A STATE SPACE SINGLE-PHASE TRANSFORMER MODEL INCORPORATING NONLINEAR PHENOMENA OF MAGNETIC SATURATION AND HYSTERESIS FOR TRANSIENT AND

PERIODIC STEADY-STATE ANALYSIS

Sigridt Garcia Aurelio Medina Facultad de Ingenieria ElCctrica, U.M.S.N.H. Edificio A, 2". Piso, Ciudad Universitaria Morelia, Michoacin, MEXICO Carlos Perez

[email protected]

amedina6iJzeus.ccu.umich.m [email protected]

An efficient state space single-phase transformer model is presented in this contribution, which incorporates the nonlinear phenomena of magnetic saturation and hysteresis. It is analyzed under transient and periodic steady-state operation conditions. The nonlinearity of saturation and hysteresis is efficiently represented with a novel and simple formulation. Fast periodic steady-state solutions are achieved with the application of a Newton technique for the acceleration of time domain computations. Comparisons are made between conventional methods used for the numerical solution of the transformer represented by a set of Ordinary Differential Equations (ODES) and the Newton method in terms of full cycles and CPU times needed to obtain the periodic steady-state solution of the network.
Abstract:

Keywords: State Space, Single-phase, Magnetic Saturation,

Hysteresis,Periodic Steady-State, Newton Method..

1. INTRODUCTION The simplest transformer model is used in single-phase studies at fundamental fiequency, such as short-circuit and load flow analysis. Here the transformers is represented by the leakage reactance in series with an ideal autotransformers. In one of the fEst models proposed [l], the transformer is represented with a x equivalent circuit incorporating the effect of the turns ratio in each branch. However, to better describe the transformer behaviour, the electric and magnetic circuits should be incorporated in the equivalent model. The magnetic circuit has a higher degree of complexity for its appropriate representation since it is dependent on the magnetic and physical characteristics. Digital models have been developed, more complex and rigorous, that combine the electric and magnetic circuits.

However, the main interest has been around the analysis of electromagnetic transients [2], since for this type of analysis as well as for the steady-state analysis of power systems under unbalanced conditions detailed transformer representations are needed [3]. Transformer models have been described in the form of shunt admittances [4] or admittance matrices 15-71, which can be obtained form excitation and short-circuit test data. The periodic behaviour of power transformers requires of a more complete and rigorous model which could appropriately predict the harmonic distortion in the power system. The transformer behaviour during transients and interruptions are more or less difficult to simulate, depending on the nature of the phenomenon involved. There are transformer models, where the simulation results are identical with those obtained with the equivalents circuits of positive and zero sequence [8]. Other mathematical model [9] for the digital simulation of multiple windings transformers is based on the decomposition and step by step linearization and is mainly applied for the simulation of transients in circuits with nonlinear elements. A basic electro-magnetic model for the magnetizing branch of a transformer was developed using a methodology based on its representation by means of a Norton harmonic equivalent [4]. The generalization i n the Harmonic Domain, representing three-phase networks containing multiple transformer units was presented in other contribution [6]. Here each electromagnetic harmonic equivalent is described by a network circuit, in the form of a Norton equivalent which interfaces with the rest of the network according to the transformer electrical connections. A procedure for the determination of the saturation and hysteresis in the transformer has been proposed in a contribution [lo], where test field data are not required. In [ 1 I] the saturation and hysteresis effects are incorporated. The numerical analysis depends closely on the material properties under study. In this contribution the harmonic analysis for a single-phase transformers is carried-out using a flux linkage
formulation; for the incorporation of magnetic saturation and

hysteresis effects is made with a simple an efficient arctangent formulation [121 is used.

0-7803-6420-1/00/$10.00 (c) 2000 IEEE

2417

2. STATE SPACE FORMULATION n e equivalent circuit of h e transformer is illusbated in the Fig. 1.

RI
Lll

It can be observed from the(7), (8) and (10) that the currents are calculated using the initial conditions for the state variables.

Rz

LIZ

3. NONLINEAR PHENOMENA
The incorporation of nonlinear effects such as magnetic saturation and hysteresis is achieved in the transformer model with the appropriate modification of the differential equations (1 1)-(13) and the algebraic equations (7)-( 10).

.-

3.1 Magnetic Saturation

Fig. 1 Transformer equivalent circuit
Applying the Kirchhoff s Law for voltages to the circuit of the Fig. 1, the following equations are obtained,

In Fig. 2 the saturation curve is illustrated. This curve is approximated with the equation, A,(i,)=
0.32atan(i,,,)

v, = R,i, + 4
dt

where i, in terms of A,, is expressed as,

(1)

v, = R,i,

d4 +dt
-. .

(2)
01

where v2 is also obtained as,

v, = -RLi2
The flux linkages A,and d2can be written as follows,

I
(4)
(5)

4. = 4
A=Li

2+

The general flux linkage law is written as,

(6)
Fig.
Fig. 2 Saturation Curve

where the currents can be calculated as,

i, =-4 - 4 L,l

(7)

(8)

The Fig. 3 shows the comparison between the experimental saturation characteristics and the approximated curve using (14) and (1 5 ) . The experimental saturation curve has been reproduced with a very good precision.
0.5
045.
0.4

Applying the Kirchhoffs Law for currents to the node 1 results in,

.
.
.

i, + i2 = i ,
where

+ i,,,

(9)

i , = il + i2 - i ,

(10)

= 2 .-.
.B e

0.35

0 . 3

0.2s. 0.2.
0.1s

From the inner loop of Fig. 1 the following equation is obtained,

dt . where a state varialde is obtained; the other state variables are obtained from (1) and (2) as, dn, = v, - R,i, -

dn,= i,R,

":I,
R

( . .
a

/ z :

....... .... Experimental

Approximated

,
8

,I
B
10

0 0

4 5 e 7 Magnetizing current

Fig. 3 Comparison of saturation characteristics

dt
dt

-= -i2(RL

+ R,)

(13)

When magnetic saturation is to be represented then (10) is modified, the branch loss is not included and (1 1 ) is not considered.

0-7803-6420-1/00/$10.00 (c) 2000 IEEE

2418

3.2 Hysteresis

The Fig. 4 illustrates a simple and efficient model for the representation of the hysteresis phenomenon. Here R, represents the hysteresis losses and L , the nonlinear inductance where saturation characteristics should be included. The magnetizing current , i can be obtained as,

The Table 1 shows the maximum error obtained during the convergence of the transformer state variables including saturation effects. The BF method is considerably slower that the ND,whereas the BF method requires of a number of full cycles (NFC)much larger than 208 to obtain the steady-state solution, the ND method requires of only 11 periods of time. Similarly in Table 2, summarizes the resulting errors during the convergence of the transformer state variables when hysteresis effects are taken in to account.
Table 2 Errors during convergence of the transformer with

hysteresis

BF
Fig. 4 Inductor model for hysteresis representation

-Ai, Lm

deriving (15) gives,

When hysteresis is included in the transformer model, the number of periods required to reach the distorted final steadystate solution is noticeably less because the system has a higher damping compared with the case of saturation.
a) Harmonic Analysis

Substituting (1 7) in (16) results in,

Then (18) remains only a function of calculated from the initial condition.
4.

and can be

The Fig. 5 illustrates the transformer steady-state operation incorporating magnetic saturation and hysteresis effects for different loads: R,RL and RC. Only primary side transformer current is plotted.

CASESTUDY

The Newton methodology described in [13-151 to accelerate the convergence of the state variables in the time domain to the Limit Cycle [161 is now applied to obtain the steady-state solution of the transformer of Fig. 1, where nonlinear effects of magnetic saturation and hysteresis are incorporated. The test system data are given in Appendix A. The required number of state variables depends on the operation of the transformer, i.e., if the transformer is operated with or without saturation. For the representation of saturation two state variables are required whereas for the modeling of hysteresis effects three sate variables are needed. The state variables are solved with the Brute Force (BF) and Numerical Differentiation (DN) methods, respectively. Both approaches used the Fourth Order Runge-Kutta integration method. The Limit Cycle and thus the periodic steady-state solution is obtained once the maximum error in the state variables is within IOe-1Op.u. Table 1 Errors during convergence of the transformer with

0 r.grrrkW . , I

(b)

Fig. 5 Primary current waveform i,. (a) With Saturation, (b)With Hysteresis

NFC
5

8
11

I I I I
I

BF
6.8811e-4 1.5563e-5

I 6.8811e-4 I 3.1429e-8

ND

l.SS84e-S 15.1847~-14

208

1.6959e-5

With the saturation characteristic represented the harmonic content varies according to the load type. In all cases there is a large third harmonic content, see Fig 6(a) where for the R load it is approximately 25% of the fienta1. The Fig. 6(b) illustrates the case when the hysteresis effect is incorporated, the harmonic content increases notoriously for the R load. For the others types of loads the harmonics do not change noticeably.

0-7803-6420-1/00/$10.00 (c) 2000 IEEE

2419

C4R
RL 0 RC

25

20 15
10 5

primary circuit going to zero, as observed from the third segment of Fig. 7. At t=0.0588 secs. the fault is cleared producing the transient shown in the fourth section. The solution process is again accelerated until reaching the new periodic steady-state of the system; identical to the obtained previously and illustrated by the second segment of Fig. 7. With this technique of acceleration of time domain computations it is possible to obtain a detailed solution of the few transient cycles of interest and once the fault is cleared, to reach the periodic steady-state solution considerably faster than digital procedures based on conventional methods for the numerical solution of ODES. For example, it is observed from Fig. 7 that despite of the large current peaks taking place in the first and second transients the periodic steadystate is obtained in a reduced number of cycles. 5. CONCLUSIONS

EiR W RL
ORC
3 5 7 9 1 1

Fig. 6 Harmonic content of the Fig. 5 , respectively (a) With saturation, (b) With hysteresis

6 ) Transient Analysis
A transient study was carried-out applying an open source perturbation. This fault consists in opening the voltage source at t~0.0539 secs.; this source supplies the transformer primary circuit. The application of this fault condition will make evident the advantage and efficiency of obtaining an accelerated time domain periodic steady-state solution once the fault is cleared.

A model for the single phase transformer incorporating the magnetic saturation and hysteresis effects has been presented. A simple an eficient arctangent formulation is used for the representation of the nonlinear saturation and hysteresis characteristics. Ihe transformer behaviour was analyzed with three different loads: R, RL and RC, respectively. Comparisons were made between a conventional method and the Numerical Differentiation technique for obtaining the periodic steady-state solution of the system. The potential of this technique for obtaining a fast periodic steady-state solution of the system after any perturbation has been demonstrated. This new alternative for transient and steadystate analysis of circuits and electric networks has the advantages of following closely the few transient cycles and accelerating the time domain computations immediately after clearing the fault to obtain a fast steady-state solution.
ACKNOWLEDGMENTS The authors gratefully acknowledge financial support from Universidad Michoacana de San NicolL de Hidalgo to carry-out this investigation.
6.
REFERENCES [I J Stagg. G.W. and El-Abiad, A.H.,Computer Me/lrods in Power Sys/ems Annlysis, McGraw-H ill, London. [2] L c h , F., Semlyen, A., Complete Transformer Model for Electromagnetic Transients, IEEE Trnnsnc/ions on Power Delivery. Vol. 9, No. I , January 1994, pp. 23 1-239. [3] Medina, A., hrrillaga. J., Simulation of Multilimb Power Transfomers in the Harmonic Domain, IEE Proceedings, Vol. 139, No. 3, May 1992, pp.269-276 [4] Semlyen. A., hcha, E. y Arrillaga, J., Harmonic Norton Equivalent for the Magnetisin Branch of a Transformer, IEE Proceedings, Part. C , Vol. 134, No. 2. March, pp. 162-169. [ 5 ] Brandwajn, V., Dommel, H.W., Dommel, I.!., Matrix representation of Three-phase N-Winding Transformers for Steady-State and Transient Studies, IEEE Trnnsncfions on Power Appnrntrrs and Sysfems,Vol. PAS101, No. 6, June 1982. pp. 1369-1378. [6] Acha, E., Arrillaga. J., Medina, A.. Semlyen A.. General Frame of Reference for Analysis of Harmonic Distortion in Systems with Multiple Transformer Nonlinearities, IEE Proceedings, Vol. 136. Part. C. No. 5, September 1989. pp. 27 1-278.

7.

I k .

Fig. 7 Transient and steady-sate solution

The Fig. 7 illustrates @eevolution in time of the current at the primary side of the transformer. In the f m t section there is a marked transient peak as a result of the transformer energization, this solution is accelerated until the periodic steady-state is achieved. The second section illustrates the fact that being in steady-state a fault was applied at 0.005 secs. by opening the source; this results in the current at the
0-7803-6420-1/00/$10.00 (c) 2000 IEEE

2420

171Medina, A., Amllaga J., "Analysis of Transformer-Generator Interaction in the Harmonic Domain", IEE Proc-Gener. Transm. Distrib., Vol. 141, No. 1, January 1994, pp. 3846. [SI Dommel, H.W., "Transformer Models in the Simulation of Electromagnetic Transients", 5th Power Systems Computation Conference, Cambridge, England, Paper 3.1/4, pp.1-16. [9] Ivanoff, D., "Simulation of Multiwinding Nonlinear Tmsformers", Escola Politkcnica, Silo Paulo-Brasil. 1977. [IO]Chin E. Lin, Jong-Bi Wei, Ching-Lien Huang, Chi-Jen Huang, "A New Model for Transformer Saturation Characteristic by Including H y s t e r e s i s Loops", IEEE Transactions on Magnetics, Vol. 25, No. 3, May 1989, pp. 2706-2712. [I 1J Hassani M.M., Lachiver G. and Jasmin G., 'Wumerical simulation of the magnetic core of a transformer in transient operations", Canadian Conference in Eleceical and Computer Engineering, September, 1989, pp. 289-291. [12] Pkrez-Rojas C. "Representation of the saturation and hysteresis, approaching through arctangent function", (in Spanish). RVPAA-99, IEEE, Acapulco, Gro., Mtxico, July, 1999. pp. 118-121. [U] Garcia, S. "Soluci6n de redes elictricas en el dominio del tiempo con aceleracih de la cowergencia al ciclo Iimite", BEng Thesis (in Spanish), Facultad de Ingenim'a Elbctrica, UMSNH, March 1998, Morelia, Mich., Mbxico. [14] A. Medina, S. Garcia, "Time Domain Accelerated Solution of Systems with Non-Linear and Time-Varying", Proceedings of the Intemacional Conference 1998 NSF Design and Manufacturing Grantees Conference, Monterrey, N.L., Mexico, January 5-8,1998, pp. 797-798. [I51 Medina, A., Garcia, S.,"Determinaci6n de la Solucih Peribdica en Estado Estacionario de Redes Elbcb-icas Mediante T h i c a Newton en el Domini0 del Tiempo", Proceeding of the Intemacional Conference IEEE RVP98 (in Spanish), Acapulco, Go., Mexico, July, 1998, pp. 223-227. [I61 Parker, T.S., Chua, L.O., '"ractical Numerical Algorithms for Chaotic Systems", Springer-Verlag,New York,1989.

[17] Krause, P. C. "Analysisof Electric Machinery" McGraw Hill, 1986.

BIOGRAPHIES
Sigridt Garcia. Obtained her BEng Degree from the Univesidad Michoacana de San Nicolis de Hidalgo UMSNH, Morelia, Mich., Mtxico 1998. At present she is an Academic Associate at the Facultad de Ingenieria Electrica, UMSNH and is working towards her Masters degree.

Aurelio Medina Obtained his Ph.D. from the University of Canterbury,

Christchurch, New Zealand in 1992. He has worked as a Post-Doctoral Fellow at the Universities of Canterbury, New Zealand (1 year) and Toronto, Canada (2 years). At present he holds a permanent academic position at the Facultad de Ingenieria Eltctrica. UMSNH, Morelia, Mexico, where he is the Head of the Division for Graduate Studies. His research interests are in the dynamic and steady-state analysis of power systems. Carlos Perez-Rojas. Obtained the B.S.E.E. from the UMSNH (Universidad Michoacana de San NicolL de Hidalgo, Mexico) in 1990. He obtained in 1993 the M. S. in Power Systems from the UANL (Universidad Authoma de Nuevo h 6 n , Mexico). At present he is a Staff member at the Faculty of Electrical Engineering of the UMSNH. He is working towards his Ph. D. Degree at the DIE-FIME-UANL, Mexico. He is an IEEE and SIAM student member.

APPENDIX A. TEST SYSTEM DATA

The parameters of the transformer are the following, [17]: R,=10, &=1.25, R,,=lOOOy %=200 PI Lf,=Lf2=0.01 105 y L,,,=0.27952 [HI V,=156sen (at) [vl

(c) 2000 IEEE 0-7803-6420-1/00/$10.00

2421