Balancing Voltage Stability and Communication Cost in Interconnected DERs: A Double Deep Q-Network Approach to Data Traffic Scheduling

In order to increase the reliability and consistency of distributed generation systems (DGSs), deep reinforcement learning (DRL) is frequently used in energy and communication systems. Previous studies presented an awareness method for data transmission in interconnected large-scale distributed energy resources (DERs) based on the delay-tolerant Kalman filter (DTKF). However, it is not practical to use an excessive amount of communication resources for data transmission, and offline scheduling strategies cannot be dynamically adapted to the state of the system. This study presents a methodology for evaluating scheduling strategies and modeling a double deep Q-network (DDQN) that balances both voltage stability and communication costs. Simulation results demonstrate the effectiveness of the proposed methodology in dynamically stabilizing the voltage with reasonable communication cost.