基于SOM-RBF神经网络的用电量预测模型研究

随着我国电网建设的高速发展,从日常电力负荷变化趋势剖析未来年度用电量已经成为电网建设的关键问题之一。根据1997~2016年湖北省年用电量及其10个影响因子的数据作为样本,提出了一种自组织特征映射神经网络(Self-organizing Feature Maps,SOM)与多变量的径向基函数(Radial Basis Function,RBF)结合的人工神经网络预测模型新方法。采用先聚类、再分类预测的方法,解决了由于RBF神经网络对于少量样本和训练样本点分散所导致的预测精度降低的问题,改进的神经网络泛化能力有所提高。结果表明:通过SOM-RBF组合算法进行预测,其相对误差维持在3%以下,平均相对误差为1.88%,预测效果较BP神经网络和RBF神经网络有较大的提升。这表明SOM-RBF组合算法可有效的用于用电量预测,具有较高的实用价值。

Research of Electricity Consumption Prediction Model Based on SOM-RBF Algorithm

With the rapid development of power grid construction in China,it has become one of the key issues to analyze the annual power consumption in the future from the trend of daily power load change.Based on the data of annual electricity consumption and its 10 influencing factors in Hubei Province from 1997 to 2016,a new method of artificial neural network prediction model combining self-organizing feature maps(SOM)and multivariable radial basis function(RBF)is proposed.The method of clustering and classification prediction is used to solve the problem of the prediction accuracy reduction caused by the scattering of a small number of samples and training samples by RBF neural network,and the generalization ability of the improved neural network is improved.The results show that the relative error of SOM-RBF is less than 3%,and the average relative error is 1.88%.The prediction effect is better than that of BP neural network and RBF neural network.This shows that SOM-RBF combination algorithm can be effectively used for power consumption prediction,and of high practical value.