基于XGBoost的输电线路火灾风险等级评估方法

XGBoost‑based assessment method for fire risk levels of transmission lines

山林火灾是威胁输电线路安全运行的威胁之一。首先,回顾湖南省2018—2020年的历史山火事件发生规律,在此基础上构建山火事件样本库,引入独热编码技术对线路植被类型等文本类特征进行数值化处理;然后,利用XGBoost技术构建输电线路山火风险评估模型,针对山火样本和正常运行样本的不平衡问题,基于代价敏感机制提出加权目标函数,减小样本不平衡导致的山火漏判问题;最后,通过湖南省永州市2020—2021年的山火事件进行测试,验证本文所提出的输电线路山火风险评估模型的有效性。

Mountain fires pose a threat to the safe operation of transmission lines.Firstly,the historical pattern of mountain fire incidents in Hunan province from 2018 to 2020 is reviewed.On this basis,a database of mountain fire incidents is constructed.One-hot encoding technique is introduced to numerically process textual features such as the vegetation type along the transmission lines.Then,XGBoost technology is utilized to build a mountain fire risk assessment model for transmission lines.Aiming at imbalanced samples between forest fire incidents and normal operations,based on a cost-sensitive mechanism,a weighted objective function is presented to mitigate the problem of overlooking mountain fires caused by sample imbalances.Finally,the proposed mountain fire risk assessment model for transmission lines is tested using the forest fire incidents in Yongzhou City,Hunan Province,from 2020 to 2021 to validate its effectiveness.