Supervisory Model Predictive Control of Flexible Building Loads with On-site Solar Generation

Abstract

In this work we consider the problem of reducing the grid impact and overall cost of electricity to the user of a commercial building with on-site solar generation. This is done through locally managing the generation-load balance to increase the attractiveness and feasibility of wide-scale solar power generation into the existing grid. To this end, a supervisory model predictive control (MPC) system is implemented on a simulated building with on-site solar generation, electrical vehicle charging stations, battery storage, lighting control, and a heating ventilation and air conditioning system. Various objective functions are proposed in an attempt to find a computationally reasonable solution that minimizes the grid impact and financial cost to the user. Three objective functions were selected to meet these criteria: the net building energy use, the peak building load, and a weighted sum of the prior two. These resulting MPC formulations were implemented with both perfect and imperfect forecasting of the available solar generation and environmental effects on the building. The results are compared and referenced to a rules-based case. While all MPC formulations outperformed the rules-based strategy under a full simulated month of use, the performance and choice of the MPC formulation show dependence on weather forecast accuracy. A discussion of the resulting operational costs, effects of imperfect forecasting, and broader implications of this local solar utilization on potential widespread solar penetration is presented.