基于相似日理论和LCSSA-BP的短期光伏发电功率预测  

Short-Term Photovoltaic Power Prediction Based on Similarity Day Theory and LCSSA-BP

作　　者：周新茂 郑焮元 于正鑫 王笑伟 曹英丽[1] ZHOU Xinmao;ZHENG Xinyuan;YU Zhengxin;WANG Xiaowei;CAO Yingli(School of Information and Electrical Engineering,Shenyang Agricultural University,Shenyang 110866,Liaoning,China)

机构地区：[1]沈阳农业大学信息与电气工程学院,辽宁沈阳110866

出　　处：《电网与清洁能源》2022年第11期88-97,共10页Power System and Clean Energy

基　　金：国家自然科学基金项目(61903264)。

摘　　要：准确预测光伏发电功率是保障含分布式电网平稳运行的关键环节。为提升反向传播神经网络(BPNN)功率预测精度,提出一种基于Logistic混沌映射的麻雀搜索算法(LCSSA)以改进BPNN的预测模型。利用相关性分析确定光伏发电功率的影响因素,并引入与天气类型密切相关的晴空指数作为选取相似日的气象因素;利用欧氏距离和马氏距离组合加权法选取训练集;建立LCSSA-BPNN功率预测模型,利用实测数据对比分析所提LCSSA-BPNN模型与SSABPNN、BPNN模型的预测精度。结果表明:在晴天、阴天、雨天3种情况下,LCSSA-BPNN模型预测值的平均相对误差率分别为9.52%、10.52%和11.56%,均优于其他对比模型,说明LCSSA-BPNN预测模型具有更好的适应性和预测性能。Accurate prediction of photovoltaic power generation is of great significance to grid dispatching management after grid connection. In this paper,a Sparrow Search Algorithm(LCSSA)based on Logistic chaotic mapping is proposed to improve the prediction model of back propagation neural network. Firstly,the correlation analysis is used to determine the key factors that affect the photovoltaic power generation,and the clear sky index closely related to the weather type is introduced as the meteorological factor for selecting similar days. Secondly,the training set is selected by the combination weighting method of Euclidean distance and Mahalanobis distance to improve the reliability of the training set. Finally,the LCSSA-BPNN power forecasting model is established, and the forecasting accuracy of the proposed LCSSA-BP model,SSA-BPNN and BPNN models is compared with the measured data. The results show that LCSSA-BPNN model has high accuracy in different weather types,and the average relative error rates of the predicted values in sunny,cloudy and rainy days are 9.52%, 10.52% and 11.55%respectively. It shows that LCSSA-BPNN model has better adaptability and prediction performance.

关 键 词：反向传播神经网络 光伏功率预测 麻雀算法 Logistic混乱映射 

分 类 号：TM615[电气工程—电力系统及自动化]