一种分布式能耗态势感知与预测方法研究  

Research on a distributed energy consumption situation awareness and prediction method

作　　者：雷培迪 于少中 许文芳 黎万 LEI Pei-di;YU Shao-zhong;XU Wen-fang;LI Wan(China Mobile Group Design Institute Co.,Ltd.,Beijing 100080,China)

机构地区：[1]中国移动通信集团设计院有限公司,北京100080

出　　处：《电信工程技术与标准化》2025年第4期45-51,共7页Telecom Engineering Technics and Standardization

摘　　要：在“3060双碳”战略与新型电力系统建设目标下,为应对工业能耗增加对节能减排目标的挑战,本文提出了一种基于XGBoost的分布式能耗资源态势感知与预测系统。该系统首先通过特征工程优化输入数据,使用XGBoost算法进行特征选择,剔除冗余特征以获得最佳输入组合;接着,采用贝叶斯优化方法自动调整模型的超参数,以提高预测准确性;最后与经典岭回归模型进行对比。实验结果显示,本文提出的XGBoost模型在交叉验证上的MAPE和测试集上的RMSE分别为22.994 8%和9.805 3 kWh,均优于岭回归模型,表明本文模型具有更高的预测精度和更强的泛化能力,为能源系统的源网协同建设提供了有效的解决方案。Under the"3060"dual-carbon strategy and the construction goals of a new-type power system,this paper proposes an XGBoost-based situational awareness and prediction system for distributed energy consumption resources to address the challenges of increased industrial energy consumption on energy-saving and emission-reduction targets.The system first optimizes input data through feature engineering and applies XGBoost for feature selection,eliminating redundant features to obtain the optimal input combination.Bayesian optimization is then used to automatically adjust the model's hyperparameters,enhancing predictive accuracy.Finally,the model’s performance is compared to that of a classical ridge regression model.Experimental results show that the proposed XGBoost model achieves RMSE values of 22.9948%and 9.8053kWh on cross-validation and test sets,respectively,both outperforming the ridge regression model.This indicates that the proposed model has higher predictive accuracy and stronger generalization capabilities,providing an effective solution for source-grid synergy in energy systems.

关 键 词：需求侧态势感知 XGBoost 能耗数据预测 

分 类 号：TN86[电子电信—信息与通信工程]