关键词: 最大信息系数, LSTNet, 关联性特征选择, 新型台区线损, 贝叶斯优化, 特征选择算法

Abstract: Aiming at the shortcomings of traditional methods for predicting line loss in substation areas, such as inability to adjust flexibly, lack of real-time performance, and accuracy affected by meteorological conditions, a new method for predicting line loss rate in substation areas based on maximum information coefficient and BO-LSTNet is proposed. The maximum information coefficient method is used to screen meteorological information in the substation area, and input variables are filtered and cleaned, and finally the line loss data is put into the LSTNet model after Bayesian optimization. The reliability is verified through simulation experiments, and the results show that compared with previous prediction methods, this model has higher adaptability to the new type of substation line loss rate prediction, and solves the problem of low accuracy in line loss prediction in practical engineering.

Key words: maximum information coefficient, LSTNet (long-and short-term time-series network), selection of correlation features, new type of substation line loss, Bayesian optimization, feature selection algorithm