Maximum Likelihood Estimation of Solid-Rotor Synchronous Machine Parameters from SSFR Test Data

The fidelity of synchronous machine models is affected by the proposed model structures, the quality of the experimental data used to identify the model's parameters, and the robustness of the estimation technique. This paper presents an approach to study the effect of noise on the estimated parameters. The robustness requirement specifies that the estimation technique should require a minimum a priori knowledge of the initial values of the parameters. Also, the estimation algorithm should not be affected by noise that is present in experimental data. These two issues can be dealt with effectively using maximum likelihood (ML) estimation, as proposed in this paper. In the companion paper [1], it was established that multiple parameter sets will be obtained when transfer functions of a solid rotor synchronous machine are estimated from noise-corrupted, frequency-domain data and then, the machine parameters are computed from the estimated machine transfer functions; time constants. Moreover, the estimated machine parameters are very sensitive to the value of the armature resistance used in the study. In this paper, a time-domain identification technique is used to estimate machine parameters. The objective is to show that the multiple solution set problem encountered in the frequency response technique can be eliminated if the time-domain estimation technique is used. For this purpose, the time-domain data are generated from the d - q axis transfer functions estimated for the SSFR test data. The maximum likelihood (ML) estimation technique is then used to estimate the machine parameters.