基于特征选择和机器学习的台区线损计算方法 刘度度.pdf

: .
电测与仪表 8 000 余个台区的历
史数据验证方法的有效性。实验结果表明，所提方法的 MSE 和 MAPE 可分别达到 和 %，对比
现有相关研究方法具有良好的计算精度。
关键词：线损计算；特征选择；机器学习；台区；LightGBM
中图分类号：TM727 文献标识码：B

A line loss calculation method based on feature
selection and machine learning algorithm

Liu Dudu1, Ren Lang1, Xiao Kun1, Wu Bangfei1, Yan Zhongzong2, Wen He2
(1. Zhangjiajie Power Supply Branch, State Grid Hunan Electric Power Company, Zhangjiajie 427000,
Hunan, China. 2. School of Electrical and Information Engineering, Hunan University, Changsha
410082, China)
Abstract: The power grid loss reduction is an important technical measure for energy conservation and emission
reduction. And line loss rate calculation is an important way for electric utilities to formulate loss reduction targets
and forecast the carbon emission. The existing research on line loss calculation mainly focuses on the construction
of the model, ignoring the issues of feature analysis. To this end, this paper proposes a line loss calculation method
in the station area based on the LightGBM. The selection of electrical features is analyzed, and then, the model
input is established by exploring the distribution of electrical feature index and its correlation with the line loss rate.
According to the results of feature engineering, a line loss rate calculation model is established based on the
LightGBM, and the influence of different model parameters and electrical features on model calculation results is
revealed. The validity of the propose