黄丹,周永刚,刘单华,梁晓伟,吴轲,李磊.基于AdaBoost集成学习算法的低压台区理论线损计算与分析[J].电测与仪表,2025,62(11):120-128. HUANG Dan,ZHOU Yonggang,LIU Shanhua,LIANG Xiaowei,WU Ke,LI Lei.Calculation and analysis of theoretical line loss in low voltage station area based on AdaBoost integrated learning[J].Electrical Measurement & Instrumentation,2025,62(11):120-128.
基于AdaBoost集成学习算法的低压台区理论线损计算与分析
Calculation and analysis of theoretical line loss in low voltage station area based on AdaBoost integrated learning
针对传统方法在计算低压台区理论线损率时出现的数据质量要求高、物理结构依赖度高等问题，文章提出一种基于自适应增强(adaptive boosting, AdaBoost)集成学习的低压台区理论线损计算方法。综合考虑低压台区网架结构参数、台区运行方式、用电负荷水平等各类因素，形成影响因素特征指标体系，进一步通过AdaBoost算法训练各分类与回归树(classification and regression tree, CART)决策树个体学习器并计算各个体学习器的权重系数，利用个体学习器的线性组合集成得到最终的理论线损计算法模型，同时在此基础上分析研究了典型重要因素对理论线损的影响规律。结果表明：所提算法具有良好的可靠性与稳定性，可以实现不同因素对台区理论线损影响规律的分析，进一步可以支撑不同因素影响程度的横向比较，可为台区的线损异常治理提供方向性指导。
英文摘要:
In order to solve the problems of high data quality and high dependence on physical structure when calculating theoretical line loss rate of low voltage station area by traditional methods, this paper proposed a theoretical line loss calculation method based on AdaBoost integrated learning. The characteristic index system of influencing factors was formed by comprehensively considering various factors such as grid structure parameters, platform operation mode and power load level of low voltage station area. The AdaBoost algorithm was further used to train individual learner of classification and regression tree decision tree and calculate the weight coefficient of each learner. The final theoretical line loss calculation model was obtained by integrating the linear combination of individual learner. On the basis of model, the influence law of typical important factors on theoretical line loss was analyzed. The results show that the algorithm has good reliability and stability, which can realize the analysis of the influence law of different factors on the theoretical line loss. By comparing the influence degree of different factors, the directional guidance for the governance of abnormal line loss can be provided.
