Data-Model Hybrid Driven Topology Identification Framework for Distribution Networks

Extensive penetration of distribution energy resources (DERs) brings increasing uncertainties to distribution networks. Accurate topology identification is a critical basis to guarantee robust distribution network operation. Many algorithms that estimate distribution network topology have already been employed. Unfortunately, most are based on data-driven alone method and are hard to deal with ever-changing distribution network physical structures. Under these backgrounds, this paper proposes a data-model hybrid driven topology identification scheme for distribution networks. First, a data-driven method based on a deep belief network (DBN) and random forest (RF) algorithm is used to realize the distribution network topology rough identification. Then, the rough identification results in the previous step are used to make a model of distribution network topology. The model transforms the topology identification problem into a mixed integer programming problem to correct the rough topology further. Performance of the proposed method is verified in an IEEE 33-bus test system and modified 292-bus system.