基于随机森林的220kV主变压器异常状态预警方法  

The Early Warning Method of 220V Main Transformer Abnormality Based on Random Forest

作　　者：刘锦新 LIU Jin-xin(Foshan Power Supply Bureau,Guangdong Power Network Co.Ltd.,Foshan 528000,China)

机构地区：[1]广东电网有限责任公司佛山供电局,广东佛山528000

出　　处：《电气开关》2023年第5期110-113,共4页Electric Switchgear

摘　　要：为提高对220kV主变压器异常状态预警的准确性,提出基于随机森林的220kV主变压器异常状态预警方法。该方法首先通过相应的传感器,对主变压器的运行数据进行采集。其次,为提高采集的数据质量,对其进行预处理。接着,基于处理后的数据,利用模拟退火算法确定异常状态数据特征。然后为提高后续预警的准确性,利用基尼系数和信息增益率确定最佳分裂特征。最后,基于此实现决策树的构建,根据置信水平确定异常状态,实现预警。结果表明,所提方法可有效提高主变压器异常状态预警的准确性,可达到97.4%,且具有较高的预警效率。In order to improve accaracy of the main transformer abnormality condition warning,put formard the abnormality conclition warning method of 220V main transformer.The method,first,collect the operation data of the main transformer by corresponding sensor,then in order to increase collection data quality,carrying out it pretreatment.Next,based on the treatment data later,use analog annealing algorithm to define abnormal condition data characteristics.Then,in order to inrease accuracy of follow-up warning,use Jini coefficient and information gain rate to define optimum split characteristics.Finally,based on this,achieve the stracture of the decision tree.According to the confidence level,define the abnormal condition,achieving follow-up warning.The results show the proposed method can effectively increase accuray of abnormal sondition warning of the main transformer.up to 97.4%and have higher pretreatment efficiency.

关 键 词：随机森林 主变压器 异常状态 分裂特征 

分 类 号：TM712[电气工程—电力系统及自动化]