Comparison of Recursive Parameter Identification Techniques for Computer Control of Power Systems
MetadataShow full item record
This paper outlines the main features of parameter identification in electric power systems. A series of recursive parameter identification techniques is presented, this containing the generalized — and extended least square techniques, the recursive maximum likelihood technique, the stochastic approximation method, and the gradient technique. These identification techniques are tested using a model of the interconnected power system of two areas to illustrate the applicability of discrete parameter identification for on-line computer control of power systems. The area control errors are considered as the controlled variables (system outputs). The changes in the governor gate position are the manipulated variables, and the variations in load are the disturbances. To evaluate the performance of the identification technique the integral of the error-squares criteria is used. This error is defined as the difference between the actual system output and the identified model output. A real time package to simulate the system and model of the identification techniques is designed and the performance of the system is recorded. PBRS is used as standard test signal during this study.