山东电力技术2024，Vol.51Issue(7)：10-18,9.DOI:10.20097/j.cnki.issn1007-9904.2024.07.002
基于3DCNN与CLSTM混合模型的短期光伏功率预测
Short-term Photovoltaic Power Prediction Based on 3DCNN and CLSTM Hybrid Model
摘要
Abstract
Photovoltaic(PV)power prediction is crucial for the formulation of power generation plans and coordinated dispatch in the power system.However,due to the stochastic and intermittent nature of PV generation,there is still significant room for improvement in the accuracy of PV power prediction.Therefore,a hybrid model based on a three-dimensional convolutional neural network(3DCNN)and convolutional long short-term memory network(CLSTM)for PV power prediction was proposed,which combines the strengths of the 3DCNN and CLSTM models to enhance prediction accuracy.The prediction model using three metrics:root mean square error(RMSE),mean absolute error(MAE),and mean absolute percentage error(MAPE)was evaluated.By applying the prediction method based on the hybrid model to the output prediction of a specific PV plant,the applicability and correctness of the model were validated.The results show that when the input time series of solar irradiance,temperature,humidity,wind speed,etc.are the same,the hybrid model based on 3DCNN and CLSTM achieves the best prediction performance.Compared to the individual 3DCNN model,CLSTM model,and backpropagation neural network(BPNN)model,the hybrid model improves MAPE by 54%,61%,and 64%,respectively.It was indicated that the hybrid model can better adapt to the stochastic and intermittent characteristics of PV generation,thus improving the accuracy of power prediction.关键词
光伏功率预测/深度学习/三维卷积神经网络/卷积长短期记忆神经网络Key words
photovoltaic power prediction/deep learning/three-dimensional convolutional neural network(3DCNN)/convolutional long and short term memory network(CLSTM)分类
信息技术与安全科学引用本文复制引用
于丹文,李山,刘航航,李广磊..基于3DCNN与CLSTM混合模型的短期光伏功率预测[J].山东电力技术,2024,51(7):10-18,9.基金项目
国网山东省电力公司科技项目(520626220004). Science and Technology Project of State Grid Shandong Electric Power Company(520626220004). （520626220004）
