关于BP算法在负荷预测相关问题探讨

The Problems of Application of BP Neural Network in Short-Term Load Forecasting

本文在研究了神经网络的模型建立问题基础上,探讨了电力负荷BP算法的建模方法,包括对BP网络模型建立中的隐含层数确定、隐含层节点数确定、训练次数与精度的关系、学习速率的选择、初始权值、训练样本的选择及归一化处理等相关问题进行了较深入定性和定量分析,并通过算例进行了比较实验,得出有益结论。

The problems about BP network applied in short - term load forecasting are discussed in this paper. As the following list：the number of hidden layers and the nodes of hidden layer, relation between learning times and precision, choice of learning rate and initializing weights, choice of train- ing data and normalize data. After making compared experiment through examples, some useful conclusions are drawn.