基于遗传算法小波神经网络的光伏电站发电量预测方法

Power Generation Forecasting Methods of Photovoltaic Power Plants Based on Genetic Wavelet Neural Network Method

针对光伏电站发电量预测不准确及多种气象因素下预测结果出现波动的问题,提出一种基于遗传算法小波神经网络(GA-WNN)的光伏电站发电量预测方法。首先,以反向传播(BP)神经网络的结构为框架,选择小波基函数作为隐含层的传递函数,将网络连接权值、小波函数伸缩因子、小波函数平移因子视为遗传个体,并通过遗传算法(GA)进行个体寻优以得到网络最优初始参数;然后,利用优化后的网络进行仿真预测,并对仿真数据进行分析;最后,将预测结果与实际发电量进行对比,以评估预测模型的误差和可靠性。实例分析表明,GA-WNN预测模型具有更小的误差和更高的预测精度,适用于精确预测光伏电站的发电量。

Aiming at the problems of inaccurate prediction of photovoltaic power plant power generation and fluctuations in prediction results caused by meteorological factors,this paper proposes a genetic wavelet neural network method is proposed for predicting the power generation of photovoltaic power plants.Firstly,using the structure of the back propagation(BP)neural network as the framework,the wavelet basis function is selected as the hidden layer.The network connection weight,wavelet function scaling factor,and wavelet function translation factor are considered as genetic individuals,and the optimal initial parameters of the network are obtained through individual optimization using genetic algorithms.Then,the optimized network is used to perform simulation prediction and the simulation data is analyzed.Finally,the prediction results are compared with actual power generation to evaluate errors and reliability of the prediction model.Experimental analysis shows that the genetic wavelet neural network prediction model has smaller errors and higher prediction accuracy,making it suitable for accurate prediction of power generation in photovoltaic power plants.