Due to the nonlinear dynamics of direct current (DC) microgrids, the existence of input constraints, and their multi-input multi-output (MIMO) nature, classical linear controllers cannot provide an appropriate performance in a wide range of operations. In this paper, to address these issues, nonlinear suboptimal controllers are systematically developed in the primary layer of DC microgrids by employing a state-dependent Riccati equation (SDRE) methodology. To this end, the whole complexities of the nonlinear dynamics and input constraints are considered in the design procedure of the proposed SDRE controllers. After designing the controllers, and for a fast yet effective fault detection/isolation, an artificial neural network (ANN) is trained to identify the closed-loop microgrid at its nominal condition. Then, the trained ANN is employed to design a fault detection/isolation mechanism. Simulation results of the developed SDRE control scheme augmented by the ANN-based fault detection/isolation mechanism demonstrate the merits of the proposed scheme.
