Content area
Load modeling and identification are important in power system stability analysis. This thesis tests and validates a stochastic dynamic load model, which has been accepted widely in academia but has not been tested using real PMU measurements data. Besides, three different methods, namely, the least square method, unscented Kalman filter method, and the nonlinear dynamic recursive method based on the Ornstein-Uhlenbeck process are tested and compared in identifying the load model parameters using the simulated data in IEEE 39-bus system. Also, the real PMU measurements collected from the GZB, SX, and WX substations in China Southern Power Grid are used to test the load model and three identification methods. Comprehensive studies considering the window size, update weight, initial value, and time step have been carried out to illustrate and compare the accuracy and robustness of the three methods. It has been observed in numerical studies that the least square method may not always guarantee good accuracy. The unscented Kalman filter method typically provides better estimations than the dynamic recursive method if the time constant of a load is less than 5s, while the dynamic recursive method is more accurate if the time constant of a load is larger than 6s. In the tests using real PMU measurements, the unscented Kalman filter method and the dynamic recursive method provide convergent estimation results.