配电网拓扑结构概念聚类及其在优化规划中的应用

聚类分析是一种无监督的机器学习方法。被广泛应用于各研究领域。在城市配电网优化规划的研究中，现有的关于网络拓扑结构分析的一些方法并不适用于配电网优化规划工作，而关于配电网拓扑结构聚类分析的研究更是鲜见报道。基于对配电网结构、运行特点以及优化规划工作实际需要的认知，提出了一种结合模糊逻辑的配电网拓扑结构概念聚类方法，对于推进配电网优化规划问题的研究具有广泛的实际意义。为检验该方法的有效性，在该方法的基础上引入几个简单的概念，构成一个基本的机器学习模块，该模块可以方便地“嵌入”基于Agent行为和范例学习的新型遗传算法中，以提高原算法的计算性能。并用算例证明了在引入基于配电网拓扑结构概念聚类的机器学习模块后，新型遗传算法具有更高的计算效率和求解质量。

CONCEPTUAL CLUSTERING OF DISTRIBUTION NETWORK TOPOLOGY AND ITS APPLICATION TO OPTIMAL PLANNING

Clustering analysis is a non supervised machine learning method and it has been applied in many fields. All current identification methods of network topology are not suitable to power distribution system optimal planning. Up to now, there is no report on network topology clustering analysis of power distribution system. Based on the cognition on the distribution system structure and operation as well as the requirement of distribution power system optimal planning, this paper presents a conceptual clustering method of distribution network topology that is combined with fuzzy logic reasoning. This method shows broadly practical significance in the study of distribution power system optimal planning. To prove the validity of the conceptual clustering method, a basal machine learning module based on the conceptual clustering method of distribution network topology is presented. This module can be embedded into the new style genetic algorithm suggested in reference expediently to enhance the computation capability of the algorithm. Finally, a numerical example proves that after embedding the machine learning module, the new style genetic algorithm has higher efficiency of optimal computation and higher quality of optimal solution. .