一种基于数据驱动递归AdaBoost的三相不平衡配电系统拓扑重构方法研究  

A Data-driven Recursive AdaBoost-Based Topology Reconfiguration Method forThree-phase Unbalanced Distribution Systems

作　　者：蔡木良 范瑞祥 王皓天 崔明建 李赢正 程宏波 CAI Muliang;FAN Ruixiang;WANG Haotian;CUI Mingjian;LI Yingzheng;CHENG Hongbo(State Grid Jiangxi Electric Power Research Institute,Nanchang 330096,China;State Grid Jiangxi Electric Power Company Limited Ganzhou Power Supply Branch,Ganzhou 342700,China;School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China;School of Electrical and Automation Engineering,East China Jiaotong University,Nanchang 330013,China)

机构地区：[1]国网江西省电力有限公司电力科学研究院,江西南昌330096 [2]国网江西省电力有限公司赣州供电分公司,江西赣州342700 [3]天津大学电气自动化与信息工程学院,天津300072 [4]华东交通大学电气与自动化工程学院,江西南昌330013

出　　处：《智慧电力》2025年第4期38-44,共7页Smart Power

基　　金：国家自然科学基金资助项目(52207130);江西省重点研发计划项目(20223BBE51013)。

摘　　要：拓扑结构的有效重构可以显著提高电网的弹性和可靠性,然而,现有的系统拓扑重构模型计算能力仍不足以求解大量变量。为此,提出了一种基于机器学习的递归自适应增强(AdaBoost)方法,用以增强不平衡配电系统的拓扑重构。首先,准备标签作为所提递归AdaBoost方法的输出,使用经典的跨越树搜索算法进行拓扑重构。其次,收集递归AdaBoost的输入进行训练,通过应对非高斯噪声来部署配电系统状态估计(DSSE)模型的迭代过程,并通过非线性最小二乘法(NLS)对DSSE进行求解。最后,通过所提递归AdaBoost方法计算出最大后验概率,说明了最优拓扑重构策略。以改进IEEE 123节点配电系统作为算例对所提方法的有效性进行了验证。Effective topology reconfiguration can significantly enhance the resilience and reliability of power grids.However,existing topology reconfiguration models still lack sufficient computational capability to handle large-scale variables.To address this,this study proposes a machine learning-based recursive adaptive boosting(AdaBoost)method to improve the topology reconfiguration in unbalanced distribution systems.Firstly,training labels are prepared as outputs for the proposed recursive AdaBoost method,and a classical spanning tree search algorithm is employed for the topology reconfiguration.Secondly,inputs for the recursive AdaBoost are collected and trained by deploying an iterative process of the distribution system state estimation(DSSE)model to address non-Gaussian noise,where DSSE is solved via nonlinear least squares(NLS).Finally,the optimal topology reconfiguration strategy is determined by calculating the maximum posterior probability using the proposed recursive AdaBoost method.The effectiveness of the proposed method is validated through case studies on a modified IEEE 123-node distribution system.

关 键 词：自适应增强 配电系统 机器学习 拓扑重构 

分 类 号：TM73[电气工程—电力系统及自动化]