Enhancing Energy Trading Between Different Islanded Microgrids A Reinforcement Learning Algorithm Case Study in Northern Kordofan State

Abstract

This paper tackles the problem of rural electrification and the lack of grid connection to large areas of Sudan. It introduces microgrids as an alternative to conventional centralized generation as they provide stability in electricity supply in addition to the environmental benefits accompanied with using renewable energy sources. A new method is introduced to facilitate the fluctuation in energy production when using renewable sources by creating a Reinforcement Learning algorithm to conduct the process of energy trading between different islanded microgrids. The goal of the trading process is to achieve stability and generation-load balance in the microgrids. The paper also presents a case study of three villages in North Kordufan State; Hamza Elsheikh, Tannah and Um-Bader. The study uses real solar irradiance and wind speed data to create a MATLAB simulation for a fully functional microgrid. An RL environment of the grids is created which can be used for future research and modelling in the field of smart grids. The paper explores Vanilla Policy Gradients VPG as a solution algorithm for the problem. The algorithm achieved generationload stability when applied to data extracted from the MATLAB simulation; satisfying the loads while also achieving profit from the trading process; reducing the return of investment period for the microgrid.