基于LSTM的光伏发电短期预测模型研究  

Research on Short-Term Forecasting Model of Photovoltaic Power Generation Based on LSTM

作　　者：林晨翔 谢炜 郑州 翁宇游 马腾 LIN Chen-xiang;XIE Wei;ZHENG Zhou;WENG Yu-you;MA Teng(Electric Power Research Institute,State Grid Fujian Electric Power Co.,Ltd.,Fuzhou 350007,China)

机构地区：[1]国网福建省电力有限公司电力科学研究院,福州350007

出　　处：《广东水利电力职业技术学院学报》2025年第2期1-4,共4页Journal of Guangdong Polytechnic of Water Resources and Electric Engineering

摘　　要：由于多种因素制约,当前短期预测准确率无法满足电力系统调度需求,对光伏发电并网影响较大。对此,研究基于LSTM的光伏发电短期预测模型,以提高电力系统调度的准确率。分析气象因素与光伏发电量的关系,将影响程度转换为数据并进行平滑处理;构建LSTM模型进行短期预测,并用时间区域原理分析预测误差;通过SOA算法校正误差,提升预测精度。测试显示,该方法预测精度高达99%,预测时间最长8分钟,具有较高应用价值。Due to various constraints,the current short-term forecasting accuracy cannot meet the scheduling requirements of the power system,which significantly affects the integration of photovoltaic power generation into the grid.In response,this study investigates a short-term forecasting model for photovoltaic power generation based on Long Short-Term Memory(LSTM)networks to improve the accuracy of power system scheduling.The relationship between meteorological factors and photovoltaic power generation is analyzed,and the degree of impact is converted into data and smoothed.An LSTM model is constructed for short-term forecasting,and the time-domain principle is used to analyze the prediction errors.The SOA algorithm is employed to correct the errors and enhance the forecasting accuracy.Testing results show that the method achieves a prediction accuracy of up to 99%,with a maximum forecasting time of 8 minutes,demonstrating significant practical application value.

关 键 词：LSTM 平滑处理 光伏发电 短期预测模型 

分 类 号：TM615[电气工程—电力系统及自动化]