A Method for Distribution Network Topology Identification Based on Mutual Information and DNN

Identifying the topology of distribution networks plays a crucial role in maintaining the stability and security of power system operations. The current research presents a technique for discerning the topology of distribution networks, relying on data gathered by smart meters. This technique incorporates mutual information and utilizes deep neural network (DNN) models to facilitate the identification process. The measured data provided by smart meters serve as the original feature set. Mutual information theory is employed to calculate the contribution of each feature to topology identification, thereby obtaining the optimal subset of features. Based on this, a DNN-based model for distribution network topology identification is constructed. The model uses the time-section measurements of some nodes in the system and topology labels to build a deep neural network, which automatically mines the intrinsic representation and internal relationships of the input features, understanding the correlation between measurement data and the underlying topological configuration. Finally, the validation of this approach's efficacy is substantiated by its application to the IEEE 33-node distribution network.