摘要
准确的光伏功率预测不仅是光伏电站并网安全运行的重要保证,还能有效减少弃光现象的发生,提高光伏能源利用效率。针对当前我国新能源功率预测3天的时间尺度与火电5-7天启停周期不匹配的情况,从提高光伏消纳的角度出发提出了一种基于云模型-LSTM的光伏功率中期预测方法。首先建立辐照度云模型,通过云模型相似度计算挖掘相似日;将相似日数据代入长短期记忆(Long-short Term Memory,LSTM)神经网模型进行训练;最后进行光伏功率预测。考虑季节差异,分别在四个季节随机选取一周进行光伏功率预测,结果表明云模型-LSTM的预测准确性较传统LSTM、SVM以及GM模型均有一定提升。
Accurate photovoltaic(PV)power forecast is not only an important guarantee for the safe operation of PV power station grid-connection,but also can effectively reduce the occurrence of light abandonment and improve the efficiency of PV energy utilization.In view of the mismatch between the 3-day time scale of new energy power prediction and the 5-7-day start-up/shutdown cycle of thermal power plants in China,this paper proposes a midterm PV power forecast method based on cloud model-LSTM.Firstly,the irradiance cloud model is established,and similar days are mined by cloud model similarity calculation.Then,the similar-day data are substituted into the LSTM neural network model for training.Finally,the photovoltaic power is predicted.With seasonal differences considered,one week is randomly selected in four seasons respectively to make photovoltaic power prediction.The results show that compared with the traditional LSTM,SVM and GM models,the forecast accuracy of cloud model-LSTM is improved to a certain extent.
作者
张晋华
黄远为
冯源
ZHANG Jinhua;HUANG Yuanwei;FENG Yuan(School of Electric Power,North China University of Water Resources and Electric Power,Zhengzhou Henan 450045,China;State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources,School of Renewable Energy,North China Electric Power University,Beijing 102206,China)
出处
《湖北电力》
2021年第2期49-55,共7页
Hubei Electric Power
基金
国家重点研发计划项目(项目编号:2019YFE0104800)。
关键词
光伏
中期预测
云模型
深度学习
LSTM
photovoltaic
midterm forecast
cloud model
deep learning
LSTM
