人工神经元网络用于负荷静态综合特性辨识

THE STATIC COMPOSITE CHARACTERISTIC IDENTIFICATION OF ELECTRIC LOADS VIA ARTIFICIAL NEURAL NETWORKS

负荷建模是一个重要而困难的问题,其模型结构一般由经验或机理方法确定(对综合负荷一般靠经验确定),而参数则由辨识获得。对综合负荷,在经验不足的情况下,很难得到好的模型结构。本文介绍采用人工神经元网络(ANN)进行负荷静态综合性辨识的初步结果,并与阻尼最小二乘法(Marquardt法)进行了比较分析。算例表明,用ANN进行负荷静态综合特性辨识可得到十分准确的结果。

The modelling of electric loads is an important and difficult problem. Their model structures are generally determined based on experience and/or the first principle(to composite loads only on experience), and the parameters are obtained by means of the system identification methods. Whth insufficient experience, it is difficult to obtain good model structures of composite loads.A new tool, the Artificial Neural Network (ANN), is applied to identify the static composite characteristics of electric loads. The simulation results and the comparison between ANN and Marquardt Method are given. Two properties are presented for the Error Back Propagation Algorithm(BP Algorithm). An example shows that it is feasible to identify the static composite characteristics of electric loads via ANN.