基于Resnet-VAE融合模型的光伏功率预测方法  

Photovoltaic power prediction method based on Resnet-VAE fusion model

作　　者：汤倩茹 张哲扬 梁子怡 王悦然 花卉 TANG Qianru;ZHANG Zheyang;LIANG Ziyi;WANG Yueran;HUA Hui(School of Electric Power Engineering,Nanjing Institute of Technology,Nanjing 211167,China)

机构地区：[1]南京工程学院电力工程学院,南京211167

出　　处：《黑龙江电力》2024年第5期389-392,400,共5页Heilongjiang Electric Power

基　　金：南京工程学院本科生科技创新基金(项目编号:TB202304006)。

摘　　要：光伏发电由于其出力受太阳辐照强度、气候等因素影响较大,其功率曲线特性往往呈现出非线性特征,传统的预测方法难以进行拟合。为此,提出一种基于残差神经网络(ResNet)与变分自编码器(VAE)相结合的融合模型,采用变分自编码器对输入数据序列进行处理,并基于机组运行历史数据,通过分析不同输入属性间的关联特性,实现了对模型输入属性的降维。模型采用机组实际运行数据进行训练,实现机组功率精准预测。Because the output of photovoltaic power generation is greatly affected by solar irradiation intensity,climate and other factors,its power curve characteristics often show nonlinear characteristics,and traditional prediction methods are difficult to fit.Therefore,a fusion model based on the combination of residual neural network(ResNet)and variational auto-encoder(VAE)isproposed.Thevariational auto-encoderisused to process the input data sequence.Based on the historical data of unit operation,the dimension reduction of model input attributes is realized by analyzing the correlation characteristics between different input attributes.The model uses the actual operation data of the unit for training to achieve accurate prediction of the unit power.

关 键 词：深度学习 电力系统 数据降维 变分自编码器 光伏发电功率预测 

分 类 号：TM615[电气工程—电力系统及自动化]