This paper deals with microgrids in an islanded location where only local generator sources are deployed. However, microgrids in isolated mode are susceptible to unplanned meteorological changes. The defy is to improve the autonomy of these microgrids; thus we start first by checking the required availability of power in the microgrid after unexpected hardware conflicts, characterizing the faults in production sources by probabilistic distributions and modeling their occurrences by using a connection of Markov chains and Petri nets. We propose also a new multi-agent control strategy where agents are employed to check the availability of sources and supply power automatically to consumers in accordance with their priorities and power requirements. The whole architecture is modeled by a modeling and verification environment named ZIZO. The simulation and experimental results are based on data collected from a Tunisian petroleum platform. The availability of energy in the microgrid is increased to 99.68%, and it could be corrected up to 100%. Although this strategy can decrease the availability rate for the uncritical loads, it prevents the microgrid from any dangerous situation.INDEX TERMS Microgid in islanded mode, Markov chain, Petri net, reliability, verification.