结合太阳辐射量计算与CNN-LSTM组合的光伏功率预测方法研究
PHOTOVOLTAIC POWER PREDICTION METHOD COMBINATING SOLAR RADIATION CALCULATION AND CNN-LSTM
王东风 1刘婧 1黄宇 1史博韬 1靳明月1
作者信息
- 1. 华北电力大学自动化系,保定 071003
- 折叠
摘要
为了提高模型预测性能,提出一种综合太阳辐射模型及深度学习的光伏功率预测模型.首先,利用太阳辐射机理建立太阳辐射模型(SRM),估算出水平面上总辐射值,再由斜面辐照度转换方法计算出光伏组件所接收的斜面辐射值.其次,通过皮尔逊相关分析法筛选出对光伏功率影响较大的主要因素,将斜面辐射计算值及主要影响因素作为输入,采用卷积神经网络(CNN)和长短期记忆网络(LSTM)建立光伏功率SRM-CNN-LSTM预测模型.分别利用春夏秋冬四季典型日的数据开展对比实验,结果表明:与几种其他方法相比,该文方法具有更好的预测效果.
Abstract
Precise photovoltaic power forecast is helpful for grid dispatching and secure operation.To enhance the forecast performance of the model,a PV prediction model combining the solar radiation model and deep learning is suggested.Firstly,the solar radiation model(SRM)is built using the solar radiation mechanism to estimate the total radiation value on the horizontal plane.Then the inclined plane radiation value received by the inclined photovoltaic panel is calculated by the inclined plane irradiance conversion method.Secondly,Pearson correlation analysis is devoted to screen out the primary factors influencing greatly photovoltaic power.Finally,the calculated value of inclined plane radiation and the major influencing factors are taken as input and derived from convolutional neural network(CNN)and long short-term memory(LSTM)network to build the PV power SRM-CNN-LSTM prediction model.Comparative experiments are carried out with the data from typical spring,summer,autumn,and winter days.The results show that the suggested method has better forecast effect compared with several other methods.
关键词
光伏发电/预测/太阳辐射/神经网络/卷积神经网络/长短期记忆网络Key words
photovoltaic power/forecasting/solar radiation/neural network/convolutional neural network/long short-term memory network引用本文复制引用
基金项目
中央高校基本科研业务费专项资金(2021MS089)
出版年
2024
NETL
NSTL
维普
万方数据