The importance of model fidelity, update rate and horizon length in model based microgrid control

In order to maximize the effectiveness of microgrids sophisticated control strategies are required. In this paper we investigate the performance of a model based power flow optimization controller for a grid connected microgrid consisting of; solar PV, battery storage, thermal generation, and a local electrical load. The controller is based on a mixed integer linear program (MILP) applied in a model predictive control framework (MPC). In the MPC problem formulation there are two design parameters that need to be considered, the update rate of the controller and the number of steps considered in the on-line optimisation problem. Previous work by the authors has demonstrated that the former is critical in order to respond to the high variability associated with solar generation. In this work we extend the previous study to highlight the effect of update rate on battery utilization and the consequential effect on cumulative system operating cost. We demonstrate that the number of steps in the controller horizon has a largely inconsequential impact due to the battery being primarily utilised to ensure efficient generator operation as opposed to providing long term energy storage. Furthermore, the importance of model fidelity is investigated and a 42% decrease in operating cost over a 1 month period is observed when generator fuel consumption non-linearities are incorporated in the controller design, as opposed to a constant efficiency fuel consumption model.