基于神经网络的负荷组合预测模型研究

给出了电力系统负荷的变权系数组合预测模型,即基于神经网络的组合预测模型。该模型利用多种方法的预测结果与实际负荷数据的非线性关系,建立相应的神经网络模型。该网络为单输出的三层网络,其中输入层为各种预测方法的预测值,输出层为实际负荷值。文中用变动量因子和变学习率的BP算法对其训练,训练后的网络便具有预测能力。同时,文中对基于遗传算法的固定权系数组合预测模型进行了简要的介绍。对几个实际系统的年、月、时负荷预测表明,该模型具有很高的预测精度。

RESEARCH OF THE COMBINATION FORECASTING MODEL FOR LOAD BASED ON ARTIFICIAL NEURAL NETWORK

This paper presents a non-stationary weights combination forecasting model(NWCFM) for load ,i.e., combination forecasting model based on artificial neural network. The corresponding artificial neural network (ANN)for this model is constructed using the nonlinear relationship between the forecasting values of various methods and the actual loads. The ANN has three layers and the output layer has only one neuron. The inputs of ANN are the forecasting values of all methods and the output is the original data. The ANN, which is trained by error back propagation (BP) algorithm with variable learning rate and variable momentum, has the forecasting function. At the same time, the stationary weights combination forecasting model(SWCFM) based on genetic algorithm is concisely introduced. The model can be used in year, month, hour load forecasting fields, and so on. The effectiveness and practicability of the models have been verified by some examples.