Dynamic Study Model For The Interconnected Power System Of Continental .

1 and steady-state data for all generation units for a peakdemand case in 2020. The model information basis descendsfrom the Network Modeling Database (NMD) of ENTSO-Eand serves as the planning basis. The use of this databasehas the advantages of an agreement between TSOs for systemplanning purposes and detailed information about generationmix and corresponding energy sources for steady-state analysis. The model is structured into 26 areas corresponding tothe included countries. It comprises 21,382 nodes and 10,829generators.In order to prepare the model for dynamic studies andavoid sharing confidential information, the load flow modelunderwent some adaptation steps, before it was distributed: adding of step-up transformers for generators directlyconnected to the transmission level, reduction of parallel lines, reduction of coupled busbars, aggregation of multiple loads at busbars and aggregation of multiple generation blocks at busbars.All these adjustments were done without influencing theload flow result.B. Standard dynamic modelsThe representation of system dynamics is accomplished byusing standard dynamic models for synchronous machines andtheir control devices. Table I gives an overview and referencesfor the used standard dynamic models for all generation units,which fulfill the defined criteria. The choice is based onprinciples with the aim to use commonly-used models withsimple structure in order to facilitate the tuning process. Thetable also specifies the variable model parameters, that will betuned at a later step to get the desired system behavior.TABLE I: Standard dynamic models for generation units and controllerselementsynchronousmachine (SM)governor (GOV)automatic voltageregulator (AVR)power systemstabilizer (PSS)modelvariable parameter for tuningstandard model,as defined in [4]TGOV1,as defined in [5]SEXS,as defined in [6]PSS2A,as defined in [7]acceleration time constant TAfor complete rotating unitsdroop R,time constants TG2 , TG3gain Kgain KS1To guarantee a common basis for the Dynamic StudyModel implementation in the different simulation tools, aharmonization process for the standard dynamic models aswell as for the load flow model was done at the beginningof the DSM development. For the comparison of the standarddynamic models a one-machine test system as shown in figure4 was introduced. It comprises a synchronous machine (SM)with a governor (GOV), an automatic voltage regulator (AVR)and a power system stabilizer (PSS). This generation unit isequipped with the defined standard dynamic models from tableI with default parameters. There are also a step-up transformer,an infinite busbar, loads and switches for test case selection.Beside a comparison of initialization values for a given loadflow case, this basic configuration of the test system is used inthree dynamic test cases as described in table II. The test casestarget the verification of the standard dynamic models amongthe simulation tools. These test cases are: voltage referencestep in no-load operation, stepwise load increase in islandoperation and a 3-phase short-circuit.Fig. 4: Test system for the comparison of standard dynamic modelsThe results of these tests are shown in figure 5 as an arrayof curves for the five different simulation tools. The resultsshow a good match although some smaller deviations duringtransients can be seen. In the steady-state condition the curvesoverlap nearly perfectly. The results of the tests were alreadypublished in [6]. In this reference all the parameters for thebranches and controllers are listed.TABLE II: Test cases for comparison of standard dynamic modelstest case 1test case 2test case 3opened S-GRID,opened S-GEN,opened S-Lstep of AVRvoltage reference UNGEN,setp 0.05 p.u.opened S-GRID,opened S-GEN,closed S-Lincrease ofdemand in LOAD PLOAD 0.05 p.u.closed S-GRID,closed S-GEN,opened S-L3-phaseshort circuitat NTHVThe applied parameters for the standard dynamic modeltests represent the basis for the parameter variation processin a further step of the DSM development process. In the nextstep the tested and compared standard dynamic models aresystematically assigned to generation units.C. Dynamic data expansion processThe dynamic data are added in each simulation tool byscripts as shown in algorithm 1 as pseudo code. Every generator i of n generators in the load flow model pass through thisprocedure. There are two mandatory conditions concerning themaximum active power Pmax,i of a generator i and its initialpower Pinit,i in the load case.The first condition assures that only large generation unitsabove Pmax,ref 250 MW are modeled as synchronous machines. Otherwise they are modeled as constant impedances.The second condition, if the first one is fulfilled, guaranteesthat only generators i with an initial active power greater thanPinit,ref 0 MW are equipped with the full appropriate controlmodels. Otherwise they only get an AVR model. In the case ofthe Dynamic Study Model the conditions particularly aim at

Fig. 5: Selected results from the comparison of standard dynamic models in three test casesAlgorithm 1 Dynamic data expansion process1: for generator i 1 n do2:if Pmax,i Pmax,ref then3:assign synchronous machine model to generator i4:if Pinit,i Pinit,ref then5:assign GOV, AVR and PSS models to generator i6:else7:assign AVR model to generator i8:end if9:else10:model generator i as negative constant impedance11:end if12: end forgenerators with negative initial power, to avoid deviations asa result of standard model choice for the governor for thermalpower plants. For instance it concerns pumped-storage hydropower plants or synchronous condensers. The machines wereidentified by maximum active power rather than rated apparentpower due to insufficient information in the load flow model.Executing this procedure, one obtains 1,013 synchronousmachine models out of 10,829 generators in the load flowmodel. Just 17 of them have only an AVR model. Theremained 9,816 generators are modeled as negative constantimpedances. The obtained Dynamic Study Model with initialparameters for the standard dynamic models was checked forcontrol limitation violations and stable time-domain simulationwithout any system event. In the next step, the parameters aretuned to influence the model behavior.D. Parameter variation processFor the next step the parameters from the tests of thestandard dynamic models were used as default [6]. The parameter variation process is merely done in one simulationtool. Therefore frequency measurements of the Wide AreaMeasurement System (WAMS) for a plant outage in Spainin October 2011 are used, see the black curves in figure 6.The outage was a loss of active power P 1 GW. Themeasured steady-state frequency deviation is f 26 mHz,before the frequency restoration reserves are activated. Thegradient of the frequency immediately after the outage variesup to the maximum of f / t 63 mHz/sec in WesternEurope and f / t 9 mHz/sec in Central Europe. Theelectromechanical wave propagation from Western to EasternEurope needs approximately two seconds. In consequence ofthe outage electromechanical oscillations occur in the EastWest axis of the power system with Central Europe as nonswinging area.The parameter variation process is primarily based on a

successive approximation of dynamic model behavior to thatmeasurements of the system event considering results of theeigenvalue analysis and mode knowledge of the power system[2], [8]. Generally the applied tuning method can be dividedinto the following steps: eady-state frequency deviation,moderate dynamics,faster dynamics anddamping.The variable tuning parameters of the standard dynamicmodels were already defined in table I. The variation of the parameters in the first steps is done simultaneously for all areas.Subsequently the variation is done individually for pre-definedareas. Figure 6 (a)-(c) shows the successive approximation ofthe model behavior for three frequency signals, in Western,Central and Eastern Europe. In black color one can see thefrequency measurement signals of these locations.In the first diagram (a) the comparison between the simulation results with initial parameters and the measurements isshown. It can be seen, that regarding the tuning method thedroop R of the governors has to be increased to get closer tothe measurements. The droop mainly affects the steady-statefrequency deviation. But increasing the droop influences thedynamic behavior of the governors and in consequence of thesystem. This can be seen in the simulation results of the seconddiagram. Therefore an adaptation of the moderate dynamics bychanging the turbine time constants TG2 and TG3 is needed. Infurther steps the faster dynamics and damping of the systemare tuned using the other variable parameters TA , K and KS1 .The frequency response in (c) shows the final tuning result,that satisfies the requirements. It is a good approximationfor the system inertia, frequency containment reserve andthe stimulation of electromechanical oscillations. The timeconstants of the oscillations correlate with the measured ones.The oscillation damping of the system cannot be modeledadequately as it is a non-constant value for the real system.The analysis of the electromechanical oscillations is alsoperformed in the linearized model by means of eigenvaluesof the system. As figure 7 shows, the characteristic modes aspresented in [8] are adequately modeled in the DSM.Fig. 7: Selected oscillatory eigenvalues for the DSMFig. 6: Measurements and simulation results for parameter variation processAfter tuning the system behavior in one simulation tool,the final parameters for all areas are distributed to users ofdifferent simulation tools. Afterwards a comparison of all toolsis performed. Figure 8 shows this comparison between themeasured and simulated frequencies for nodes in Western,Central and Eastern Europe, again as array of curves for thefive different simulation tools.It can be seen a good fit of simulated curves among thedifferent simulation tools. The smaller deviations may havethe same source as seen in the standard dynamic model tests.Hence the predetermined aims of the Dynamic Study Modelconcerning the dynamic behavior and the overall conformityof different tools are fulfilled.

impact (wind, HVDC, FACTS), e.g. [11]–[13].The major benefit of the Dynamic Study Model is the entiretopology information of the synchronously interconnectedpower system of Continental Europe under consideration ofdynamics, so that it can be used in analysis of global phenomena in stability studies. For analyzing local phenomenain the defined ”observability area” a detailed model can beimplemented modularly while keeping the global frequencybehavior. It is also possible to transfer the concept and thedevelopment process of the Dynamic Study Model to otherpower systems.V. ACKNOWLEDGMENTSThe authors gratefully acknowledge the constructive contributions, comments, discussions and technical support ofthe TSOs Amprion (Germany), ČEPS (Czech Republik), Energinet.dk (Denmark), Mavir (Hungary), PSE (Poland), REE(Spain), RTE (France), Swissgrid (Switzerland), TenneT (Germany), Terna (Italy) and TransnetBW (Germany).R EFERENCESFig. 8: Measured and simulated frequency after a plant outage in SpainIV. C ONCLUSION AND O UTLOOKAs shown in the paper, the novel approach of the DynamicStudy Model is suitable for the representation of the globaldynamic behavior of the synchronously interconnected powersystem of Continental Europe concerning the system inertia,the frequency containment reserve and the dominant inter-areaoscillation modes. It is available in different simulation toolsproducing similar results. The deviations to measurements andbetween the simulation tools are insignificant with regard tothe objectives of the model.In future the degree of the model accuracy can be augmented by considering existing standard dynamic models for: generation units with respect to their energy source(steam, gas, hydro etc.), e.g. [5], loads with dynamic impact, e.g. [9], [10], and infeeds connected via power electronics and transmissiontechnologies employing power electronics, with dynamic[1] ENTSO-E, “Ten-Year Network Development Plan 2014,” ENTSO-E,Brussels, 2014.[2] M. Larsson, P. Korba, W. Sattinger, P. Owen “Monitoring and Control ofPower System Oscillations using FACTS/HVDC and Wide-area PhasorMeasurements,” CIGRE Session, Paris, 2012.[3] elsexchange/standards/Pages/default.aspx, last access: 08/10/2014.[4] G. Herold, Electrical Power Systems III (German: Elektrische Energieversorgung III), Wilburgstetten: J. Schlembach Fachverlag, 2008.[5] IEEE, “Dynamic Models for Turbine-Governors in Power System Studies,” IEEE, 2013.[6] ENTSO-E, “Documentation on Controller Tests in Test Grid Configurations,” ENTSO-E SG SP&D Report, Brussels, 2013.[7] IEEE, “IEEE Recommended Practice for Excitation System Models forPower System Stability Studies,” IEEE Std 421.5-1992, 1992.[8] E. Grebe, J. Kabouris, S. Lopez Barba, W. Sattinger, W. Winter “LowFrequency Oscillations in the Interconnected System of ContinentalEurope,” IEEE PES General Meeting, Minneapolis, 2010.[9] IEEE, “Standard Load Model for Power Flow and Dynamic PerformanceSimulation,” IEEE Trans. Power Syst., Vol. 10, No. 3, Aug 1995.[10] CIGRE, “CIGRE Technical Brochure 566: Modelling and Aggregationof Loads in Flexible Power Networks,” CIGRE Working Group C4.605,Feb 2014.[11] IEC, “IEC 61400-27-1: Wind turbines - Part 27-1: Electrical simulationmodels Wind turbines,” IEC standard 61400-27-1 ed. 1 CDV, Nov 2013.[12] C. Hahn, A. Semerow, M. Luther, O. Ruhle “Generic Modeling of a LineCommutated HVDC System for Power System Stability Studies,” IEEEPower Engineering Society Transmission and Distribution Conference,Chicago, 2014.[13] S. Arabi, “A Versatile FACTS Device Model for Powerflow and StabilitySimulations,” IEEE Trans. Power Syst., Vol. 11, No. 4, Nov 1996.

the entire power system is expanded with standard dynamic models for generators and their appropriate control devices by means of allocation criteria. Afterwards the initial parameters Fig. 3: The development process of the Dynamic Study Model of the standard dynamic models are tuned on the basis of measurement values of a system event as well .