Deep Neural Networks based Power Flow Calculation in Distribution System Using Clustering

Abstract

In distribution systems, complexity and uncertainty have increased due to the integration of distributed energy resources. Fast and accurate power flow calculation is required to operate the distribution system stably. A deep learning-based power flow calculation method was proposed using distribution system data. To improve the performance of the deep learning method, we propose a clustering-based deep learning model. The proposed method uses voltage profiles to group similar buses. Simulation result using 33-bus and 69-bus models shows that the proposed model outperforms the plain deep learning model in terms of accuracy and robustness to uncertainties.