摘要
针对短期光伏功率预测中影响因素挖掘不充分,以及传统反向传播神经网络(back propagation neural network,BPNN)预测模型动态参数更新难以抵达全局最优解的问题,提出考虑相似日优选的精英保留策略遗传算法(elitist preservation genetic algorithm,EPGA)改进BPNN短期光伏功率预测模型。首先,根据灰色关联分析(grey relation analysis,GRA)对影响短期光伏出力预测准确度的气象因素进行深入分析,引入GRA和皮尔逊相关系数(Pearson correlation coefficient,PCC)综合指标选取相似日;其次,采用具有全局搜索特性的EPGA算法进一步改进BPNN参数更新的方法,从而构建了EPGA-BPNN模型;最后,利用光伏发电功率历史数据集进行算例分析。结果表明,改进模型EPGA-BPNN在训练速度与短期光伏功率预测精度方面表现更为优异,从而验证了所提算法模型的有效性和优越性。
To address the issues of limited exploration of influencing factors in short-term photovoltaic power prediction and the challenge of dynamically updating parameters in traditional back propagation neural network(BPNN)prediction models to achieve the global optimal solution,the elitist preservation genetic algorithm(EPGA)is proposed to improve the short-term photovoltaic power prediction model of BPNN by considering the similar day preference.Firstly,the meteorological factors affecting the accuracy of short-term PV output prediction are analyzed in depth according to the grey relation analysis(GRA)algorithm,and the combined index of GRA and Pearson correlation coefficient(PCC)is introduced to select similar days.Secondly,the EPGA algorithm with global search characteristics is used to further improve the method of BPNN parameter updating,so that the EPGA-BPNN model is constructed.Finally,an arithmetic example is analyzed using the PV power history dataset.The results show that the improved model EPGA-BPNN performs better in terms of short-term PV power prediction accuracy,thus verifying the effectiveness and superiority of the proposed algorithmic model.
作者
王丹
薛激光
旋璇
张笑怡
杨婷
王玉莹
WANG Dan;XUE Jiguang;XUAN Xuan;ZHANG Xiaoyi;YANG Ting;WANG Yuying(Marketing Service Center of State Grid Liaoning Electric Power Company,Shenyang 110179,China;Department of Electrical Engineering,Nanjing Institute of Technology,Nanjing 211167,China)
出处
《供用电》
北大核心
2024年第5期80-87,共8页
Distribution & Utilization
基金
江苏省产学研合作项目(BY2022056)。
