基于Shapley值理论的能源系统需求预测方法  

Energy Demand Forecasting Method Based on Shapley Value Theory

作　　者：李娜[1] 刘树勇[1] 曾鸣[2] 刘丽霞[1] 李源非 韩旭[2] 

机构地区：[1]国网天津市电力公司经济技术研究院,天津市300000 [2]华北电力大学经济与管理学院,北京市102206

出　　处：《电力建设》2016年第1期15-22,共8页Electric Power Construction

基　　金：国家自然科学基金项目(71271082)~~

摘　　要：合理而准确的能源消费预测对于科学制定能源规划、优化调整能源与产业结构具有重要意义。针对传统能源预测方法预测精度低、未充分计及环境政策影响的缺点,提出了基于Shapley值理论的多情景修正组合预测模型。首先,根据能耗预测的要求和特点选取3个单项预测模型,并通过博弈论Shapley值理论确定单项模型在组合模型中的权重从而得到基准预测结果;然后,量化技术进步、经济发展、政策变动3个环节为修正项和修正系数,进一步改进模型函数,得到不同情景下的修正预测结果;最后,基于T市生活能耗数据进行算例分析,结果表明所提方法能够实现预测值曲线与实际值曲线的高度拟合,在充分考虑环境政策影响的基础上提高能源预测精度,为有关部门进行能源规划提供决策依据。The reasonable and accurate prediction of energy consumption is of great significance for scientifically making energy plan and optimizing the structure of energy and industry. Aiming at the shortcomings of the traditional energy forecasting method,which has lowprediction accuracy and not be fully accounted for the influence of environmental policies,this paper presents a combined forecasting and scene correction model based on the Shapley value theory. Firstly,according to the requirements and characteristics of energy consumption forecasting,we select three single forecasting models,and determine the weight of the single model in the combined model through the Shapley value theory to obtain the basic forecasting result. Then,three main aspects of technological progress,economic development and policy change are quantified as the correction term and coefficient to further improve the model function and obtain the modified prediction results under different scenes. Finally,the case of life energy consumption in T City is studied. The results showthat the forecasted value curve and actual value curve are highly fitted in the proposed method,which can improve the accuracy of energy forecasting based on the full consideration of environmental policy influence and provide decision basis for the energy planning of related departments.

关 键 词：组合模型 能源消费预测 SHAPLEY值 情景修正 

分 类 号：F206[经济管理—国民经济] F407.61