基于LGWO-XGBoost-LightGBM-GRU的短期电力负荷预测算法  

Short-term Power Load Forecasting Algorithm Based on LGWO-XGBoost-LightGBM-GRU

作　　者：王海文 谭爱国 彭赛 黄佳欣怡 田相鹏 廖红华 柳俊 WANG Haiwen;TAN Aiguo;PENG Sai;HUANG Jiaxinyi;TIAN Xiangpeng;LIAO Honghua;LIU Jun(College of Intelligent Systems Science and Engineering,Hubei Minzu University,Enshi 445000,China;Urban Power Supply Center,State Grid Enshi Power Supply Company,Enshi 445000,China)

机构地区：[1]湖北民族大学智能科学与工程学院,湖北恩施445000 [2]国网恩施供电公司城区供电中心,湖北恩施445000

出　　处：《湖北民族大学学报(自然科学版)》2025年第1期73-79,共7页Journal of Hubei Minzu University:Natural Science Edition

基　　金：国家自然科学基金项目(62163013)。

摘　　要：针对历史负荷特征提取困难所导致的短期电力负荷预测精度不高的问题,提出了基于堆叠泛化集成思想的逻辑斯谛灰狼优化-极限梯度提升-轻量级梯度提升机-门控循环单元(logistic grey wolf optimizer-extreme gradient boosting-light gradient boosting machine-gated recurrent unit, LGWO-XGBoost-LightGBM-GRU)的短期电力负荷预测算法。该算法首先使用逻辑斯谛映射对灰狼优化(grey wolf optimizer, GWO)算法进行改进得到LGWO算法,接着使用LGWO算法分别对XGBoost、LightGBM、GRU算法进行参数寻优,然后使用XGBoost、LightGBM算法对数据的不同特征进行初步提炼,最后将提炼的特征合并到历史负荷数据集中作为输入,并使用GRU进行最终的负荷预测,得到预测结果。以某工业园区的负荷预测为例进行验证,结果表明,该算法与最小二乘支持向量机(least squares support vector machines, LS-SVM)算法相比,均方根误差降低了68.85%,平均绝对误差降低了69.57%,平均绝对百分比误差降低了69.97%,决定系数提高了8.42%。该算法提高了短期电力负荷预测的精度。To address the problem of low precision of short-term power load forecasting caused by the difficulty of historical load feature extraction,a logistic grey wolf optimizer-extreme gradient boosting-light gradient boosting machine-gated recurrent unit(LGWO-XGBoost-LightGBM-GRU)short-term power load forecasting algorithm was proposed based on the idea of stacked generalisation integration.The grey wolf optimizer(GWO)algorithm was enhanced through the application of the logistic map,resulting in the LGWO algorithm.Subsequently,the LGWO algorithm was employed to fine-tune the parameters of XGBoost,LightGBM and GRU algorithm.The XGBoost and LightGBM were utilized to extract distinct features from the dataset.These features were then integrated into the historical load dataset as input for further analysis.The GRU was leveraged for the final load forecasting,generating prediction results.The efficacy of the algorithm was validated through load forecasting in an industrial park.The results showed that,in comparison to the least squares support vector machines(LS-SVM)algorithm,the proposed algorithm decreased the root mean squared error by 68.85%,the mean absolute error by 69.57%,and the mean absolute percentage error by 69.97%,and improved the coefficient of determination by 8.42%.The proposed algorithm significantly enhanced the precision of short-term electricity load forecasting.

关 键 词：短期负荷预测 集成学习 灰狼算法 极限梯度提升 轻量级梯度提升机 门控循环单元 

分 类 号：TM715[电气工程—电力系统及自动化]