Understanding Dynamic Model Validation Of A Wind Turbine Generator And .

appropriately. Three options for reactive compensation can bechosen for operation: constant power factor, constant reactivepower, and constant voltage s of WTGs: a WPP may consist of two groupsof WTGs—one of Type 1 WTGs and another ofType 3 WTGs—and each group is about the samesize in total power rating. Line impedance: the electrical distance betweenone turbine and the main substation at the point ofinterconnection (POI) differs among the turbineswithin a WPP. As such, even for the same windcondition, the voltage and phase angle at theterminals of each WTG may be slightly differentfrom one turbine to another. Control setting: the control setting of one group ofturbines may be different from another group ofturbines—for example, to compensate for thevoltage drop within the collector system. Onegroup of turbines may be controlled to generate atunity power factor while another group iscontrolled to regulate the voltage at the POI.The diversities listed above have different impacts on theWPP’s response to different types of disturbances. The relayprotection settings at each turbine are normally customizedaccording to the recommended values from the manufacturerbased on the regional or local grid codes and/or the requestfrom the system operators or project developers. As a result,during a fault, each turbine will experience different voltageand current levels, and some of the turbines—usually thoseclosest to the POI—will get disconnected from the grid whileothers stay connected. Thus, upon a disturbance, it can beexpected that a group of turbines within a WPP may bedisconnected from grid while others stay online. In a way, thismakes a WPP more resilient or more forgiving to disturbanceevents. For example, [4] summarized an observation in a WPPin Texas for a period of one year and concluded that in amajority of faults only 14% of the events disconnect the entireWPP. And for approximately 80% of the events, only 15% ofthe turbines were disconnected from the grid. Thus, during thevalidation, we need to understand this, and we can expect thatthe pre-fault generation may be different from the post-faultgeneration. This fact needs to be reconciled during thevalidation process. Another option is to screen the data andvalidate the dynamic model using only the available data thathas the same output power before and after the faults, anindication that no turbines were disconnected from the grid. CollectorBusG-boxGridPFCCapacitors(a) Type 1 formerVariableResistorGrid(b) Type 2 r Converter(c) Type 3 WTGCollectorBusPWM ConverterPMSGTurbineTransformerGrid(d) Type 4 WTGFig. 1. Different types of WTGsInitialization of the dynamic simulation takes place in thepower flow stage of the simulation; thus, the initial values ofthe power generation (both real and reactive power) are set atthe actual generation [1].B. Wind Power Plant RepresentationThe WPP must be represented according to the actual WPP.The most common method is to represent hundreds of windturbines as a single turbine. The method of equivalencingmany turbines into a single turbine has been documented [1]–[-4], and it will not be repeated here.Bus 1ReplayRecordedV(t), f(t)At this busC. Diversity RepresentationA WPP covers a very large area; thus, there is diversitywithin a WPP. Diversity in a WPP can be in different forms: Wind resource: a group of wind turbines in onecorner will experience different wind speeds due tothe spatial difference or due to the landscape andthe turbine locations.230 kV LineR1, X1, B1230/34.4 kVsubstationtransformerBus 2R t, X tBus 3230/34.5 kVcollectorsystemRe, Xe, Be34.4/0,6 kV GSUtransformerRte, XteBus 4Bus 5100 MWequivalentwind turbinegeneratorWTGSubstation levelshuntcompensationTurbine levelshuntcompensationFig. 2. Single-turbine representationD. Multiple-Turbine RepresentationAs mentioned before, the WPP must be represented accordingto the actual WPP. Consideration should be given to the typeof study being conducted. For planning studies, the worst-casescenarios are often considered; thus, a single-turbine2This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

representation of a WPP, as shown in Fig. 2, is commonlyused. Other types of studies may need to use a multipleturbine representation. For example, if the interaction amonggroups of turbines is the main interest of study, a multipleturbine representation should be considered. For example, take aWPP which consists of 60% of Type 1 WTGs, and the rest areType 4 WTGs operated at unity power factor (refer to Fig. 3).Bus 4aBus 3faults). Thus, in the validations, it is preferable to use recordeddata corresponding to the symmetrical faults. Also, in powersystem planning, dynamic model validation is designed tostudy the worst-case scenario; thus, often a single-turbinerepresentation is used.B. Case 1: Wind Speed ConstantThe WPP is represented as a single turbine. The powersystem network is modeled up to the POI, and the controlparameters are set to represent the actual settings. An exampleof the WPP validation is shown in Fig. 4.Bus 5aWTGBus 1ReplayRecordedV(t), f(t)At this bus230 kV LineR1, X1, B1230/34.4 kVsubstationtransformerBus 2Rt, Xt34.5 kVcollectorsystemRe, Xe, Be34.4/0,6 kV GSUtransformerRte, XteBus 4bBus 5b40 MWequivalentType 4 WTGWTGSubstation levelshuntcompensationTurbine levelshuntcompensation60 MWequivalentType 1 WTGFig. 3. Multiple-turbine representationIn this case, we need to represent the WPP using a two-turbinerepresentation so that the significantly unique characteristicsof each turbine type are included. However, , if the same WPPcontains several Type 3 WTGs representing less than 2% ofthe total power of the WPP, these turbines do not need specialrepresentation because the impact of the Type 3 WTGs on theoverall behavior of the WPP will be negligible. The smallnumber of Type 3 WTGs can be lumped into therepresentation of the Type 4 WTGs because their behavior isthe closest to that of the Type 3 WTGs. Thus, therepresentation of the WPP must be unique and include asignificant proportion of the power of the total size of theWPP. Examples of multiple-turbine representations in WPPmodeling have been documented in several sources [5].(a) Real powerIII. DYNAMIC SIMULATIONS TO VALIDATE DYNAMIC MODELS(b) Reactive powerVery often the state estimations of a power system that arecaptured during the short duration of a transient fault for theduration of a disturbance are not available to re-create theevent for the entire system; thus, conducting a validation foran entire interconnection is not feasible or necessary, and thedynamic model validation is usually conducted for one plant ata time. The validation is normally done by using the datacaptured at the POI of the WPP. The voltage at the POI is thenreplayed to drive the simulated WPP, and the response iscompared to the recorded data during the event.Fig. 4. Real and reactive power comparison of the WPP validated for Case 1Given the same voltage and the frequency at the POI, thereal and reactive power from the simulation match the real andreactive power data recorded at the POI. This is a goodvalidation example wherein the operating condition is normal,and it follows the assumptions made in the dynamic modelrepresenting the WTG. In this case, it is important that thesimulation is initialized to the same operating condition at thePOI where the data is recorded. Note that in Case 1, the windspeed during the disturbance was constant, as is theassumption made in the dynamic model representation.A. Availability of DataThe data to validate the WPP dynamic model are not easyto get. The recent proliferation of synchrophasor units, alsoknown as phasor measurement units (PMUs) in many parts ofthe power system network, makes it easier to harvest data thatcan be used to validate the WPP dynamic model [6]. When weobtain the data, the next step is to find the disturbance eventswithin them. Depending on the severity of the disturbance,these events are good candidates to validate the dynamicmodels.Power system planning is commonly conducted usingpositive-sequence-based power system software such as PSSE,PSLF, and PowerWorld. These programs are intended to solvepositive-sequence cases (such as three-phase-to-groundC. Case 2: Wind Speed Varies During the Window ofObservationThe WPP is represented as a single turbine. Case 2 isdifferent from Case 1 in that the wind speeds vary during therecorded observation (refer to Fig. 5). The dynamic modelused here does not allow modeling at varying wind speeds;thus, as shown, there is a mismatch of real power between therecorded data and the simulated output. Note that thevariation of the wind speed is not large enough to affect thereactive power control. As shown here, the reactive poweroutput of the simulation matches the recorded data very well.3This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

E. Case 4: Partial Drop-Off of the WTGsAs discussed in the previous section, the diversity within aWPP makes the operating condition at individual turbinesunique. Case 4 is used to illustrate the impact of diversitywithin a WPP. In the recorded data, it was observed that thepre-fault data of the real power output of the WPP is higherthan the real power output during the post-fault condition.This is an indication that some turbines disconnected from theWPP during the transient fault. Note that in both the pre-faultand post-fault conditions, the wind speed is steady. This isevidence that the real and reactive power do not fluctuate.Thus, it is appropriate to model the WPP with two groups ofWTGs: one representing the WTGs that stay connected to thegrid (91% of the total) and another group representing theWTGs that disconnected during the fault (9% of the total).This 9% of the WTGs perhaps represents the WTGs closest tothe POI where the impacts of the transmission faults are worsethan they are in the rest of the WPP due to its diversity.Representing the WPP with a single turbine will not reflect theactual situation recorded during the fault event. Fig. 7 shows asingle-line diagram of this WPP to represent the circuitconfiguration as the sequence of events unfolded.(a) Real power(b) Reactive powerFig. 5. Real power mismatched and reactive power matched in the WPPvalidation for Case 2D. Case 3: Mismatch on Both the Real and Reactive PowerIn Case 3, the wind speed is shown to vary within a largepower range (refer to Fig. 6). In Case 2, the small variation ofreal power does not significantly impact the match to thereactive power. However, as shown in Case 3, the variation ofthe real power output of the WPP is very large, and as suchthis type of recorded data is not suitable for the validation ofthe generic dynamic model that we used (the wind speed isassumed to be constant).POI or Connectionto d-mountedTransformerEquivalentTwo TurbineRepresentationW91% WTGs stays“on” after the fault.W9% WTGs weredropped of lineduring the fault.Fig. 7. Multiple-turbine representation and the recorded voltage and frequencyreplayed at the POI for Case 4At the beginning of the simulation, both of the generatorsare connected. When the fault occurred, the voltage at theterminal of the 9% of the WTGs drops below the undervoltagerelay setting that triggered the disconnection of this generatorfrom the WPP while the rest of the generators (91%) stayconnected. Fig. 8 shows the recorded voltage and frequency(at the point of interconnection) used to drive the simulation.(a) Real power(b) Reactive powerFig. 8. Recorded voltage and frequency at the POIFig. 6. Real and reactive power mismatched in the WPP validation infor Case 34This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

As shown in Fig. 9 (a) and Fig. 9 (b), the real and reactivepower output of the WPP match the simulation results. Notethat in addition to a single-turbine representation, thesimulation result includes a simulation of all the WTGs withinthe WPP. All show a good match between the simulation andthe measurement, especially during the post-fault recovery,which is the most important part of the simulation.reactive power control), and the post fault should return theWPP to the pre-fault generation (none of the WTGs isdisconnected), or all of the WTGs are disconnected. Thedynamic model validation requires several data sets measuredat the POI of the WPP, representing commonly occurringevents in the area. During the process, the parameters of thedynamic model are tuned so that the simulated results matchthe actual measured data. With more data available, theaccuracy of the dynamic model in representing the actualWTG or WPP will be improved.Without understanding the process of dynamic modelvalidation, the correct data used in validation, and theassumption made in the dynamic model, the validated modelmay not be representative of the actual WTG and WPP.V. ACKNOWLEDGMENTThis work was supported by the U.S. Department ofEnergy under Contract No. DE-AC36-08-GO28308 with theNational Renewable Energy LaboratoryVI. REFERENCES(a) Reactive power[1][2][3][4][5](b) Reactive power[6]Fig. 9. Comparison between recorded data and simulation data.Western Electricity Coordinating Council Wind Generator ModelingGroup, WECC Wind Power Plant Power Flow ModelingGuide, nt%20Power%20Flow%20Modeling%20Guide.pdf, May 2008.J. Brochu, C. Larose, and R. Gagnon, “Generic equivalent collectorsystem parameters for large wind power plants,” IEEE Trans. EnergyConvers., vol. 26, no. 2, June 2011.E. Muljadi, C.P. Butterfield, A. Ellis, J. Mechenbier, J. Hochheimer, R.Young, N. Miller, R. Delmerico, R. Zavadil, and J. C. Smith,“Equivalencing the collector system of a large wind power plant,”presented at the 2006 IEEE Power Engineering Society 2006, http://www.nrel.gov/docs/fy06osti/38940.pdf.E. Muljadi, Z. Mills, R. Foster, J. Conto, and A. Ellis, “Fault analysis ata wind power plant for one year observation,” in Proc. 2008 IEEEPower Energy Soc. Gen. Meeting (PES GM), pp. 1–7.E. Muljadi, S. Pasupulati, A. Ellis, and D. Kosterov, “Method ofequivalencing for a large wind power plant with multiple turbinerepresentation,” presented at the 2008 IEEE Power Engineering ��24,2008, http://www.nrel.gov/docs/fy08osti/42886.pdfY. Zhang, E. Muljadi, D. Kosterev, and M. Singh, “Wind power plantmodel validation using synchrophasor measurements at the point ofinterconnection,” IEEE Trans. Sustain. Energy, vol. 6, no. 3, July 2015.IV. CONCLUSIONDynamic model validations need to be done periodically toensure that the dynamic models sent to the regional reliabilityorganizations represent the latest setup of the WPP controlparameters. In validating the dynamic model, we need torecreate the actual representation of the network connection,the sequence of events, and the correct representation of theWPP (initialization, control settings, protection settings). Thedata needed for the validation must be selected to represent theassumptions adopted for the dynamic model of the WTG.For a single-turbine representation of a WPP, the idealmeasurement data that should be used to validate a WPPdynamic model is from the event with a steady wind speed,the fault event is a symmetrical fault event, the reactive powercontroller should match the actual setting (voltage control or5This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

dynamic model to ensure that different WTGs from different manufacturers can be represented as accurately as possible to the actual turbines. Section II presents a discussion on WPP representation. Section III presents the dynamic model validation, followed by Section IV, which presents the dynamic simulations to validate the dynamic models.