Cooperative Markov Decision Process for Voltage Control of Grid-Tied Photovoltaic Generators With Reactive Power Support

The problemof voltage control (VC) by means of inverter reactive support in distribution grids with photovoltaic (PV) generation is addressed. A novel decentralized approach for VC is proposed, which consists in heuristic specification of a distributed reactive support cooperative partially observable Markov decision process (DRS-C-POMDP) whose agents control groups of PV nodes clustered according to voltage correlation. The proposed method employs offline policy optimization, which yields closed-form reactive power setpoint control laws. Hence, the online processing time required by most VC methods is avoided and controller implementation is simplified. The method is validated via simulation on a 123-bus feeder, in which it is compared to other decentralized heuristic methods from the literature. The results indicate that, for identical local inverter controls, the proposed method yields superior performance.