基于Cox比例风险模型的电力电缆故障影响因素分析  

Analysis of Influencing Factors on Power Cable Failures Using Cox Proportional Hazards Model

作　　者：王航[1] 付光攀 杨斌[3] 姜伟[3] 周文俊[1] 田智[3] 唐泽洋[4] 周承科[1] 

机构地区：[1]武汉大学电气工程学院,武汉430072 [2]广州供电局有限公司,广州510220 [3]国网湖北省电力公司武汉供电公司,武汉430072 [4]国网湖北省电力公司电力科学研究院,武汉430077

出　　处：《高电压技术》2016年第8期2442-2450,共9页High Voltage Engineering

基　　金：国网湖北省电力公司科技项目(SGTYHT/14-JS-188)

摘　　要：电力电缆故障信息的深层次挖掘可提高对电缆故障影响因素的分析。因此,针对某供电公司10 k V电力电缆故障数据,运用统计学模型—Cox比例风险模型,定量分析了电缆故障影响因素,用以指导电缆采购、施工、运行和维护。为确保数据分析的准确性,提出了电缆数据预处理原则,探讨了合适的样本量大小。运用Cox比例风险模型对电缆故障影响因素进行单因素分析;运用Logistic回归模型确定了电缆故障影响因素类别,并统计计算了各电缆故障影响因素对应的电缆故障率,确定了各影响因素组成元素的相对危险程度,最终证明了Cox比例风险模型分析结果的正确性。结果表明:本体生产厂家M1、附件生产厂家N1、施工单位I3对应的电缆故障率最高分别为0.33、0.29、0.218,企业在进行电缆采购、施工、维护时应着重关注这3家单位。Analysis or data mining of power cable faults information can improve the analysis of cable failure factors. Focusing on 10 kV voltage power cables from a certain power supply company, and using a statistical model and the Cox proportional hazards model, we quantitatively analyzed factors of cable failures to guide the procurement, construction, operation and maintenance of power cables. Firstly, we presented prepossessing of cable data, which was followed by determination of the required sample size, in order to ensure the suitability of the data. Then, we analyzed each of the factors related to cable faults using Cox proportional hazard models. Finally, we used the Logistic Regression model and statis- tical calculations on the relative degree of risk of each element containing the influential factors of cable failures to prove that the analysis results of Cox proportional hazards model were correct. The results showed that cable failure rates of the body manufacturer M1, the accessory manufacturer N1, the construction company I3 were the three highest, which were 0.33, 0.29, 0.218, respectively. The enterprise on cable procurement, construction, and maintenance should focus on these three companies.

关 键 词：电力电缆 故障影响因素 样本 COX比例风险模型 LOGISTIC回归模型 电缆故障率 

分 类 号：TM75[电气工程—电力系统及自动化]