Deep learning-aided joint DG-substation siting and sizing in distribution network stochastic expansion planning

Abstract

The rapid growth of distributed generation (DG) and load has highlighted the necessity of optimizing their ways of integration, as their siting and sizing significantly impact distribution networks. However, little attention has been paid to the siting and sizing of new substations which are to be installed. This paper proposes deep learning-aided joint DG-substation siting and sizing in distribution network stochastic expansion planning. First, as the model depends on an accurate forecast, Long Short-Term Memory (LSTM) deep neural network is used to forecast DG output and load, where electricity growth rate, bidding capacity of the electric expansion, and industrial difference are all considered. Then, a two-stage stochastic mixed integer bilinear programming model was established for joint DG-substation siting and sizing under uncertainties, where multiple objective functions are comprehensively addressed. By using the Fortuny-Amat McCarl Linearization, the resultant bilinear model is equivalently transformed into a mixed integer linear program, which can be efficiently solved. Finally, stochastic power flow calculation in the IEEE 69-node system is conducted to analyze the influence of electric expansion and DG integration on the node voltage and power flow distribution of the power system. The effectiveness of the proposed method is also verified by simulation tests.