ. When kE = 1 (source voltage modules are equal), formula (4) is simplified to the following formula:

Setting Methods of The Impedance Type Power Swing Blocking Functions Applied in Distance Protections

Rs (t ) 

X a  X b  2

 ctg

� X �   ctg 2 2 2

(5)

After transformation of this formula, angle δ can be calculated according to the formula:

 2 R (t )  �  2  arc ctg s   X 

(6)

where angle δ is expressed in [deg].

Thus, after substitutions, time Δt of transition between two (outer and inner) zones of the power swing blocking function can be calculated according to the formula:

t 

� 1� 0 2  360  f 360  f

  2R   2R    arc ctg s1   arc ctg s0    X   X  

(7)

where: time Δt is expressed in [s]; angles δ 1 , δ0 are expressed in [deg], Δƒ – slip frequency is expressed in [Hz], Rs1 , Rs0 , X are expressed in [Ω].

Some Z(t) – type blocking functions have permanent resistance intervals ΔR = (Rs0 – Rs1) set by the device manufacturer, or can not be set freely due to strong motion load of the line between the inner and outer zones of the blocking function and the set time ΔT of transition between these zones in order to distinguish between faults and power swings. In this case, the maximum slip frequency Δƒmax , which is detected by the blocking function for specified settings and system parameters, may be calculated. It can be done using the following formula:

f max 

 2  2R   2R    arc ctg s1   arc ctg s0   360  T   X   X 

(8)

As can be seen, in addition to the above parameters this frequency depends on the equivalent reactance X = (Xa + Xb) of the entire system where the blocking function is installed.

In complex power systems it is very difficult to calculate exact impedance value Xa and Xb (Fig. 2a), which are needed to set the resistance reach of the blocking function outer and inner zones, and to set the blocking function timer. Reactances (impedances in the general case) of the system vary depending on the grid configuration, e.g. in the case of switching on the generating units, and switching off or on the line in the system. The system impedance (as seen from the relay point) may vary significantly during large disturbances (and after their elimination). It should be noted, however, that setting the power swing blocking function is very simple (as shown above) if the impedance of the system do not change and if it is easy to calculate them. Usually, however, complex system stability tests should be performed to analyse various situations that may occur, and to select appropriate equivalent impedances of the system, needed to set conventional Z(t) type power swing blocking functions. Such tests are very expensive and can never predict every situation that may occur in the system.

65

66

Adam Smolarczyk / Warsaw University of Technology

3. RESET TIME OF THE OPERATIVE POWER SWING BLOCKING FUNCTION One of the parameters set in the power swing blocking functions is reset time (maximum blocking time, time after which operative power swing function is blocked). Resetting is based on unconditional automatic release of the blocking function (regardless of other criteria) when the set time passes (referred to as reset time).

The reset time should be long enough to ensure that the distance protection is not unblocked (blocking function release) during the typical synchronous and asynchronous swings.

When setting the reset time it is helpful to know how long the impedance trajectory remains inside the distance protection zone during typical synchronous and asynchronous swings. Analysis of issues regarding the time of the impedance trajectory in the characteristics of the distance protection can be found in the studies [1, 4] and will not be quoted in this article. The formulated recommendation is as follows: the time during which the impedance trajectory Z(t) remains inside the distance protection characteristics for synchronous swings may be much longer than in the case of asynchronous swings.

The literature [1, 4] suggests that the reset time should be set to no less than 2 s for the protection of lines inside the power system and no less than 5 s for the protection of intersystem lines because too quick a release of the blocking function would cause unnecessary activation of the protection and disconnection of the intersystem line.

According to the recommendations available in the manufacturers’ descriptions of protection equipment it is typically assumed (factory settings) that the reset time is 2 s. Equipment manufacturers usually stress the fact that this time was set in electromechanical and electronic power swing blocking function solutions and older digital solutions.

Resetting (blocking) operative power swing function is used for the two following purposes: • Unblocking the distance protection function if for some reason the unblocking when the impedance trajectory exits from inside the power swing blocking function inner characteristic fails.

• Unblocking the distance protection function (when the additional criteria in the power swing blocking function algorithm fail to do so) if the impedance trajectory did not exit from inside the power swing blocking function characteristics at the set time, for example as a result of another three-phase fault occurring during power swings.

In the latter case, the unblocking allows the distance protection activation when the set reset time passes.

A major challenge faced by the algorithms used in the power swing blocking functions is the detection of internal three-phase faults occurring during power swings. In order to increase reliability of detecting an internal fault occurring during power swings, modern protection equipment manufacturers have introduced additional criteria and algorithms enabling unblocking of distance protection (blocking function “release”) without having to wait for the reset time to pass. The above-mentioned are, e.g. algorithms for tracking the continuity of the measured signals. If the signal continuity is disturbed by faults, the distance function of the protection in unblocked. Due to the use of these new algorithms for fault detection, the role of resetting in modern digital protective equipment is less significant than in electromechanical or electronic protection. For example, a Siemens 7SA511 relay includes the option for setting this time, though it is factory set to infinity. In a 7SA513 relay, the reset time can be set by the persons servicing the relay, not by the users of the relay (an additional service password is needed). In this case, the reset time is also factory set to infinity. In a 7SA522 relay it is not possible to set the reset time at all, either by an ordinary user or by a service technician.

If it is not possible to set the reset time in the relay, and the user of the device does not trust the algorithms used in the protection device and wants the reset time to be counted, the solution to this problem would be to form a proper logic (connecting the inner signals of the relay between each other) using the so-called internal programmable logic of the device. This logic is available in digital protection devices manufactured by leading brands.

Other manufacturers, despite the use of more and more efficient algorithms for the detection of various disturbances (including three-phase faults) that occur during power swings, still include the option for setting the reset time [1]. Therefore, it is important to explain how this time is counted in a particular protection device solution. This is due to the fact that in various versions of firmware in the relays, the manufacturers may change (and often change) algorithms of various functions (including power swing blocking functions), and thus the reset time can be counted in different ways [8].

Setting Methods of The Impedance Type Power Swing Blocking Functions Applied in Distance Protections

According to the author of this article, the best way to obtain an answer to how accurate the reset time calculation is, is to perform tests on properly configured protection equipment, using microprocessor testers equipped with appropriate software. This is the surest way to check whether the given protection function (in this case, the power swing blocking function) works as expected. All the more so that the block diagrams of functions and their descriptions available in user manuals are often quite brief or difficult to analyse.

4. SUMMARY Z(t) – type impedance blocking functions are still used for blocking the operation of distance protection during power swings. The most important parameters that are set in the blocking functions of this type are resistance and reactance reaches of the inner and outer zones of the blocking function, setting for time of transition of impedance between the zones (in order to distinguish between faults and power swings), and setting for reset time (blocking the operative power swing blocking function after a specified time).

There are different approaches to power blocking function settings (which is probably a result of the algorithms used in them). In some manufacturers’ devices (e.g. relay REL531 produced by ABB), the number of parameters associated with setting power swing blocking functions is high compared to the above mentioned.

Their proper setting (other than factory settings) requires considerable experience. Other manufacturers set blocking functions on their own (e.g. Siemens relay 7SA522), and users do not have access to settings (of course apart from the basic settings, such as selection of the zone to be blocked during power swings).

Due to the large complexity of digital protection equipment, the most certain way to see if the power swing blocking function works according to the user’s expectations is to perform tests using specialist equipment, including microprocessor testers, amplifiers of current and voltage signals and software for simulating the dynamic conditions in power systems.

REFERENCES 1. Machowski J., Smolarczyk A., Brzeszczak L., Opracowanie zasad nastaw blokad przeciwkołysaniowych zabezpieczeń pod kątem odbudowy systemu, The Institute of Electrical Power Engineering, Warsaw University of Technology, research and development study commissioned by PSE-Operator SA, Agreement No. SR/RB/IS/008/05, Warsaw 2005.

2. Machowski J., Selektywność działania zabezpieczeń w trakcie kołysań mocy w systemie elektroenergetycznym, Part 3, Zabezpieczenia rozcinające sieć przesyłową i systemy zabezpieczeń związanych z kołysaniami mocy, Automatyka Elektroenergetyczna, Issue 2/2007.

3. Smolarczyk A., Blokady przeciwkołysaniowe stosowane w zabezpieczeniach odległościowych, Wiadomości Elektrotechniczne, Issue 10/2010.

4. Machowski J., Selektywność działania zabezpieczeń w trakcie kołysań mocy w systemie elektroenergetycznym, Part 2, Zabezpieczenia odległościowe i ich blokady przeciwkołysaniowe, Automatyka Elektroenergetyczna, Issue 1/2007.

5. IEEE PES: Power swing and out-of-step considerations on transmission lines. A report to the Power System Relaying Committee of IEEE Power Engineering Society, Report located at: http://www133.pair.com/psrc/ (2005, Published Reports/Line protections).

6. Mooney J., Fischer N., Applications guidelines for power swing detection on transmission systems, SEL 2005, 20050920, TP6228-1.

7. Ziegler G., Numerical Distance Protection. Principles and Applications, Siemens, Erlangen 2006.

8. Smolarczyk A., Nastawianie impedancyjnych blokad przeciwkołysaniowych typu Z(t), Automatyka Elektroenergetyczna, Issue 3/2010.

67

68

Grzegorz Widelski / ENERGA-OPERATOR SA Sławomir Noske/ ENERGA-OPERATOR SA

Autorzy / Biografie

Grzegorz Widelski Gdańsk / Poland

Sławomir Noske Elbląg / Poland

The Head of the Property Development Department in ENERGA-OPERATOR SA. Has been employed in the energy company since 1999. Graduate of Faculty of Electrical and Control Engineering at Gdańsk University of Technology, speciality in electric power. He completed postgraduate studies in the field of power distribution system management at his home university. Currently engaged in MBA management studies, organized by the Gdańsk Foundation for Management Development.

Works as a chief engineer for research and development for ENERGA-OPERATOR SA. Has been employed in the energy company since 1991.

Graduate of Poznań University of Technology, Faculty of Electrical Engineering. He completed postgraduate studies in the legal and management field at Gdańsk University of Technology and MBA management studies, organized by the Gdańsk Foundation for Management Development.

In 2006 he was admitted into the Ph.D. program “Diagnosis of medium voltage cable lines with the use of partial discharge test using the self-damping voltage wave method” .

Towards Smart Gird – Pilot Project “Smart Peninsula”

TOWARDS SMART GIRD – PILOT PROJECT “SMART PENINSULA” Grzegorz Widelski / ENERGA-OPERATOR SA Sławomir Noske / ENERGA-OPERATOR SA

1. INTRODUCTION Looking at the distribution network it may be noticed that there have been no significant changes in the power system for over thirty years. Of course, there has been some development in the equipment used in the power system, but it has not introduced significant changes in the structure of the power system. The situation is different in the case of energy, both in terms of applied technologies, changing lifestyles, as well as evolving environmental awareness and transformation of the legislation. Changes in telecommunications and information technology (both hardware and software) can be regarded as revolutionary. They affect our private lives and the way businesses operate. The consumer expectations regarding quality and reliability of the supplied energy and the method of its use are significantly growing. There have also been significant changes regarding the law: among others, the European Parliament adopted the package of laws “3x20”, which assumes that by 2020 the green energy share in electricity production will increase to 20 percent, greenhouse gases will be reduced by 20 percent, and electricity consumption will be reduced by 20 percent.

In the face of such great changes it seem necessary to also develop new solutions in the field of energy itself. New solutions and the philosophy of rebuilding the distribution system with the use of IT infrastructure are defined as the Smart Grid technology. Smart Grids are considered to be the primary means for achieving an efficient and safe power system. These solutions are seen as an opportunity for changing the production, supply and use of electricity. Smart Grid intelligently integrates the activities of all participants in the processes, i.e.: generation, distribution and use, in order to provide electricity in an economical, consistent and safe manner.

Smart Grid is currently the main area of interest in the development of power grids; this trend can be observed throughout the world.

2. TOWARDS SMART GRID Realizing the need to use the opportunities offered by the Smart Grid, ENERGA-OPERATOR SA is planning to transform the existing traditional grid into a Smart Grid. Two key projects currently being implemented in this area are: the construction of Advance Metering Infrastructure (AMI) and the construction of Smart Grid in the pilot area.

AMI Project In 2010, remote reading of meters of industrial consumers was introduced. The project covers 18 500 business consumers. This way, consumers can, for example, gain access to the readout data, and take actions to optimize the electricity consumption. The implemented system can read data at 15-minute intervals. Communication in the system is realized through GPRS. It is the largest project of this type in Poland.

In another project, ENERGA-OPERATOR SA plans to implement AMI in the entire operation area of the company from 2011 to 2017. The project includes about 2.5 million municipal consumers and about 0.3 million

Abstract ENERGA-OPERATOR SA has introduced changes in the organization and management of network assets, implemented an integrated IT system supporting the management of network assets. The changes above are the basis for further steps in the field of grid management

and development. The planned changes are aimed at switching from a traditional grid to a Smart Grid. Two key projects currently being implemented in this area are: the construction of Advance Metering Infrastructure (AMI) and the construction of a Smart Grid.

69

70

Grzegorz Widelski / ENERGA-OPERATOR SA Sławomir Noske/ ENERGA-OPERATOR SA

business consumers. Implementation of the system will take place on three levels: the implementation of smart metering, the implementation of telecommunication solutions (depending on location - PILC, GPRS and WiMAX technologies) and the implementation of information systems. The first stage, implemented in 2011, assumes the implementation of AMI in three pilot areas and will cover over 100 thousand consumers. One of the pilot areas is the Hel Peninsula, the area where the Smart Grid construction project began. AMI implemented in the field of telecommunications is being prepared so as to meet the requirements of the future Smart Grid. When implementing AMI, ENERGA-OPERATOR SA does not focus solely on achieving benefits for itself. According to arrangements with the Energy Regulatory Office, some of the benefits obtained from the implementation of AMI will be passed on to the electricity consumers.

Smart Grid Project ENERGA-OPERATOR SA perceives the implementation of Smart Grid as the next stage of the distribution system development. This is the first implementation of Smart Grid in the area of the distribution network in Poland. The Hel Peninsula was chosen as the pilot area. It covers about 150 km of MV lines, 80 MV/LV substations, and 100 km of LV lines. The grid supplies about 15 thousand consumers. The project is divided into two stages.

The first stage is based on preparing the concept for switching from traditional grid to Smart Grid. Its completion is scheduled for October 2011. The second stage includes the construction of Smart Grid in the pilot area. The beginning of this stage is planned for 2012.

Basic requirements related to Smart Grid include: • Construction of a model control system for the Smart Grid area. The primary function will be an integrated system for voltage control and reactive/active power management, aimed at enabling the adjustment of the grid load level to the distribution capabilities and energy parameters at a particular time (by adjusting the load or generation characteristics of individual connected entities to the grid conditions) • Creating opportunities for maintaining a separate area in the island operation in the case of power failure in the national power system • Creating opportunities for cooperation between Smart Grid and smart buildings equipped with microgeneration • Implementation of pilot installations for charging electric cars in a manner adapted to the existing conditions and the distribution network load • Equipping the grid with an adequate infrastructure, including measurement systems for remote reading of measuring data, and control of energy supply to consumers. This solution is to enable the introduction of new products and services by the companies engaged in electricity trading.

In the first stage, ENERGA-OPERATOR SA cooperates with the Institute of Power Engineering, Gdańsk Branch. A document containing key information necessary for the beginning of the grid construction was prepared for the purpose of implementing Smart Grid in the pilot area: “The concept of construction and implementation of Smart Grid solutions in the ENERGA-OPERATOR SA Grid on the Hel Peninsula.” This document covers the following subjects: • the Smart Grid construction and operation concept • model tests of the grid operation • feasibility study for the implementation of the Smart Grid project.

The Smart Grid construction and operation concept includes an analysis of the current condition of the power infrastructure, its load, the connected sources, as well as the characteristics of the connected consumers. For this purpose, the inventory of the current grid condition was made in the scope relevant to the project, including: conventional and renewable generation, consumption, including consumption suitable for control, automatics and network protection, cable and overhead power supply lines (HV and LV grids), transformers, capacitors, etc., and the existing telecommunications infrastructure.

On the basis of the inventory, a quantitative and qualitative assessment of the existing power and telecommunications infrastructure as well as the connected sources will be performed regarding the use in the implementation of Smart Grid on the Hel Peninsula.

Inspection of the functional aspects will take into account the following:

Towards Smart Gird – Pilot Project “Smart Peninsula”

• The use of AMI systems in Smart Grid • Prosumer and active network in Smart Grid solutions • Smart buildings as part of Smart Grid • Electric cars as part of Smart Grid • Energy sources connected to the distribution network (DER) as part of Smart Grid • Consumption (DSM, Demand Response) • Management and control in microgrids • Improving the reliability of microgrid supply • Independent operation of microgrids (island operation).

The inspection of currently available technologies in terms of suitability for use in Smart Grid will be also carried out as part of this task. Particular emphasis will be placed on telecommunications and information technology due to its important role in constructing Smart Grid. An assessment will be performed regarding the ability to use the above technology in the project for the construction of Smart Grid on the Hel Peninsula. The technical and functional concept for implementation of Smart Grid on the Hel Peninsula will be prepared based on the developed material.

The concept will include such elements of Smart Grid as: • management system • monitoring • grid control • charging system for electric vehicles • island operation.

The second step will include model tests for the grid operation and the development of algorithms for the control of Smart Grid on the Hel Peninsula. Algorithms will be presented in descriptive form, and, if possible, should be tested on a simulation model, and will also be formulated in the form of a code suitable for inclusion in the simulation model in order to perform tests.

The completion of work included in the first stage will be followed by preparation of a feasibility study on the implementation of the Smart Grid project, which will be the basis for moving to the second stage of the Smart Grid construction on the Hel Peninsula. The study will include, among other things, the project implementation timetable.

3. SUMMARY According to the strategic objectives, ENERGA-OPERATOR SA intends to participate in the development of “Smart Energy” in a broad sense, in practical terms meaning the construction of an integrated supply and demand market, including Smart Grid. These activities are to transform the distribution network, bringing benefits to all participants of the energy market. The expected effects of the Smart Grid implementation will include: • more efficient management of the power grid • reduction of network losses • preventing and minimizing emergency situations in the distribution network • possible inclusion of any source of renewable energy in the grid • maintaining high quality electricity offered to consumers • expanding the opportunities regarding new products and services offered to consumers in order to actively participate in the demand management (the idea of prosumer) • using the opportunities brought by new technologies in the area of smart buildings (e.g. new household appliances, electric cars) • environmental protection through the promotion of distributed resources.

71

,1)250$7,21 )25 $87+256 :H DFFHSW RQO\ WKH DUWLFOHV WKDW KDYH QHYHU EHHQ SXEOLVKHG EHIRUH (DFK WH[W VHQW WR $FWD (QHUJHWLFD LV VXEMHFW WR VFLHQWLğF UHYLHZ 7KH HGLWRULDO PHHWLQJ GHFLGHV ZKHQ WKH WH[WV DUH SXEOLVKHG :H GR QRW VHQG WKH WH[WV EDFNV WR WKH DXWKRUV 3OHDVH VHQG WKH PDWHULDOV ZKLFK PXVW FRQVLVW RI WKH HQWLUH SDFNDJH RI IRXU FRPSRQHQWV DUWLFOH DEVWUDFW ELEOLRJUDSKLFDO QRWH SKRWRJUDSK RI WKH DXWKRU HOHFWURQLFDOO\ WR WKH IROORZLQJ DGGUHVV UHGDNFMD#DFWDHQHUJHWLFD RUJ NOTE! ( PDLO FRQWHQW PXVW LQFOXGH FRQWDFW GDWD JLYHQ QDPH DQG VXUQDPH GHJUHH WHOHSKRQH QXPEHU VWDWLRQDU\ DQG PRELOH DQG H PDLO DGGUHVV 1. ARTICLE ľ 7H[W OHQJWK 1RW PRUH WKDQ VWDQGDUG SDJHV FKDUDFWHUV VL]H LQWHUOLQHV FROXPQ ľ )RUPDW :25' ğOH DQG 3') DV D PXVW ľ )RUPXOD SDWWHUQV 3OHDVH EH YHU\ FDUHIXO ZLWK SXQFWXDWLRQ VWDQGDUGV ķ×” LV XVHG DV D VLJQ RI multiplication ([DPSOHV

ľ )RRWQRWHV $W WKH ERWWRP RI HDFK SDJH ([DPSOHV -RXUQDO RI ODZV 1R SRV RI 2FWREHU 7KH H[DPSOHV DUH GHDOW ZLWK LQ WKH ERRN )LHGRU % *UDF]\N $ -DNXEF]\N = 5\QHN SR]ZROHñ QD HPL VMÚ ]DQLHF]\V]F]Hñ QD SU]\NïDG]LH 62 Z HQHUJHW\FH SROVNLHM >3ROOXWLRQ (PLVVLRQ $OORZDQFHV 0DUNHW LQ 3ROLVK 3RZHU (QJLQHHULQJ RQ WKH ([DPSOH RI 62 @ :\GDZQLFWZR (NRQRPLD L ¥URGRZLVNR %LDï\VWRN SS ļ ľ /LWHUDWXUH $W WKH HQG RI WKH WH[W ([DPSOHV /DUVHQ ( 9 6ZDQQ ' $ $SSO\LQJ 3RZHU 6\VWHP 6WDELOL]HU ,((( 7UDQV 3RZHU $SSDU 6\VW YRO SS ļ 0DGDMHZVNL . 6REF]DN % 7UÚEVNL 5 3UDFD RJUDQLF]QLNöZ Z XNïDGDFK UHJXODFML JHQHUDWRUöZ V\QFKUR QLF]Q\FK Z ZDUXQNDFK QLVNLFK QDSLÚÊ Z V\VWHPLH HOHNWURHQHUJHW\F]Q\P PDWHULDï\ NRQIHUHQF\MQH $3( ij *GDñVN

2. ABSTRACT ľ 7H[W OHQJWK 1RW PRUH WKDQ FKDUDFWHUV ZLWKRXW VSDFHV

ľ )RUPDW :25' ğOH DQG 3') DV D PXVW %,%/,2*5$3+,&$/ 127( ľ 7H[W OHQJWK &D FKDUDFWHUV ZLWKRXW VSDFHV

ľ )RUPDW :25' ğOH DQG 3') DV D PXVW 3+272*5$3+ ľ )RUPDW &RORXU RU EODFN DQG ZKLWH SKRWRJUDSK MSJ RU WLII GSL

etica.org

energ www. acta