Microgrids of electricity distribution can "smartly" improve local reliability and power quality, while moderating local greenhouse gas emissions and costs of power supply by the exploitation of renewable sources and storage. In this paper, we consider a microgrid composed of a middle-size train station with integrated photovoltaic power production, a small urban-sized wind power plant and a residential district, and optimize its energy management in presence of uncertainties in the environment and mechanical failures. We use Agent-Based Modeling (ABM) and Robust Optimization (RO), and evaluate system performance in terms of typical reliability (adequacy) indicators for energy systems such as Loss of Load Expectation (LOLE) and Loss of Expected Energy (LOEE).