基于LSTM网络的光伏发电功率预测研究  

Research on Photovoltaic Power Generation Power Prediction Based on LSTM Network

作　　者：何启鹏 孙明扬 杨苗苗 徐连生 李健 贺晓阳[2] HE Qipeng;SUN Mingyang;YANG Miaomiao;XU Liansheng;LI Jian;HE Xiaoyang(Bijie Qixingguan Industrial Development Co.Ltd.,Bijie 551700,China;School of Information Science and Engineering,Dalian Polytechnic University,Dalian 116034,China)

机构地区：[1]毕节七星关工业发展有限公司,贵州毕节551700 [2]大连工业大学信息科学与工程学院,辽宁大连116034

出　　处：《照明工程学报》2025年第1期72-79,共8页China Illuminating Engineering Journal

基　　金：黔科合平台人才“贵州致福光谷科技企业孵化器”(QYFHQ[2021]003)。

摘　　要：随着全球向新能源的转型,清洁能源需求不断增加,光伏发电作为其中的重要组成部分,其功率预测的准确性对于电网的稳定运行和裕度分配至关重要。基于物理模型、相似日方法和统计方法等传统预测模型因缺乏历史记忆能力而导致自身鲁棒性较差、适应能力较弱。为了解决上述问题,本文提出了一种基于LSTM网络的光伏发电功率短期预测方法。在预处理过程中,本文先将得到的数据中的坏点和离散点进行清洗;利用斯皮尔曼相关系数分析光伏发电功率与环境影响因子之间的相关性,得到输入变量;建立LSTM网络的光伏发电功率短期预测模型,并通过与BP和决策树模型的比较,验证了其在时序数据动态捕捉和预测精度上的显著优势。LSTM模型所提供的预测结果精准可靠,能有效辅助电力调度决策。With the global transition to new energy sources,the demand for clean energy is continuously increasing.Photovoltaic(PV)power generation,as an important component of this shift,has its power prediction accuracy crucial for the stable operation of the power grid and the allocation of spare capacity.Traditional prediction models based on physical models,similar day methods,and statistical methods suffer from poor robustness and weak adaptability due to their lack of historical memory capabilities.To address these issues,this paper proposes a short-term prediction method for PV power generation power based on Long Short-Term Memory(LSTM)networks.In the preprocessing stage,this paper first cleans the bad and discrete points in the obtained data;it then uses the Spearman correlation coefficient to analyze the correlation between PV power generation power and environmental impact factors,obtaining the input variables.A short-term prediction model for PV power generation power based on LSTM networks is established,and its significant advantages in dynamic capture of time series data and prediction accuracy are verified by comparison with BP and decision tree models.The prediction results provided by the LSTM model are accurate and reliable,effectively assisting in power dispatch decision-making.

关 键 词：光伏发电功率预测 LSTM 斯皮尔曼相关系数 

分 类 号：TM615[电气工程—电力系统及自动化]