基于R语言的负荷预测ARIMA模型并行化研究  

Research on ARIMA Model Parallelization in Load Prediction Based on R Language

作　　者：麦鸿坤 肖坚红[2] 吴熙辰 陈驰[3] 

机构地区：[1]中国科学技术大学计算机科学与技术学院,安徽省合肥市230026 [2]国网安徽省电力公司,安徽省合肥市230022 [3]河海大学能源与电气学院,江苏省南京市210098

出　　处：《电网技术》2015年第11期3216-3220,共5页Power System Technology

基　　金：国家电网公司科技项目(521200135 1TR)~~

摘　　要：自回归求和移动平均(autoregressive integrated moving average,ARIMA)模型常在R语言环境下被用于电力负荷数据的分析和预测。然而,面对海量数据背景下的工程应用,R环境下ARIMA模型的运行效率无法达到令人满意的程度。针对此问题,通过JAVA与R的实时通信,充分利用JAVA丰富的开源资源与R强大的统计计算功能,在JAVA中进行程序的逻辑判断,在R中进行数值计算,采用混合编程,最后完成ARIMA模型接口的封装,实现了基于负荷数据预测的ARIMA模型的串行化程序。在串行程序完成的基础上,根据电力负荷特性,对数据进行划分,结合JAVA多线程技术,实现了ARIMA模型的并行化。最后,结合文中提出的复合评价指标,对安徽地区随机获取的用户电力负荷进行了测试。结果表明,ARIMA模型并行算法预测准确率好,代码执行效率高。ARIMA Model is usually used to analyze and predict power load data in R language. However, for application in context of mass data, efficiency of this model is not satisfactory. To meet demand of engineering application, this paper utilizes real-time communication between R and JAVA, making full use of abundant JAVA resources and statistical computing capacity of R language, performing logic judgment in JAVA and numerical calculation in R. This paper accomplishes interface encapsulation and realizes programmed ARIMA model. On basis of serial program, power load data is divided according to its characteristics, and ARIMA model parallelization based on multi-thread technology in Java is presented. At last, considering comprehensive evaluation indexes proposed in this paper, experimental data picked up in random from Anhui Province is tested. Experiment results show that the algorithm is accurate and efficient, and the method is valid for data prediction and project application.

关 键 词：电力负荷预测 ARIMA JAVA R语言 并行化 

分 类 号：TM71[电气工程—电力系统及自动化]