基于小波变换和神经网络的同步电机参数辨识新方法

A New Method for Parameters Identification of Synchronous Electric Machine Based on Wavelet Transform and Neural Network

准确地辨识同步电机参数,是研究分析电力系统运行和控制系统设计的前提。神经网络具有信号分离能力,但传统的人工神经元模型不适合分离同步电机的三相突然短路电流。为精确辨识同步电机的瞬态参数,文中提出了一种改进的人工神经元模型,并将小波变换和改进的线性人工神经元结合起来,对采集到的同步电机三相突然短路电流进行分析处理。利用小波变换对短路电流进行预处理,并辨识得到各个时间常数：根据辨识得到的时间常数来设定神经元激发函数中时间常数的迭代初始值,用改进的人工神经元模型对短路电流进行分离,得到其中的直流、基波和二次谐波电流分量,通过简单代数运算便得到电机的瞬态参数。仿真分析和实机试验表明,该方法能够有效地分离出短路电流中的信号成分,并且提高了电机参数的辨识精度。

It is the precondition for investigating power system running and controlling system design to determine the electromagnetic parameters of synchronous electric machine exactly. The artificial neural network possesses the ability of separating a signal. But the traditional artificial neuron model is not fit to deal with the sudden short-circuit current. In order to determine transient parameters of synchronous electric machine exactly, an improved artificial neuron model is presented in this paper. By combining the wavelet transform with the improved artificial neuron model, we provide a new method for transient parameters identification. The following approach is adopted. Firstly, precondition the sudden short-circuit current by using the wavelet transform. In this stage, the time constants of the short-circuit current are identified with low accuracy. Secondly, choose the initial values of time constants of neural nodes according to the result obtained in the first step. Finally, by using the improved artificial neuron model to separate the short-circuit current, then the direct current, the fundamental component as well as the second-order harmonic of the short-circuit current are obtained. Meanwhile, the transient parameters of synchronous electric machine ar~ determined easily with higher accuracy by some simple calculations. The simulation and practical test results show that the method developed in this paper is valid for determining the transient parameters of synchronous electric machine.