改进随机森林算法在用电预测中的应用  

Application of Improved Random Forest Algorithm in Electricity Consumption Forecasting

作　　者：熊洁 牛燕 刘伟 XIONG Jie;NIU Yan;LIU Wei(Ezhou Power Supply Company,State Grid Hubei Electric Power Co.,Ltd.,Ezhou 436000,China)

机构地区：[1]国网湖北省电力有限公司鄂州供电公司,湖北鄂州436000

出　　处：《自动化仪表》2023年第10期95-100,共6页Process Automation Instrumentation

摘　　要：全社会用电量对于电力企业的经营和管理具有重要作用。提高全社会用电量的预测精度,有利于合理调配电力资源,提前为“迎峰度夏”等特定用电场景作好供电准备。针对全社会用电预测难度较大的问题,提出利用K-means聚类对行业用电数据进行有效区分。根据不同类型行业的波动特点,采用季节性自回归整合滑动平均(SARIMA)和随机森林(RF)的混合模型预测出各类型行业的用电指数,以预测全社会用电量发展趋势,从而提高预测准确率。以某省2018年1月至2021年6月全社会及各行业月用电量数据作为样本数据,测算发现各行业用电波动有明显差异。研究结果显示:该模型能够对不同类型行业的用电特点进行修正,具有较好的稳定性;全社会用电量的预测结果准确,相对误差控制在2.0%以下。Electricity consumption of the whole society plays an important role in the operation and management of electric power enterprises.Improving the prediction accuracy of electricity consumption of the whole society is conducive to the rational deployment of power resources and preparing power supply in advance for specific scenarios of electricity consumption such as“peak summer”.Aiming at the problem that it is difficult to predict the electricity consumption of the whole society,it is proposed to utilize K-means clustering to effectively differentiate the industry electricity consumption data.According to the fluctuation characteristics of different types of industries,a hybrid model of seasonal auto regressive integrated moving average(SARIMA)and random forest(RF)is used to predict the electricity consumption index of each type of industry to forecast the development trend of the whole society’s electricity consumption,so as to improve the prediction accuracy.The monthly electricity consumption data of the whole society and each industry from January 2018 to June 2021 in a province is used as sample data,and the fluctuation of electricity consumption in each industry is found to be significantly different.The results of the study show that the model can correct the characteristics of electricity consumption in different types of industries and has good stability;the prediction results of the whole society’s electricity consumption are accurate,and the relative error is controlled below 2.0%.

关 键 词：用电量预测 K-MEANS聚类 混合模型 季节性自回归整合滑动平均 随机森林 

分 类 号：TH-39[机械工程]