基于RF-CPSO-LSSVM的日线损率置信区间预测研究Confidence Interval Forecasting of Daily Line Loss Rate Based on RF-CPSO-LSSVM
郁家麟,顾韬,沈浚,周金飞,汪东耀,徐晓丁
YU Jialin,GU Tao,SHEN Jun,ZHOU Jinfei,WANG Dongyao,XU Xiaoding
摘要(Abstract)：
针对当前台区拓扑数据不全所造成的传统理论线损计算方法不适用的现状,提出了基于RF(随机森林)-CPSO(混沌粒子群)-LSSVM(最小二乘支持向量机)算法的日线损率置信区间预测研究方法。采用RF算法对造成线损的特征按重要性进行排序,计算各特征的累计贡献率并进行筛选;利用CPSO算法对LSSVM算法的惩罚因子C, g进行参数寻优以获得最佳预测模型,选取95%置信度下的理论线损置信区间作为预测结果。以浙江某市台区线损数据进行验证,结果表明,RF-CPSO-LSSVM算法比传统LSSVM, PSO-LSSVM和APSO-LSSVM算法预测日线损率置信区间结果更加精确,改进效果明显。
Given the inapplicability of the traditional theoretical line loss calculation method caused by the incomplete topological data of the current station area, a confidence interval prediction study method of daily line loss rate based on RF(random forest)-CPSO(chaos particle swamp optimization)-LSSVM(least square support vector machine) algorithm is proposed. Firstly, the RF algorithm is used to rank the importance of the features causing line loss, and the cumulative contribution rate of each feature is calculated for screening.CPSO algorithm is used for optimizing penalty factor C and g of LSSVM to obtain the best prediction model;confidence interval of theoretical line loss under the confidence level of 95% is selected for forecasting and line loss data of a station area in Zhejiang is verified. The result shows that the RF-CPSO-LSSVM algorithm is more accurate and advanced in daily line loss rate forecasting in contrast to LSSVM, PSO-LSSVM and APSO-LSSVM algorithm.
关键词(KeyWords)：
日线损率;置信区间;LSSVM算法
daily line loss rate;confidence interval;LSSVM algorithm
基金项目(Foundation): 国网浙江省电力有限公司科技项目(5211JX170007)
作者(Author):
郁家麟,顾韬,沈浚,周金飞,汪东耀,徐晓丁
YU Jialin,GU Tao,SHEN Jun,ZHOU Jinfei,WANG Dongyao,XU Xiaoding
DOI: 10.19585/j.zjdl.202007011
参考文献(References)：
- [1]晏荣煜，李向阳，高秉强.国家电网公司“互联网+”下的信通支撑架构和运营模式研究[J].电力信息与通信技术，2018(2):11-12.
- [2]朱永利，李莉，宋亚奇，等.ODPS平台下的电力设备监测大数据存储与并行处理方法[J].电工技术学报，2017,31(9):14-16.
- [3]杨柳，罗璇，肖宝辉，等.配电网线损在线监测系统与计算分析研究[J].自动化与仪器仪表，2019,231(1):31-34.
- [4]张勤，马玮，付锦，等.中压配电网规划态线损计算方法[J].电气技术，2018,18(12):57-63.
- [5]吉杨，郑华，段云琦，等.基于积分电流法的配电网理论线损计算[J].东北电力技术，2017,37(9):38-42.
- [6]邹云峰，梅飞，李悦，等.基于数据挖掘技术的台区合理线损预测模型研究[J].电力需求侧管理，2015,16(4):25-29.
- [7]廖永锋.基于复杂网络理论的电网结构脆弱性评估指标研究[D].成都：电子科技大学，2015.
- [8]HARTIGAN J A,WONG M A.Algorithm AS 136:A K-means clustering algorithm[J].Journal of the Royal Statistical Society,1979,28(1):100-108.
- [9]张贵畅.配电网线损分析和降损策略的研究[D].兰州：兰州理工大学，2018.
- [10]AHMADIZAR F,SOLTANIAN K,AKHLAGHIANTAB F,et al.Artificial neural network development by means of a novel combination of grammatical evolution and genetic algorithm[J].Engineering Applications of Artificial Intelligence,2015(39):1-13.
- [11]陈洋洋.配电网理论线损智能计算方法及应用研究[D].广州：华南理工大学，2016.
- [12]彭宇文，刘克文.基于改进核心向量机的配电网理论线损计算方法[J].中国电机工程学报，2011,31(34):120-126.
- [13]吴鸿亮，门锟，董楠，等.基于J2EE架构的线损理论计算与诊断分析系统开发与应用[J].电力系统保护与控制，2016(5):108-116.
- [14]刘健，段璟靓.配电网极限线损分析及降损措施优化[J].电力系统保护与控制，2013(12):27-35.
- [15]BODON F.A fast APRIORI implementation[C]//Proceedings of the IEEE Icdm Workshop on Frequent Itemset Mining Implementations,2003(3):70-71.
- [16]陈海涛，周鑫，雷才嘉.基于激光点云的中压配电网日同期线损检测系统设计[J].电网与清洁能源，2019,35(10):28-33.
- [17]明东岳，李振华，李俊，等.配网线损同期化计算与多维分析系统[J].电网与清洁能源，2019,35(4):30-36.
- [18]刘东升，代盛国，商学斌，等.基于压缩感知理论的缺失数据集下线损预测模型[J].广东电力，2019,32(2):80-86.