SYNC NOUS MA TERS IDENTIFICATION USING TEST DATA L Jose Anthnio Jardim, E E E Edson da Costa Bortoni, E E E DEE-FEG-UNESP bortoni@feg.unesp.br PEA-POLI-USP jardini@pea.usp.br BRAZIL Abstract - Titis work shows a computational m e ~ o d o i ofor ~ the determination of synchronous machines parameters using load rejection test data. The quadrature axis parameters are obtained with a rejection under an arbitrary reference, reducing the present difficulties. I. INTRODUCTION In the last decades, many methods have been developed in order to obtain synchronous machines parameters. Two of then gained prominence due to their practice, low risk level imposed to the machine under test and moreover, the quality of the obtained data. One is the frequency response method [11 and the other is the load rejection test [ 2 ] . The frequency response test is carried out by applying currents about 0.5% of the rated current., with frequencies varying in the range from 0.001 Hz to 1000 Hz. The rotor must be properly positioned in order to obtain the direct and quadrature axis parameters. Using the frequency response data one can evaluate the operational reactances and consequently determine the parameters and time constants currently used in power systems studies. The second method also allows the determination of direct and quadrature axis parameters by executing load rejection in two special operational points, where the components of current do exists in the axis of interest only. The direct axis load point can be obtained under exciting the machine as it has been synchronized to the system. The machine must be running at negligible active power and driving a considerable amount of reactive power from the system to obtain non-saturated parameters and avoiding undesirable overvoltages during the tests. The process to locate the quadrature axis is not so hivial, since one must find a loading point in which the absolute value of the power factor angle (9)be equal to the power angle (6). In practice it can be found after successive load rejections with different reactive powers, aiming to ” i z e the field current variations. An alternative procedure is the employment of a power angle meter [3]. These “ H i e s can, sometimes, make this test impracticable. This work will show that the data obtained with a load rejection under an arbitrary reference is sufiicient to detennine the transient parameters using numerical methods. The identification will be made using the LevenbergMArquardt Algorithm [5]. 0-7803-39464/97/$10.00 0 1997 IEEE. IL TRANSIENT PROCESS MODELING The synchronous machine transient process modeling will be made in three steps. The first takes into account only the steady state, before the load rejection; the second analyse the transient process after the rejection and in the former, a composite of the previous results will be done in order to obtain the complete behavior of the machine under a load rejection [4]. In these analysis all the variables are in per unit. A. Steady State In a three-phase system with perfectly symmetric voltages, balanced loads, with no damping winding currents, neglecting the armature resistance and taking the angular speed (0)equal to 1 P.u., one can obtain the initial conditions of the linkage fluxes for the direct and quadrature axis: where U is the terminal effective voltage and 6 is the angle between the terminal voltage (reference) and the q-axis. B. Transientprocess Using a second order model to represent the direct axis of a synchronous machines, one can obtain: where L d is the direct axis synchronous inductance, L’d, L”& T’& and T’’& are the direct axis transient and subtransient inductances and open circuit time constants. For a first order quadrature axis model and applying the same procedure used for the direct axis, yields: A y s = i, I . L, - (L, - L:). e- tlT$ 1 (3) where L, is the quadrature axis synchronous inductance,Yq and T”, are the quadrature axis subtransient inductance and the subtransient open circuit time constant. WB1-1.1 C. Description of The Load Rejection Phenomena The machine behavior in a load rejection transient process can be obtained by composing the phenomena that occur before and after to the load rejection. Thus, the linkage flux in the direct axis will be: vd + A y d = -U- cos& +id .[Ld- ( L -~ L:). vdo - ( ~ -h 1. e-t/TL 1 (4) In the same way, for the quadrature axis: = \yqo + Avs = U. sm6 + i, . \y, -t/T& ] (5) The equation of the voltage variation in a load rejection is: U, = U . sin(at - S) - i, -id kd- . I . L, ( L -~cd). - (L, - U;). e-t’T;o] - cos(wt) - (L: - vi).e-t’TL 1. where yt and Yda) are the and dcuIztted d u e s at the instant t, a is parameter vector iterativelly calculated by the following equation. au = au-1- h . (10) It is important to note that any identification method has its performance improved if the iterative process begins with adequate initial approximations for the unknown parameters. In this case, they could be obtained by using basic relationships [2]. Typical values found in text books can also be wed as messes. ms technique will be employed for the direct and quadrature axis time constants, avoiding the laborious graphical approximation. (6) lV.EXAMPLEOF APPLICATION sin(ot) The proposed method will be applied to one rounded rotor synchronous machine, the rated characteristics are presented in Table I. m e n there is no ament component in the axis (iq=Oand M),one can obtain: . .cs - Table I Rated- c % = 9375 kVA UN = 13.8 kV 1~=392A (7) In the other hand, when there is no direct axis current component, results: coscp~= 0.8 Vmc = 125 V Imc = 368 A It was not possible to determine the quadrature axis load point, however, load rejection tests under an arbitraxy axis and under the direct axis was made. Fig. 1 shows the transient (la) and subtransient (Ib) response for the direct These are the basic to the load rejection axis. Fig. 2 shows the transient (za) and s u b w e n t (2b) description and they could be used to identify the responseforthearbitrary~. synchronous machines parameters. Thus,two procedures are proposed in order to simplify the identitication process, and A. mcjsparameters spread the use of load rejection tests: U, =U.sin(ot-6)-ip. -tlT& ].cos(mt) (8) Determine all the synchronous machines parameter with a load rejection under an arbitrary condition using (6); or, 0 Make a load rejection test in the direct axis, using (7) to determine the direct axis parameters, and make another load rejection under an arbitrary axis to calculate the quadrature axis parameters with (8), using the previous calculated direct axis parameters as constants. The quantities involved in the load rejection process was: P 0 p.u. Q, = -2.930 M v ~= -0.3125 p.u. = 0.972 p.u. U, =13416V I, = 126.1 A = 0.321 p.u. U, =84OOV = 0.609 p.u. Po EOMW The following typical values was adopted as initial guesses for the time constants: III. PARAMETERSIDENTIFICATION The syncbronous machines parameters will be identified in order to minimize the error between the theoretid model (6, 7 or 8) and the experimental data obtained. The henberg-hhquardt method [5] will be employed to minimize the goodness function defined as: TAo = 5 5 TJ0 = 0-05s Applying (7), the refined parameters were obtained with relative error smaller than 1%, after 6 iterations. WB1-1.2 SUBTRANSIENT STATE TRANSIENT STATE 0.00 2.00 4.00 Time [SI 6.00 8.00 (4 0.00 10.00 - Time [SI "0° 0.90 1 .eo 7.20 2.00 @> - Fig. 1 LOAD REJECTION DLRECT A X I S 1 ! 0.92 1 0.80 0.40 4 0.92 TRANSIENT STATE i 0.84 o'88 .... 0.80 I I 0.m 0.88 '---STEADYSTATE , 1 2.00 ' 4.00 1 ' Time (4 Ld = 1.1375 P.U. p d = 0.1876 P.U. L2 = 0.1074 D.U. 1 8.00 8.00 [SI ' 0.84 1 I 0.00 10.00 , I 0.40 ' 0.80 I ' I 7 7.20 Time [SI ' .oo I 2.00 (b) Fig. 2 - LOAD REJECTION - A R B m U R Y m s Table III - Adjusted quamature axisparameters T<, = 0.074s Lq = 1.055 p a . TAo = 49653 s T;Io = 0.0222 s L i = 0.1492 P.U. - 6, = 23.11' V. CONCLUSIONS B. Quadrature axis parameters The work showed a mathematical modeling of the load rejection phenomena, allowing the parameter identification with numerical methods. This technique avoids the employment of a load angle meter, making possible its generalizeduse. The identification process explores all the transient, subtransient and steady state information, and not only specified points, reducing the errors and conducting to more representative parameters. The electrical quantities are: Po = 2.810 MW Qo = -1.770 M v U, = 13390 V I, = 143 A U, = 11400 = 0.300 p.u. ~= -0.189 P.U. = 0.970 p.u. = 0.365 p.u. = 0.826 p.u. VI. REFERENCES The power angle will be also identified. Its initial value was obtained using the following approximation: [11 IEEE std 1 15 A - "Test Procedures fa SynchronousMachines" [2] F.P. de Melio; J.R Ribeiro; "Derivation of synchronous machine parameters ffom tests" - IEEE T m on PAS- 96, n 4, pp 1211-1 218. [3] F.P. de Mello; "Measurement of synchronous machine rotor angle from analysis of zero sequence harmonic components of machine terminaI voltage" IEEE Trans. on PWRD- 9, n 4, oct 1994, pp 1770-1777. [4] T.H.Ling T.N. Wen; T.D. Sheng, L.J. Yu; "Mathematical model of synchmnous machine in transient process" in ICEM88. [5] W.H. Press;B.P. Flannq, S.A Tenkolsky; W.T. Vetterling ''Nuiuerical Recipes in C"- CambridgeUnivmity Press, 1988. UK. [6] Krause, P.C.; No& F.; Skvarenina, T.L.; Olive, D.W.; "The theory of neglecting statortransients". IEEE Trans.on PAS- 98, No1, pp 141-148. - Using (S), after 7 iterations, the adjusted parameters with relative error smaller than 1% parameters are: WB1-1.3