主成分分析结合神经网络的光伏发电量预测

PV Generation System Forecasting Model Based on Neutral Network and Principal Components Analysis

针对光伏发电量预测模型主要以气象因素、历史发电量等作为BP神经网络的输入,输入量多、数据冗余、网络难以收敛。利用主成分分析法PCA(principal components analysis)分析原来多个输入变量反映的个体信息,提取较少的几项综合性变量,减少预测模型的输入量。同时利用遗传算法优化BP网络的权值阈值建立预测模型,克服了神经网络算法的局部收敛、训练速度慢等问题。实验结果表明,该方法提高了预测精度,为解决光伏系统发电量预测提供了一种可行方法。

Most of photovohaic （PV） generation forecasting models take meteorological factors as the input parameters of back propagation （BP） neural network. However, the input parameters and redundant data cause neural network to converge difficultly. The principal components analysis （PCA） is adopted to analyze individual information of the initial input variables, then the input variables are minimized as extracting the really comprehensive variables. Thus the prob- lem of the BP neural network can be overcame by the combination of genetic algorithm and BP neural network. The ex- perimental results indicate that principal components analysis can significantly improve the precision of power predic- tion, and it provide an effective way to forecast generation power of PV system.