Markov Processes for the Management of a Microgrid

In this work, a stochastic modelization of wind and photovoltaic power productions is coupled with a two-level optimization in which the operating costs of a hybrid and isolated micro-grid are minimized. First, the power demand is modeled and predicted using an auto-regressive moving average model (ARMA), and the renewable productions are modeled using Markov reward processes. Then, the optimization problem is solved through a stochastic unit commitment and an economic dispatch. The results show that the stochastic models correctly capture the behavior of renewable sources in all system configurations proposed in the different scenarios. Furthermore, the different impacts caused by wind and photovoltaic sources and battery energy storage system on operating costs are also highlighted, which is more punctual for the first and more regular and smoother for the second.