On-line estimation of transient and sub-transient parameters with evolutionary programming

This paper presents an artificial intelligence approach of using evolutionary programming to estimate the transient and subtransient parameters of a generator under normal operation. Several cases with different operating points and noise variances are studied. The evolutionary programming is robust to search for the real values of parameters in all of these cases. The estimation using evolutionary programming is compared with that using a corrected extended Kalman filter. The comparison shows that when the data are highly contaminated by noise, evolutionary programming still gives satisfactory results; while estimation using the extended Kalman filter tends to diverge with such data.