loth -13'h September, Porto, Portugal Initialization of Wind Turbine Models in Power System Dynamics Simulations J .G. Slootweg' H. Polinde? Member. IEEE I Electrical Power Systems, Electrical Power Processing Faculty of Information Technology and Systems, Delft University of Technology P.O. Box 503 1,2600 GA Delft, The Netherlands Phone: +31 15 278 6219, Fax: +31 15 278 1182, e-mail: j.g.slootweg@its.tudelft.nl Abstract: As a result of increasing environmental concern, increasing amounts of electricity are generated fromrenewable sources. Oneway of generating electricity fromrenewablesources is to usewind turbines. A tendency to erect morewind turbines can be observed. As a result of this, in the near future wind turbines may start to influence thebehavior of electrical power systems. To investigate the impact of increasing wind turbine penetration, power systemdynamics studies need to becarried out. To this end, power systemdynamics simulation software is used, in which wind turbine models must be integrated to enable the investigation of increasing wind turbine penetration on power systembehavior. If wind turbine models are to be integrated into power system dynamics simulation software, it must bepossible to calculate the initial conditions of thedynamic model fromload flow data to be able to initialize thedynamic simulation correctly. In this paper, a solution to this problemis presented. Keywords: wind turbine, modeling, simulation, power system dynamics, grid integration, model initialization, initial conditions 1. INTRODUCTION As a result of increasing environmental concern, the impact of conventional electricity generation on the environment is being minimized and efforts are made to generate electricity fromrenewable sources. The main advantages of electricity generation from renewable sources are the absence of harmful emissions and the infinite availability of the prime mover that is converted into electricity. One way of generating electricity fromrenewable sources is to use wind turbines that convert the energy contained in flowing air into electricity. Up to this moment, the amount of wind power integrated into large scale electrical power systems only covers a small part of the total power system load. The rest of the power system load is for the largest part covered by conventional thermal, nuclear and hydro power plants. Wind turbines often do not take part in voltage and frequency control and if a disturbance occurs, the wind turbines are disconnected and reconnected when normal operation has been resumed. Thus, notwithstanding the presence of wind turbines, frequency and voltage are maintained by controlling the large power plants as would have been the case without any wind turbines present. This is possible, as long as wind power penetration is still low. However, a tendency to increase the amount of electricity W.L. Kling' Member. IEEE generated fromwind turbines can be observed and they may begin to influence overall power systembehavior, making it difficult to operate a power systemby only controlling large scale power plants. It is therefore important to study the behavior of wind turbines in an electrical power systemand their interaction with other generation equipment and with loads. To investigate the impact of increasing wind turbine penetration on power system behavior, power system dynamics studies need to be carried out. To this end, power systemdynamics simulation software is used and to enable the investigation of increasing wind turbine penetration, wind turbine models must be integrated in power systemdynamics simulation software packages. One of the problems that must be solved to make this possible, is the calculation of the initial conditions of the wind turbine from load flow data in order to be able to initialize the dynamic model correctly. The structure of this paper is as follows. First, the problem of the initial condition calculation is commented upon. Then, models of the three most important actual wind turbine concepts are described and the system equations for each concept are given. To conclude, a solution to the problem of the calculation of the initial conditions for each of these concepts is discussed and applied to an example wind turbine. 11.THE MODEL INITIALIZATION PROBLEM A. Power system dynamics simulation software Power system dynamics simulation software is used to investigate the transient behavior and small signal stability of power systems. An example of a power system dynamics simulation package is PSSIE. Power system dynamics simulation software can be used when the phenomena of interest have a frequency of about 1 to 10 Hz. When the frequencies of interest are higher, subtransient simulation software, such as ATP, EMTP or the MATLAB Power Systems Blockset should be used. These packages contain more detailed and higher order equipment models than power systemdynamics simulation software. Some advanced software packages offer both transient and subtransient simulations and are able to automatically switch between equipment representations appropriate for the subtransient and transient domain, depending on the event to be simulated. 0-7803-7139-9/01/$10.00 02001 IEEE The main difference between dynamics simulation software and subtransient simulation software is that in the first only the fundamental harmonic component is taken into account. This approach enables the representation of the network by a constant impedance or admittance matrix, like in load flow calculations. Further, it reduces the number of differential equations, because no differential equations are associated with the network and fewer with generating equipment and because it enables the use of a larger simulation time step [l]. As a result of this approach, power system dynamics simulation software alternately executes a load flow, to calculate node voltages and branch flows, and a dynamic calculation, to calculate the response of the dynamic equipment models to changes in their terminal voltage, current andor frequency and to determine the new values of the state variables. This sequence is depicted in figure 1. Loadflow Dynamics Start calculatio? calculation Numsncal cDmp1expower values of state variables - - Figure I . Dynamic simulation activity sequence B. Model initialization As can be seen from figure 1, a simulation run using a power system dynamics simulation package starts with preparing a load flow case and solving this using the load flow calculation module of the software. Thereafter, the dynamic simulation model must be initialized. This comes down to calculating the initial conditions of the dynamic equipment models from the load flow data. If this is not done correctly, state variables and other quantities do not stay at the value at which they are initialized, but start changing at the start of the dynamic simulation. In this situation, it may take much time to reach a new equilibrium and numerical instability can occur before an equilibrium is reached. Further, if an equilibrium results, it can be different from the original load flow case that was initially planned to be investigated [2]. These outcomes are undesirable and therefore attention should be paid to the calculation of the initial conditions of dynamic equipment models. Some of these models have been used for long, such as synchronous generator models and exciter and governor models. The calculation of the initial conditions of these model categories does not pose special problems and is integrated in power system dynamics simulation software. However, in order to be able to integrate new equipment models into dynamic simulation software, not only an equipment model must be available, but the problem of the calculation of the model's initial conditions must be solved as well. For models of the three currently most widely used wind turbine concepts, a solution to this problem is presented in this paper. 111. MODELS OF ACTUAL WIND TURBINE CONCEPTS A. Introduction The three most important currently applied wind turbine concepts are the following: A constant speed wind turbine, which consists of a directly grid coupled squirrel cage induction generator. The wind turbine rotor is coupled to the generator through a gearbox. The power extracted from the wind is limited using the stall effect. This concept will be further referred to as concept 1. A variable speed wind turbine with doubly fed (wound rotor) induction generator. The rotor winding is fed using a back-to-back voltage source converter. Like in concept 1, the wind turbine rotor is coupled to the generator through a gearbox. In high wind speeds, the power extracted from the wind can be limited by pitching the rotor blades. This concept will be further referred to as concept 2. A variable speed wind turbine with a direct drive synchronous generator. The synchronous generator may have a wound rotor or may be excited using permanent magnets. It is grid coupled through a back- to-back voltage source converter or a diode rectifier and voltage source converter. The synchronous generator is a low speed multi pole generator, therefore no gearbox is needed. Like in concept 2, the power extracted from the wind is limited by pitching the rotor blades in high wind speeds. This concept will be further referred to as concept 3. For a more elaborate description of these wind turbine concepts and their advantages and disadvantages, ref: [3] should be consulted. Further, three remarks must be made. First, when calculating the initial conditions of a dynamic model from load flow data, the model is calculated through from the output to the input. The initial values of the output are determined in the load flow case and the initial values of the state variables and other values have to be calculated from these load flow data. Therefore, the description and modeling of the components of the each of the wind turbine concepts will be discussed starting from the grid connection point. Second, when calculating the initial conditions, the system is assumed to be in steady state. Therefore, all time derivatives are equal to zero and will not be included in the wind turbine model equations. Further, the following is assumed when deriving the equations describing each of the systems: Magnetic saturation is neglected in the generator models Flux distribution is sinusoidal Converter losses are neglected Third, the rotor is modeled as an algebraic relationship between wind speed and rotor. power. This is considered detailed enough for power system dynamics studies [4]. In this way, the use of the blade element method which would result in lengthy computations and the need for detailed rotor blade descriptions can be avoided. It is difficult to calculate the per unit value of the power extracted from the wind, because mechanical wind turbine characteristics such as rotor diameter and nominal rotor speed come into play. Therefore all equations are in per unit, except the equations describing the rotor. These equations are only valid for a 2 MW wind turbine with the characteristics given in table 1 . For wind turbines with different nominal power and/or mechanical characteristics, new numerical approximations for the power coefficient cp should be developed. Wind turbine characteristic Value for concept 1 Rotor diameter [m] 75 Value for concepts 2 and 3 1 5 Nominal rotor speed [RF'M] Minimumrotor speed [WM] 16.7 18 9 n.a. B. Steady state equations describing concept I Seen from the grid, the first component is the squirrel cage induction generator, which is described by the following per unit equations in the d-q reference frame using the generator convention, i.e. positive currents are outputs [ 11. Voltage equations: v*= - R,i,+ oS((Lsa+ Lm)iqs+ Lmiqr) vqs= - Rsiqs- q( ( Lsa+ Lm)idp+ Lmib) O= - Rriqr- sws((L,+ Lm)idr+ Lmih) O = - Rrib+ sos((Lnr+ Lm)iqr+ Lmiqs) (1) Real and reactive power: P= vdrirtr+ v i Q= vqsidF- vJqS (2) qs 4s In these equations v is voltage, i is current, R is resistance, L is inductance, omega is frequency and s is slip. The indices d and q stand for direct and quadrature component respectively, m stands for mutual, s for stator, r for rotor and U for leakage. P is active power and Q is reactive power. The rotor slip equals Nominal wind speed [ ds] (3) I 12 14 in which p is the number of poles and om is the mechanical frequency of the generator rotor. The generator is coupled to the rotor with a shaft. Wind turbine shafts are relatively flexible, therefore it is important to incorporate the shaft in the model in power system dynamics studies [5]. In steady state, the following per unit equation describes the wind turbine shaft: (4) in which T is torque in P.u., y is the angle between the both ends of the shaft [el. rad] and K, is the shaft stiffness. The rotor is modeled using the following algebraic relation between wind speed and mechanical power extracted: P, =p 2 Cp(a)AR v; cP(A)= (46.4((-- A 0.01)- 2.0))e - 16.52- ( 5) with 1 (6) in which P, is the power extracted from the airflow [W], p the air density [kg/m3], cp the performance coefficient or power coefficient, 3L the tip speed ratio vt/vw,, the ratio between blade tip speed v, and wind speed upstream the rotor v, [ ds] and A, the area covered by the rotor [m2]. C. Steady state equations describing concept 2 In this concept, the first components seen from the grid are the doubly fed induction generator and the converter feeding the rotor winding. The voltage equations describing the doubly fed induction generator are the following [6]: v&= - R,i,+ ms((Lsa+ Lm)iqs+ Lmiql) vqs= - R,iqs- as((Lsa+ Lm)idr+ Lmi*) vqr= - Rriqr- sos((Lra+ Lm)idr+ Lmidr) v*= - Rri,+ sos((Lra+ Lm)iqr+ Lmiqs) (7) A doubly fed induction generator is capable of generating reactive power. Reactive power can be delivered by the stator or by the grid side of the converter. It is assumed that the grid side of the converter operates at unity power factor and that any reactive power is delivered by the stator. In that case, the following equations apply for real and reactive power: P= Ps+ Pr= vdFi,+ v i +vdFidc+ vqsiq, Q= vqSi*- vdriqs (8) in which the index c stands for converter. The other symbols have the same meaning as in equations ( 1) and (2). Note that from (S), it follows that the converter only takes part into active power generation, not in reactive power generation. In the case of a variable speed wind turbine, the shaft can be neglected, because the power electronic converter decouples the electrical and mechanical behavior. Therefore, the shaft properties are hardly reflected in the wind turbines response to wind speed changes or grid faults. Therefore, no equations describing the wind turbine shaft will be taken into account. The rotor is modeled using the following equations: qs qs P,= p 2 cp(a,e)ARv; (9) with - a 0.8- a I L ' g0. 6- $ 0 4 - I 2 - 12.5 - (10) 116 cp(;l,e)= 0.22(--- 0.48- 5) e *i AI and 1 - 1 0.035 (11) The meaning of the symbols is the same as in equations (4) and (5). However, the blade pitch angle 8 [deg] has been added to the equations, because in the variable speed wind turbines the power extracted from the wind is limited using pitch control. Using these equations, the system has not been fully determined yet. This is caused by the fact that concept 2 is a variable speed wind turbine. Therefore, the blade tip speed v, varies in concepts 2 and 3, in opposition to concept 1, where it is approximately constant. Thus, A is not determined. To link A to equations (7) to (1 l), the speed versus power control characteristic can be used. The control characteristic of the 2 MW example wind turbine used here is depicted in figure 3 and will be further referred to as P(w,,,). -- -- A~ A+ 0.088 e+ 1 I I 8 10 12 14 16 18 20 Rotor speed [RPM] Figure 3. Rotor speed versus power control Characteristic used in concepts 2 and 3. the semiconductor switches. The equations describing a synchronous generator can be found in [I ]. The damper windings need not to be taken into account, because the current through the damper windings is zero in steady state. The direct drive synchronous generator is coupled to the generator side of the converter. The generator terminal voltage vg can be derived from the control strategy V,(o,) and the rotor speed. Voltage equations: vg= /= (13) vdF =om(Lqm +LsJiqs- R,i* v =om(- (L*+ Lso)idr+ LhGd)- Rsiqs 5- Real and reactive power: (14) In these equations, the index fd stands for field and g for generator terminal. The main inductance in the d and the q axis may be different in opposition to what applies to the induction generators used in concepts 1 and 2. The other indices have the same meaning as in equations (1) and (2). The generated reactive power Q, can be controlled through the rotor winding excitation and depends on the reactive power generation control characteristic. Normally, it is dependent the rotor speed om andor the terminal voltage v,. The reactive power control characteristic will be further referred to as Q,(w,,v,). The last equation of (13) only applies to synchronous generators with a wound rotor. Further, when a rotor with permanent magnets is used, the product of mutual inductance in the d-axis L,, and the field current i, has to be replaced by the flux of the permanent magnets coupled to the stator winding in the equation of (1 3) giving vqs. This results in the following voltage equations when a rotor with permanent magnet excitation is used: D. Steaa'y state equations describing concept 3 In this concept, the first component seen from the grid is a voltage source converter. A voltage source converter can generate both active and reactive power. The equations describing the converter in steady state operation are Pc=vdci dc+v i (12) qc qc Qc= 'qcidc- 'dciqc Then, the generator side of the converter is modeled. It can be modeled as a voltage source. The value of the voltage is determined by the control strategy. Normally, it varies with the generator speed and its value can be derived when the generator speed is known. The relation between terminal voltage and generator speed will be further referred to as When the generator side of the converter consists of a diode rectifier, its voltage is controlled indirectly through the grid side of the converter andor the generator if it has a wound rotor. When the generator side of the converter consists of a voltage source converter as well, like the grid side of the converter, its voltage can be controlled directly by controlling V,(O. v =/ v*+ vis g v&= om(Lqm+ Lso)iqs- RsiA (15) vqs =om(- (Lm, +L,,>i* +qpm> - Qqs in which $pm is the flux of the permanent magnets linked by the stator winding. The wind turbine shaft is neglected again, because the :<haft properties are hardly reflected in the grid interaction. The rotor model and speed versus power control characteristic are the same as those of concept 2. IV. CALCULATION OF INITIAL CONDITIONS A. Number of equations and unknowns f or each concept In table 2, the variables with a known value, the variables with an unknown value and the number of equations for each of the wind turbine concepts are given. Concept 3 is addressed both for a wound rotor and a permanent magnet synchronous generator. From this table, it can be concluded that the initial conditions of all concepts can be calculated. The substitution of the equations in each other is basic calculus and will therefore not be carried out here. Note that because P and Q are quadratic quantities, the equations are not linear. A symbolic mathematics package such as MAPLE can be used advantageously to solve the equations. Equations in which the rotor speed versus power characteristic P(o,,,) occurs, had to be solved numerically, because it is difficult to obtain a closed form solution. and equations f or each of ihe Concept 1 KnOWn P, vds, Vqs variables Unknown iqs, iqr, Q>s, Y Number of (1)+(2)+(4) equations =7 concepts Concept 2 Concept 3 Concept 3 with with wound permanent rotor magnets P, Q, vds, Pc, Qo v d o Pc, Qc , vds? vq\ vqo p, Vqa ldsr l qsr l dr r Vdr. vqs. Q@ ."&. vqs3 p,, $pn . . . . . . iqr, Vdri Vqrr I do l qo Ids, I d 0 I q o I ds s, U, Iqs. wm, vfdi l qsr om, vg &, Q, vg (3)+(7)+(8) (12)+(13) (12)+(14) +P(%) +(I41 +(I51 =8 +P(O,) +P(%) +V,(%) +V,(%J=9 +Q,(%VJ =1 I B. Example generator characteristics In the next paragraphs, for each concept the initial conditions will be calculated. The parameters of the generators used in concept 1 and 2 can be found in [7] and of the generator used in concept 3 in [8] respectively and are not reproduced here. C. Initial conditions of concept I The initial conditions of a concept 1 model can be calculated using algorithm depicted in figure 4. As an example, this algorithm is applied to a wind turbine of 2 MW nominal power that is generating 1 MW of real power, i.e. 0.5 P.u., with a terminal voltage of vd,=l.02 p.u. and v,,=0.25 p.u. The results of each of the steps are given in table 3. D. Initial conditions of concept 2 The initial conditions of a concept 2 model can be calculated using the algorithm depicted in figure 5. In this algorithm the generator losses are neglected by setting the mechanical power extracted from the wind equal to the electrical power generated in the load flow case. This very much simplifies the calculations and introduces only a minor error in the initial conditions, because the generator losses are small. As an example, the algorithm is applied to a wind turbine of 2 MW nominal power that is generating 1 MW of real power, i.e. 0.5 p.u. and 0.4 MVAr reactive power, i.e. 0.2 p.u. at a terminal voltage of vd,=l.02 p.u. and vq,=0.25 p.u. The results of each of the steps are given in table 3. I 1. Use (1) and (2) to calculate ids, io, ido io,, s and Q I 1 1 2. Calculate copper losses using p,=~~i,:+i,:,+~,(i,,'+i,,Z) I 3. Calculate rotor speed using (3) I 1 I 4. Calculate torque using P=Tw I 1 I 5. Calculate y using (4) I 1 I 6. Calculate wind speed using (5) and (6) I Figure 4. Algorithm f or calculation of initial conditions of a concept 1 wind turbine from loadflow data. PCP", P=P", 2. Calculate wind speed using (IO) and ( I 1) 1 4. Calculate v, vqp idp i, i& iqs using (7) and (6) Figure 5. Algorithm f or calculation of initial conditions of a concept 2 wind turbine from loadflow daia. E. Initial conditions of concept 3 The initial conditions of a concept 3 model can be calculated using the algorithm depicted in figure 6. In this algorithm, the generator losses are again neglected, like in the case of concept 2 an$ for the same reasons. As an example, the algorithm is applied to a wind turbine of 2 MW nominal power that is generating 1 MW of real power, i.e. 0.5 p.u. and 0.4 MVAr reactive power, i.e. 0.2 p.u. .at a terminal voltage of v,,=1.02 p.u. and v,,=0.25 p.u. Further, it is assumed that and that in case of the permanent magnet rotor bpm equals 1.3 p.u. The lower equation of (16) only applies to a generator with a wound rotor. Other control strategies for terminal voltage and reactive power can be used without rendering the proposed algorithm invalid. The results of each of the steps are given in table 3. F. Initial conditions of concepts 2 and 3 at nominal power When concepts 2 and 3 generate nominal power in the load .flow case, there is no unique relation between generated power and rotor speed, as can be seen from the right part of figure 3 (up from 18 rpm). This is caused by the fact that at wind speeds above nominal the pitch angle controller becomes active and limits the power extracted from the airflow to the nominal power, no matter what the actual wind speed is. Therefore, when a concept 2 or concept 3 wind turbine generates nominal power in the load flow case, the actual rotor speed and the wind speed or the pitch angle have to be given. The other values can then be calculated in the same way as when the wind turbine is initialized below nominal power. In figures 5 and 6, the right arrow corresponds to an alternative algorithm that can be used in that situation. V. CONCLUSIONS In this paper, a method was presented to calculate the initial conditions of dynamic models of the most important actual wind turbine concepts from data provided in a load flow case. This is an important result that facilitates the integration of wind turbine models in power systems dynamics simulation software. REFERENCES [l ] P. Kundur, Power system stability and control, New York, US: McGraw-Hill, Inc., 1994. [2] P. Lacassel, L.Gerin-Lajoie, D. McNabb, P.-J . LagacC, J . Mahserediian, "New Initialization method for controls in transient analysis", 200 1 IEEE Power Engineering Society Winter Meeting, J an. 28-Feb. 1, 2001, paper 66-09. [3] S. Heier, Grid integration of Wind Energy Conversion Systems, Chicester, UK: J ohn Wiley & Sons Ltd., 1998. [4] V. Akhmatov, "Some aspects of new windmill technologies on voltage stability", Seminar on Control Concepts of Wind Turbines, 23 Oct., 2000. [5] V. Akhmatov, H. Knudsen, A.H. Nielsen, "Advanced simulation of windmills in the electric power supply", International J ournal of Electrical Power & Energy Systems, v01.22, no.6, 2000, pp.421-434. [6] P.C. Krause, 0. Wasynczuk, M.S. Hildebrandt, "Reference frame analysis of a slip energy recovery system", IEEE Transactions on energy conversion, vo1.3, no.2, J une 1988, pp.404-408. [7] J .G. Slootweg, H. Polinder, W.L. Kling, "Dynamic Modelling of a Wind Turbine with Doubly Fed Induction Gcnerator", 2001 IEEE Power Engineering Society Summer Meeting, J uly 15-19, 2001. [8] J .G. Slootweg, H. Polinder, W.L. Kling, "Dynamic Modelling of a Wind Turbine with Direct Drive Synchronous Generator and Back to back Voltage Source Converter and its Controls", 2001 European Wind Energy Conference and Exhibition, Copenhagen, Denmark, J uly 2-6,2001, P=P,, 2. Calculate wind speed using (10) and (1 1) 4. Calculate v and for wound rotor also Q, using&,(w,) and Q,(w,,v,) J. 5 Calculate vd., vq,, id., ig,. v,, and i,, for wound rotor using (13) usina (11) and 115) and (14) or Q,, v,,, vq, ids, i for permanent magnet rotor Figure 6. Algorithm for calculation of initial conditions of a concept 3 wind turbine,from loadjlow data. Table 3. Results yielded when the solution algorithms are applied to an ?dotherwise. Concept 3 with wound rotor ample. Va h Step f unless indic Concept 2 are in per un Concept 1 Concept 3 with permanent magnets 0,=15.2 WM BIOGRAPHIES J.G. Slootweg received his MSc degree in electncal engineenng from Delft University of Technology on September 23rd, 1998. During his education he stayed in Berlin for six months, to hear lectures at TU Berlin and to conduct research at the Dynamowerk of Siemens AG He is currently working on a PhD on large scale integration of dispersed generation into existing electric grids at the Electncal Power Systems Laboratory of Delft UT H. Polinder received his MSc degree in electncal engineering in 1992 and his PhD degree in 1998 both fromDelft University of Technology. Currently, he is an assistant professor at the Electrical Power Processing Laboratory at the same university, where he gives courses on electncal machines and dnves. His main research interest is the field of generator systems in renewable energy, such as wind energy and wave energy W.L. Kling received his MSc degree in electncal engineenng fromthe Technical University of Eindhoven in 1978. Currently he is a part time professor at the Electnc Power Systems Laboratory of Delft UT. His expenence lies in the area of planning and operation of power systems He is involved in scientific organizations such as Cigre and IEEE He is the Dutch representative in the CigrC Study Committee 37 Planning and Development of Power Systems. 1 i,,=O. 3 7 iq,=0.49 idr=-0.47 iqr=-O. 16 Q=-0.40 ~=-0.0049 w,=15.2 RPM a,= 15.2 RPM 2 P,=0.0062 v,=9.5 m/s v,=9.5 mJ s w,=1.0049 s=-0.015 idp0.51 t i,g-O.072 v,=9.5 mls id,=0.5 1 iqc=-O.072 3 4 T=0.50 i,= 0.51 i,,=-0.072 v,,=-O.O 10 v,,=-0.009 id,=-0.61 i,;0.42 v,=0.72 Q,=O v,=0.72 vd,=0.27 v,,=0.67 i,=0.25 i,,=0.64 v,=1.32 i,=1.09 vd,=0.32 vq,=0.65 idq= 0.26 iq,=0.64 Q,=-0.03 5 5 y=l.68 el. rad 6 vW=9.2m/s I I