Microgrid Economic Dispatch for Building Networks using Learning Based Technique

Building networks create unique challenges in the domain of economic power dispatch with efficient control and scheduling of renewable and non-renewable local resources in conjunction with supply from the main grid. This work involves an intelligent microgrid control scheme that employs building thermal models for computing the approximate power requirement over bounded horizons and accordingly tunes local generation resources while learning the uncertainties in building models with learning based techniques. This learning in the loop mode of building load scheduling alleviates the requirement of accurate building thermal models or large data set of building power consumption to learn such models. We validate the effectiveness of our methodology by simulating a Model Predictive Control (MPC) based microgrid control scheme with a data-driven building thermal model and the actual building load.