基于小波变换和优化的Elman神经网络的光伏功率预测

Photovoltaic Power Prediction Based on Wavelet Transform and Optimized Elman Neural Network

准确地预测光伏发电功率,有利于提高电网系统运行的可靠性和经济性。分析各个气象因素对光伏发电功率的影响,确立了关键的气象因素,并利用小波分析获得气象因子样本集和光伏功率样本集不同频带下的小波系数作为神经网络的输入训练集,结合Elman神经网络建立不同天气条件下的光伏功率预测模型。提出基于自适应遗传算法优化的Elman神经网络模型,优化后的Elman神经网络在晴天、阴天、雨天3种情况下预测值的平均相对误差率分别为5.43%、8.26%、14.15%,相较于Elman神经网络分别降低了13.16%、16.61%、17.33%,改善了Elman神经网络的预测精度,提高了Elman神经网络的学习能力和泛化能力,验证了所提方法的有效性。

Accurate prediction of photovoltaic power generation is conducive to improving the reliability and economy of the system operation of the power grid.In this paper,the influence of various meteorological factors on photovoltaic power generation is analyzed,and the key meteorological factors are established,and the sample set for meteorological factors and photovoltaic power sample set under different frequency bands of wavelet coefficients are obtained as input of the neural network training set by using the wavelet analysis.Combined with Elman neural network,the photovoltaic power prediction model under different weather conditions is established.An optimal Elman neural network method based on adaptive genetic algorithm is proposed.The average relative error rate of the optimized Elman neural network prediction in sunny,cloudy and rainy days is5.43%,8.26%and 14.15%respectively.Compared with Elman neural network,they are reduced by 13.16%,16.61%and17.33%,respectively.The prediction accuracy of Elman neural network is enhanced,and the learning ability and generalization ability of Elman neural network are improved,thus the effectiveness of the proposed method is verified.