Dynamic Spatio-Temporal Planning Strategy of EV Charging Stations and DGs Using GCNN-Based Predicted Power Demand

As a sustainable participant in the modernization of transportation systems, electric vehicles (EVs) call for a well-planned charging infrastructure. To meet the ever-increasing charging demands of EVs, an efficient dynamic spatio-temporal allocation strategy of charging stations (CSs) is necessary. With newly allocated CSs, additional distributed generators (DGs) are required to compensate for the load increase. Given a budget to be allocated over a certain time horizon, we formulate the joint spatio-temporal CSs and DGs planning problem as a multi-objective optimization problem. During each planning period, the allocation strategy aims at minimizing the total power generation costs and CSs/DGs installation costs while satisfying budgetary and power constraints and ensuring a minimum level for the charging requests satisfaction rate. In this regard, we first predict the future power demand of EVs using a graph convolutional neural network (GCNN). Then, using the power demand forecast, we obtain the optimal number and locations of CSs and DGs at each time stage using reinforcement learning. A case study of the proposed allocation strategy over 6 time stages for the 2000-bus power grid of Texas coupled with 720 initially existing CSs is presented to illustrate the performance of the planning strategy.