Cooperative secondary voltage control of static converters in a microgrid using model-free reinforcement learning

Agent-based secondary voltage regulation in an islanded MicroGrid is complicated by non-linear system dynamics, state couplings and uncertain communication network topology information. This paper proposes an off-policy learning algorithm for cooperative secondary voltage control which can synthesize an optimal feedback controller in real-time without knowledge of the system model. A simulation model has been developed using MATLAB/Simulink, which demonstrates a working controller. Results from the simulations are included, and practical considerations regarding implementation on a real system discussed.