Micro-grid energy management: a computational approach based on simulation and approximate discrete abstraction

We address the problem of designing an energy management system for a small scale micro-grid comprising a chiller plant and a cooling load representing some zone. The energy management task involves distributing the cooling power request among the chillers constituting the chiller plant, and modulating the temperature set-point of the zone so as to save energy while preserving comfort. The problem can be decoupled into a static optimization problem and a dynamic programming problem for a discrete time stochastic hybrid system. The latter one is here addressed by abstracting the stochastic hybrid model to a (finite) controlled Markov chain, where costs associated to transitions are computed by simulating the original model and determining the corresponding energy consumption. A numerical example shows the efficacy of the approach.