基于人工神经网络的综合负荷模型

指出了BP神经网络应用于动态综合负荷建模时存在的缺陷。提出了一种适合描述综合负荷动态特性的具有内反馈功能的动态Elman神经网络负荷模型，并采用改进遗传算法作为优化算法对某220 kV变电站综合负荷采集样本进行建模。大量建模实践表明，文章所提出的动态Elman神经网络综合负荷模型具有结构简单、参数少、应用简便、对综合负荷动态特性描述能力强等优点；Elman神经网络不仅对动态负荷建模具有良好的实用价值，也是一种很适合于电力系统其他动态非线性辨识的神经网络模型结构。

Compositive Load Model Based on Artificial Neural Network

The authors point out the defects in the application of BP neural network in dynamic compositive load modeling; then propose a dynamic Elman neural network based load model that is suitable to describe dynamic characteristics of compositive load and possesses internal feedback function, and conduct the modeling for acquired samples of compositive load of a certain 220 kV substation by using improved genetic algorithm as optimization algorithm. A lot of modeling practices show that the proposed dynamic Elman neural network based compositive load model possesses such advantages as simple structure, less parameters, convenient to apply and strong ability to describe dynamic characteristics of compositive load etc. The Elman neural network is not only practicable for the modeling of dynamic load, but also a kind of neural network structure suitable to other dynamic nonlinear identifications in power system.