The Study Of Parameter Identification Method For The Generator's Excitation Systems

Accurate excitation system models allow for more precise calculations of the power system control and stability limits. It calls for the necessity that excitation systems be modeled in detail. The desired models are those representing the actual excitation system performance for both large as well as small system disturbances. Excitation system models are intended to represent the overall dynamic behaviors of composite system components such as voltage regulator, commutator exciter, alternator, terminal voltage transducer, etc. The representation for the effects of each individual component has yielded several different types of models. Consequently, various models have been proposed and standardized. Once a model structure is set, the next step is to derive unknown parameter values in the model structure. This thesis concentrates in parameter identification method suitable for excitation systems. The structure and characteristics of various excitation systems are introduced firstly, then the traditional identification methods, frequency domain identification (Fast Fourier Transform / Least Square Equation) and time domain identification (Piece Linear Polynomial Function), are studied. The thesis introduces the principium of the two methods and analyzes their merit and defect in real application. Based on principle induction and large number of simulation tests, the guidelines of parameter settings for the time domain method and frequency domain method are summarized. It develops the efficiency of the two traditional identification methods. The traditional parameter identification methods for the generator's excitation systems can't identify nonlinear systems. So the paper introduces genetic algorithm (GA) in the parameter identification that makes the nonlinear system parameters identification possible. The results of research and application indicate that the methods can validly identify non-linear systems and has the advantages of high accuracy and good convenience for operation. On the basis of study for those methods, the software packet based on MATLAB/SIMULINK that integrates three parameter identification methods is produced, which has the practical content and friendly interface.