摘要
为提高光伏短期出力预测精确问题,提出一种基于图卷积神经网络(GCN)的光伏短期出力预测方法。首先,构建考虑多气象影响因素的光伏短期出力模型,开展光伏出力影响因素和出力特性分析。其次,对光伏出力历史时序数据进行图形化转换和数据重构,构建邻接矩阵并提取光伏短期出力图形化特征数据。在多时间尺度场景下,建立基于GCN的光伏出力预测模型,并与基于长短期记忆网络(LSTM)、反向传播网络(BP)、图注意力模型(GAT)等算法的预测模型做比对分析。最后,以某地区光伏出力实测数据开展仿真验证研究,仿真结果表明所提方法具有良好的预测效果。
To improve the accuracy of short-term photovoltaic output prediction,a photovoltaic short-term output prediction method based on graph convolutional network(GCN)is proposed.Firstly,construct a short-term photovoltaic output model that considers multiple meteorological factors,and conduct an analysis of the influencing factors and output characteristics of photovoltaic system.Secondly,graphical transformation and data reconstruction are performed on the historical time series data of photovoltaic output,and an adjacency matrix is constructed to extract graphical feature data of short-term photovoltaic output.Establish a photovoltaic output prediction model based on GCN in a multi time scale scenario,and compare and analyze it with prediction models based on LSTM,BP,GAT and other algorithms.Finally,simulation validation research is conducted using measured photovoltaic output data from a certain region,and the simulation results show that the proposed method has good predictive performance.
作者
张亮
周立洋
徐晓春
李荣
李睿
Zhang Liang;Zhou Liyang;Xu Xiaochun;Li Rong;Li Rui(School of Electric Power Engineering,Nanjing Institute of Technology,Nanjing 211167,China;State Key Laboratory of Smart Grid Protection and Control(Nari Group Corporation),Nanjing 211106,China;State Grid Jiangsu Electric Power Co.,Ltd.Huai’an Power Supply Branch,Huaian 223001,China)
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2024年第8期289-294,共6页
Acta Energiae Solaris Sinica
基金
国网江苏省电力公司科技项目(J2022049)
江苏省重点研发计划(BE2017169)。
关键词
光伏发电
图卷积神经网络
图形数据结构
多时间尺度
photovoltaic power
graph convolutional neural network
graph data structure
multi-time scale