INTERNATIONAL JOURNAL OF ENERGY RESEARCH Int. J. Energy Res., 23, 719 } 726 (1999) SYNCHRONOUS MACHINE PARAMETERS ESTIMATION USING A NEW GENETIC-BASED ALGORITHM KHALED M. EL-NAGGAR* AND HOSAM K. M. YOUSSEF Electrical Engineering Department, College of Technological Studies, Hawalli 32084, P.O. Box 5378, Kuwait SUMMARY A new genetic-based algorithm (GA) for estimating synchronous machine parameters from frequency tests is presented in this paper. GAs are general search techniques based on biological concepts and are very suitable for solving optimization problems. The proposed method uses a set of digital measurements for the direct axis impedance magnitude and phase as functions of frequency for estimating both the d- and q-axis parameters, such as direct reactance and time constants. The problem is formulated as an optimization problem and solved using the proposed method. Two di!erent models along with di!erent "tness functions are suggested to be used with the genetic algorithm. A practical example from the literature is used to test the proposed algorithm. The results obtained are compared with those given in the literature using other methods. The results and comparison show that the new algorithm is very applicable and highly accurate. Copyright  1999 John Wiley and Sons, Ltd. KEY WORDS: genetic algorithms; synchronous machine parameters; state estimation 1. INTRODUCTION Synchronous machine parameters can be estimated by collecting a set of test data and then searching for the best set of parameters which produce results as close as possible to the measured data. Di!erent functions are usually used to measure the accuracy of the "tness. Accordingly, estimating synchronous machine parameters and determining their characteristics has been performed in the literature through di!erent techniques. Some of these techniques are based on results obtained from standstill tests, while others are based on results obtained from frequency tests (De Mello and Hannet, 1983). Among the techniques used with results obtained from frequency tests are the least-error- squares estimation technique, Kalman "lter algorithm and the non-iterative least absolute value parameter estimation technique (Eitelberge and Harley, 1987; Mamb et al., 1981; Soliman et al., 1989). An application of a discrete time-dynamic "lter based on least absolute value approximation is introduced for parameter estimation from frequency test in El-Naggar (1996). Due to the nature of this estimation problem, it seems to be another "eld where genetic algorithms (GAs) can perform satisfactorily since they have recently received much attention as being robust stochastic search algorithms. This class of methods is based on the mechanism of natural selection which combines the notion of survival of the "ttest with random and yet structural search parallel evaluation of points in the search space. GAs have been successfully applied in various areas such as load #ow problems, fault estimation, unit commitment and reactive power control (Goldberge, 1989; Wen and Zhenxiang, 1995; Lee et al., 1995; Nims et al., 1997). This paper presents a new application of GAs in estimating synchronous machine parameters using data obtained from frequency tests. The proposed method is tested using a practical example to demonstrate its * Correspondence to: K. M. El-Naggar, Electrical Engineering Department, College of Technological Studies, Hawalli 32084, P.O. Box 5378, Kuwait. Email: knaggar@ctsms.paaet.edu.kw CCC 0363}907X/99/080719 } 08$17.50 Received 16 October 1998 Copyright  1999 John Wiley & Sons, Ltd. Accepted 29 October 1998