节能调度下短期预估分布式光伏发电功率模型研究  

Study of Short-Term Estimated Distributed Photovoltaic Power Model Under Energy-Saving Scheduling

作　　者：李丽 张振臣 刘国军 LI Li;ZHANG Zhenchen;LIU Guojun(Luannan County Power Supply Branch,State Grid Jibei Electric Power Co.,Ltd.,Tangshan 063500,China)

机构地区：[1]国网冀北电力有限公司滦南县供电分公司,河北唐山063500

出　　处：《自动化仪表》2023年第11期106-110,共5页Process Automation Instrumentation

摘　　要：在节能调度的背景下,分布式光伏发电站易受天气和环境的影响,使得光伏发电功率存在一定的不稳定性和随机性,难以实现精准预测。将长短时记忆(LSTM)网络应用到节能调度下的分布式光伏发电功率短期预估中。将影响分布式光伏发电功率的外部因素看作主要因素,对映射后的新数据序列采用主成分分析法降维处理,以填补缺失和损坏的发电功率信息。归一化处理所有数据,以提高准确度。在LSTM框架中,建立分布式光伏发电功率短期预估模型,通过设置相关参数后实现精准预测。对4个不同季节连续3天光伏发电功率预测的试验证明,所提方法预测结果接近实际结果、误差较小。In the context of energy-saving scheduling,distributed photovoltaic power plants are susceptible to weather and environmental influences,which makes the photovoltaic power a certainty unstable and stochastic,and makes it difficult to realize accurate prediction.The long short-term memory(LSTM)network is applied to the short-term prediction of distributed photovoltaic power under energy-saving scheduling.The external factors affecting distributed photovoltaic power are regarded as the main factors,and the mapped new data series are processed using principal component analysis to reduce the dimensionality and fill in the missing and damaged power information.Normalization is used to process all the data to improve the accuracy.In the LSTM framework,a short-term prediction model of distributed photovoltaic power is established,and accurate prediction is realized after setting relevant parameters.Experiments on photovoltaic power prediction for three consecutive days in four different seasons prove that the prediction results of the proposed method are close to the actual results with small errors.

关 键 词：分布式光伏发电 长短时记忆网络 功率预估 主成分分析法 降维处理 归一化处理 发电功率 节能发电 

分 类 号：TH39[机械工程—机械制造及自动化]