基于极限梯度提升树的输电线路缺陷风险预报

Prediction of Transmission Line Defects Risk Based on Extreme Gradient Boosting Tree

输电线路是电力系统远距离输送电能的重要设备。由于输电线路长期暴露在野外环境下,极易受暴雨、台风等极端灾害性天气影响,设备容易发生缺陷和安全隐患,危害电力系统安全可靠运行。针对灾害性天气输电线路缺陷风险难以预报问题,基于气象因子和输电线路缺陷之间的关联关系,采用极限梯度提升树XGBoost（e Xtreme Gradient Boosting）集成学习框架,构建了输电线路缺陷风险预报模型。根据气象因子的变化,预报电力系统输电线路缺陷类别和缺陷风险发生概率。以某市电力系统气象数据和缺陷数据为应用背景进行验证,实验结果表明基于XGBoost的输电线路缺陷风险模型能够根据天气预报信息快速、准确地判断输电线路缺陷类别,预报缺陷发生的概率。

Transmission lines are very important equipment for the long-distance transmission of electrical energy. As the transmission lines exposure chronically to the wild environment, it is highly vulnerable to heavy rain, typhoons and other extreme catastrophic weather conditions, which leads to power system equipment defects even endanger safe and reliable operation of power system. Due to the difficulty in predicting the risk of defective weather transmission lines, a risk forecasting model of meteorological factors and transmission line defects is built based on the relationship between meteorological factors and transmission line defects using the ensemble learning framework of the extreme gradient boosting tree（called XGBoost）. It can predict the defect state and probability of the transmission line of power system according to the change of meteorological factors. The meteorological data and defect data of a city power system are validated for the application background. The experimental results show that the risk model based on XGBoost can identify the defect category of the transmission line quickly and accurately, and predict the probability of the occurrence of the defect.