Deep Reinforcement Learning Enabled Voltage Control at Grid Edge Technologies

The installation of distributed technologies such as variable renewable energy sources, energy storage system and flexible electric vehicle charging stations at edge of the distribution grid has increased drastically. Therefore, centralized management through distribution management system faces numerous operational challenges and control issues. Furthermore, highly dependent of communication system also make the grid vulnerable to cyber threats. Hence, an artificial intelligence-assisted voltage control strategy that relays on local measurements has been introduced. To develop the local control framework, the deep reinforcement learning (DRL) based artificial neural networks has been utilized. The proposed DRL-based local control works for grid edge technologies (GET) to support the voltage profile. The proposed DRL framework for voltage management has been validated through a modified IEEE 13 bus distribution test system connected with solar photovoltaics and electric vehicle charging station at edge of the grid. The simulation results of OpenDSS model interfaced with Python for the considered test illustrate that the proposed DRL framework able to control the voltage profile of GETs under uncertainties scenarios and transient events.