Identification of exciter constants using a coherence function based weighted least squares approach

Experience with generator exciter modeling at the Ta-Kuan pumped storage power plant in Taiwan is reported. To identify exciter parameters, frequency response data are first obtained by injecting wide bandwidth signals into the system. Then a modified weighted least squares (WLS) approach using a coherence function is developed to determine exciter constants. A novel feature of the proposed coherence function based WLS method is that it can eliminate measurement dependent inaccuracies by placing less weights on measurements that are more severely corrupted by noise. Based on the established excitation system model, computer simulations are performed to examine the dynamic responses of the generating unit under small and large disturbance conditions. Field test results are also provided