一种基于多神经网络的组合负荷预测模型

针对BP神经网络、RBF神经网络和小波神经网络应用于负荷预测时所遇到的问题,提出了一种基于各种神经网络的组合预测模型。该模型为单输出的3层神经网络,即将3种神经网络的预测结果作为神经网络的输入,将实际负荷值作为神经网络的输出,使训练后的网络具有预测能力。该模型能降低单个神经网络的预测风险,提高预测精度。仿真结果表明,所提出的组合预测模型的精度高于其中任一单一网络模型,也高于传统的线性组合预测模型。

A Combination Load Forecasting Model Based on Multi-Neural Networks

To solve the problems existing in the load forecasting by back propagation (BP) neural network, radial basis function (RBF) neural network and wavelet neural network, an artificial neural network (ANN) based combined load forecasting model is proposed. This model is an ANN with three layers and its output has only one neuron, i.e., taking the forecasted results of above-mentioned three neural networks as the input of the combined ANN and the actual value of load as the output, the proposed ANN possesses forecasting ability after training. With this model the forecasting risk of single neural network can be reduced and the forecasting accuracy can be improved. Simulation results show that the accuracy of the proposed combined forecasting model is higher than that of any single network model and also higher than that of traditional linear combined forecasting model.