Approximate dynamic programming, building-integrated solar energy systems, emulation, forecast uncertainty
This study presents an emulation framework for the optimal utilization of solar energy in buildings with integrated solar systems under uncertain solar irradiance forecast. An open plan office space is used as a test-bed, in which the energy system includes a building-integrated photovoltaic-thermal system serving as the source side of a solar-assisted heat pump, and the heat pump provides hot water to a thermal energy storage tank connected to a radiant floor heating system. To model the solar irradiance as non-Gaussian stochastic disturbance to the energy system and quantify the uncertainty, we developed a probabilistic time-series autoregressive model, using the sky-cover forecast as input. An emulator that couples physical system models in TRNSYS with a stochastic model predictive controller developed in Python and MATLAB is built to evaluate the closed-loop operation of the integrated energy system. The controller receives uncertain solar irradiance forecast, and predicts the optimal heat pump operation schedule that minimizes the energy consumption over a prediction horizon while maintaining indoor thermal comfort. A new approximate dynamic programming methodology is deployed in the controller to solve the stochastic optimal control problem, and achieve good solution quality. The results show that the proposed approach saves up to 44% of the electricity consumption for heating in a winter month, compared to a well-tuned rule-based controller, while imposing less uncertainty on thermal comfort violation.