# EXECUTIVE SUMMARY# This document reviews several
optimization methods that can be applied to control a microgrid, taking into
account both technical and economical criteria. A microgrid control system must
supply the inner energy demand, minimizing the cost and maximizing the quality
of service, considering its own energy resources – which are usually based on
renewable energy- and the grid constraints. To evaluate which method best suits
to manage a microgrid, instead of addressing the whole problem, the subproblem
of scheduling loads is considered. The first section describes the problem of
managing a microgrid, globally, and then focuses on the scheduling load problem
that turns to be an NP-hard problem. The next sections discuss a set of
optimization methodologies considered likely to solve this problem. For each
methodology, after a brief description, an explanation on how to apply the
method to the considered problem is presented. The list of methods discussed
includes: greedy search, simulated annealing algorithm, particle swarm
optimization, ant colony optimization, Monte Carlo tree search, A* search,
dynamic programming, integer linear
programming and minimum cost network flow . The last section summarizes the
advantages and disadvantages for each method and compares the performance of
most methods on a test sample. The conclusion of the study is that the
Simulated Annealing algorithm and Ant Colony optimization show the best
performance with a basic tuning process. However, with a fine tuning process,
better results might be expected with other methodologies such as Particle
Swarm Optimization, as described in literature. |