基于机理模型和XGBoost的短期光伏功率区间预测  

Short-term photovoltaic power interval prediction based on mechanism model and XGBoost

作　　者：袁威 裴玮[1] 曾锃 张瑞 滕昌志 赵振兴[1] YUAN Wei;PEI Wei;ZENG Zeng;ZHANG Rui;TENG Changzhi;ZHAO Zhenxing(Institute of Electrical Engineering,Chinese Academy of Sciences,Beijing 100190,China;Information&Telecommunication Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210000,China)

机构地区：[1]中国科学院电工研究所,北京100190 [2]国网江苏省电力有限公司信息通信分公司,江苏南京210000

出　　处：《电工电能新技术》2025年第2期89-97,共9页Advanced Technology of Electrical Engineering and Energy

基　　金：国家电网公司总部科技项目(5108-202218280A-2-394-XG)。

摘　　要：随着分布式光伏发电规模的不断扩大,准确预测光伏发电功率对于电力系统的安全稳定运行至关重要。为了提高光伏功率预测模型的准确性,提出了一种基于机理模型和极限梯度提升(XGBoost)算法的集成模型应用于短期分布式光伏功率概率区间预测。首先,结合气象数据,应用基于密度的空间聚类算法设计光伏功率数据治理方法,以剔除历史数据中的异常数据;其次,基于筛选后的优化样本构建集成模型,具体来说,基于机理模型构建基础预测模型对光伏功率进行初步预测,将预测结果和其他气象数据作为XGBoost模型的输入变量,进而对基础预测模型产生的预测误差进行修正;分别提取不同的特征数据对机理模型和XGBoost模型进行训练及预测。最后,通过非参数核密度估计建立预测误差概率密度函数,并在一定置信水平下预测光伏功率的波动范围。所提方法的有效性已通过光伏电站的实际数据及对比试验得到验证。With the increasing scale of distributed photovoltaic power generation,accurate prediction of photovoltaic power generation is very important for the safe and stable operation of power systems.In order to improve the accuracy of photovoltaic power prediction,an integrated model based on mechanism model and extreme gradient boosting algorithm is proposed for short-term distributed photovoltaic power probability interval prediction.Firstly,combined with the meteorological data,a density-based spatial clustering of applications with noise algorithm is used to design a photovoltaic power data governance method to filter out abnormal data in historical data.Then,the integrated model is constructed based on the optimized samples after screening.Specifically,the basic photovoltaic power prediction model is constructed based on the mechanism model for preliminary prediction,and the prediction results and other environmental data are used as input variables of the XGBoost model,and then the prediction error generated by the basic prediction model is corrected.Different feature data are extracted to train the mechanism model and XGBoost model and forecast respectively.Finally,the prediction error probability density function is established by non-parametric kernel density estimation,and the fluctuation range of photovoltaic power is predicted at a certain confidence level.The accuracy and effectiveness of the method are verified by the actual data of photovoltaic power station.

关 键 词：光伏发电预测 集成模型 XGBoost 概率区间预测 

分 类 号：TM615[电气工程—电力系统及自动化]