摘要
针对当前光伏发电功率预测存在预测误差较大的问题,提出一种基于NARX神经网络-小波分解组合预测方法,该方法通过小波分解将历史光伏序列分解为高频和低频分量,将高、低频数据作为NARX神经网络输入,光伏输出功率作为神经网络输出进行训练得到预测输出分量,随后对其进行小波重构推出光伏发电预测数据。仿真结果表明:新的组合预测算法预测均方误差相较于传统BP神经网络降低了1.47%,并且新的预测算法将运行时间缩短近5 s。
Aiming to reduce large forecast error in current photovoltaic power generation forecast,based on wavelet decomposition-NARX neural network combined prediction method was proposed.The historical PV sequence was decomposed into high frequency and low frequency components by wavelet decomposition,and the high and low frequency data were used as NARX neural network input,and PV output power was used as neural network output to train and obtain the predicted output.Then wavelet reconstruction was used to derive PV power generation prediction data.The simulation results showed that prediction error of new combined prediction algorithm was 1.47%lower than that of traditional BP neural network,and new prediction algorithm could reduce running time by nearly 5 s.
作者
史如新
王德顺
余涛
薛金花
冯鑫振
窦春霞
SHI Ruxin;WANG Deshun;YU Tao;XUE Jinhua;FENG Xinzhen;DOU Chunxia(Changzhou Power Supply Company,State Grid Jiangsu Electric Power Co.,Ltd.,Changzhou 213000,China;Nanjing Branch,China Electric Power Research Institute,Nanjing 210009,China;Institude of Advanced Technology,Nanjing University of Posts and Telecommunications,Nanjing 210009,China)
出处
《郑州大学学报(工学版)》
CAS
北大核心
2020年第6期79-84,共6页
Journal of Zhengzhou University(Engineering Science)
基金
国家自然科学基金重点项目(61533010)
国网江苏省电力有限公司科技项目(储能提高规模化光伏发电消纳和用户用电质量的关键技术研究)。
关键词
光伏发电功率预测
小波分解
NARX神经网络
小波重构
BP神经网络
prediction of photovoltaic power
wavelet decomposition
NARX neural network
wavelet reconstruction
BP neural network