摘要
为提高光伏发电功率预测的准确性,提出一种基于改进自适应因子与精英反向学习策略的改进灰狼算法(IGWO),用以优化长短期记忆网络(LSTM)预测模型。利用IGWO优化LSTM全连接层参数,建立IGWO-LSTM组合模型预测光伏功率,具有较好的收敛速度与求解效率,也可有效避免局部最优解。最后基于常州某光伏发电站实时数据进行仿真,实验结果表明IGWO-LSTM相对于LSTM光伏功率预测更具准确性。
Improving the accuracy of PV power prediction is important for improving the operational efficiency of PV power plants and ensuring the safety and stability of grid-connected PV operation.Therefore,an improved gray wolf algorithm(IGWO)based on improved adaptive factor and elite backward learning strategy is proposed to optimize the long short-term memory network(LSTM)prediction model.The IGWO is used to optimize the LSTM fully connected layer parameters and build a combined IGWO-LSTM model to predict PV power,which has better convergence speed and solution efficiency,and also can effectively avoid local optimal solutions.Finally,based on the simulation of real-time data from a PV power station in Changzhou,the experimental results show that the IGWO-LSTM has more accuracy than the LSTM PV power prediction.
作者
薛阳
燕宇铖
贾巍
衡雨曦
张舒翔
秦瑶
Xue Yang;Yan Yucheng;Jia Wei;Heng Yuxi;Zhang Shuxiang;Qin Yao(College of Automation on Engineering,Shanghai University of Electric Power,Shanghai 200090,China;Shanghai Solar Energy Engineering Technology Research Center,Shanghai 200241,China)
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2023年第7期207-213,共7页
Acta Energiae Solaris Sinica
基金
国家自然科学基金(52075316)
上海市2021年度“科技创新行动计划”(21DZ1207502)
国网浙江省电力有限公司科技项目(5211HZ17000F)。
关键词
光伏发电
长短期记忆网络
优化算法
灰狼算法
精英反向学习策略
PV power generation
long short-term memory
optimization
gray wolf optimizer
elite backward learning strategy
