规避坏数据影响的母线负荷预测新策略  

Novel strategy for bus load forecast by preventing the effect of bad data

作　　者：陈新宇[1] 康重庆[1] 陈刚[2] 程芸[2] 杨军峰[2] 

机构地区：[1]清华大学电机系电力系统国家重点实验室,北京100084 [2]国家电力调度通信中心,北京100031

出　　处：《中国电力》2009年第9期27-31,共5页Electric Power

基　　金：高等学校博士学科点专项科研基金资助课题(200800030039);国家电网公司科技资助项目

摘　　要：坏数据的存在对母线负荷预测的精度有较大影响。针对在实际工作中坏数据辨识和修补的结果不令人满意的情况,提出了一种新的规避坏数据影响的预测策略。提出了完全可信信息集的概念,在这一概念下将历史负荷进行合理划分,并分析了相互之间的横向联系和纵向联系;由此提出了以完全可信信息集为基础的预测策略,避免将修补后数据直接用于预测,妥善处理坏数据对预测效果的影响。描述了预测方法的实现过程,算例表明这一预测策略可以取得较好的效果。The existence of Bad data is a key component ＇affecting the accuracy of bus load forecasting. However, it has been a huge challenge to identify and restore bad data in practice. A novel strategy was proposed by which the impact of bad data on bus load forecasting result could be effectively mitigated. A new conception, perfect reliable information set, was introduced. Historical load data were divided into different categories, and the interrelations as well as distinctions among different categories were analyzed. On this basis, a bad data circumventing strategy in bus load forecasting was proposed to avoid direct utilization of bad data in the process of forecasting. Besides, a relative algorithm was developed to support the utilization of the strategy, and a numerical case was studied to testify its effectiveness.

关 键 词：母线负荷预测 坏数据 特征曲线 预测误差 

分 类 号：TM715[电气工程—电力系统及自动化]