Model predictive load scheduling using solar power forecasting

In this paper a model is developed to solve the on/off scheduling of (non-linear) dynamic electric loads based on predictions of the power delivery of a (standalone) solar power source. Knowledge of variations in the solar power output is used to optimally select the timing and the combinations of a set of given electric loads, where each load has a desired dynamic power profile. The optimization exploits the desired power profiles of the electric loads in terms of dynamic power ramp up/down and minimum time on/off of each load to track a finite number of load switching combinations over a moving finite prediction horizon. Subsequently, a user-specified optimization function with possible power constraints is evaluated over the finite number of combinations to allow for real-time computation of the optimal timing and switching of loads. A case study for scheduling electric on/off loads with switching dynamics and solar forecast data at UC San Diego is carried out.