Multi agent reinforcement learning based distributed optimization of solar microgrid

We consider grid connected solar microgrid system which contains a local consumers, solar photo voltaic (PV) systems, load and battery. The consumer as an agent continuously interacts with the environment and learns to take optimal actions through a model-free Reinforcement Learning algorithm, namely Q Learning. The aim of the agent is to optimally schedule the battery to increase the utility of the battery and solar photo voltaic system and thereby aims for the long term objective of reducing the power consumption from grid. Multiple agents sense the states of environment components and make collective decisions about how to respond to randomness in load and intermittent solar power by using a Multi agent reinforcement algorithm, namely Coordinated Q Learning (CQ Learning). Each agent learns to optimize individually and contribute to global optimization. Grid power consumed when solar PV system operates individually, by using Q learning is compared with operation of many such solar PV systems in a distributed environment using CQ learning and it is proved that the grid power requirement is considerably reduced in CQ learning than in Q learning. Simulation results using real numerical data are presented for a reliability test of the system.