Research On Parameter Identification Method Of Steam Turbine Generator Excitation System

The parameters of generator excitation system are one of the four parameters of power system operation.The accurate measurement of generator excitation system parameters and the correct mathematical model are the important foundation to ensure the stable operation,safety and reliability of power system.Compared with the traditional time domain and frequency domain identification methods,the modern artificial intelligence algorithm can identify the nonlinear part of the excitation system more effectively,so as to solve the problem of parameter identification better.Taking the excitation system of a 200 MW steam turbine generator set in Heilongjiang Province as an example,this paper combined genetic algorithm and particle swarm optimization algorithm to identify important parameters in the system.The details are as follows:(1)Genetic algorithm is adopted to identify the single link and the whole system respectively.The genetic algorithm sets several sets of parameters,and then applies the same excitation signal to each set of parameters to get the corresponding output,and then compares the output result with the actual output result of the system.For a single link,the average identification error of system parameters is about 20%,and the identification error of some parameters is relatively large.For the whole system,the identification error is about 30%,which is larger than that of a single link.(2)Due to the shortcomings of genetic algorithm in the late identification period,such as slow calculation speed and prematurity,particle swarm optimization and its improved algorithm are adopted in this paper to identify the single link and the whole system.In order to overcome the shortcomings of PSO,such as precocity convergence and poor local search ability in late evolution,based on standard PSO,this paper introduces a second-order oscillation link and adds 4 new random parameters ξ1,ξ2,ξ3 and ξ4 to the updating formula of PSO evolution speed.This algorithm increases the complexity of particle swarm,improves the global convergence,and further improves the calculation accuracy.Compared with genetic algorithm,particle swarm optimization algorithm has smaller identification errors for single link and the whole system.The average identification error of particle swarm optimization algorithm and its improved algorithm for single link is less than 2%,and the average identification error of the whole system is less than20%,and the improved particle swarm optimization algorithm for the whole system is less than(3)In order to solve the global convergence and local optimization problems caused by genetic algorithm and standard particle swarm optimization algorithm in the identification of partial parameters,this paper combines the improved second-order oscillating particle swarm optimization algorithm with an adaptive GA algorithm to further improve the global convergence and calculation accuracy.Compared with the single optimization algorithm,the identification accuracy and accuracy of the improved GA-PSO algorithm are significantly improved,no matter for the identification of a single link or the identification of the whole system.Moreover,the improved GA-PSO algorithm can converge quickly and achieve fast solution.In this paper,an improved GA-PSO generator excitation system parameter identification algorithm can be used to identify synchronous generator excitation system model more accurately,which is of great significance for the study of generator stability and the formulation of reasonable operation mode.