Pareto-Optimized Thermal Control of Multi-Zone Buildings Using Limited Sensor Measurements

This paper presents a control-oriented building thermal model and optimization framework. Space heating is used as an illustrative example of a flexible building load, with temperature setpoints as a control input. The presented framework is applicable to practical building systems where measurements are limited by cost and installation burden. An Unscented Kalman Filter estimates parameters and disturbance inputs of a multi-zone thermal circuit. Forecast models of multiple exogenous input sources are created from disturbance proxies and estimated disturbance inputs. Zone-level controllers in the thermal circuit simulation estimate the heating system response based on forecasted exogenous thermal inputs and proposed temperature setpoint profiles. Genetic algorithm-based operations are used to find an approximate Pareto set, i.e., the best trade-offs in the objective space. The focus of this work is reducing energy usage from space heating, while maintaining or improving thermal comfort. The full framework is demonstrated using data collected from a university building. Results predict that the proposed method provides a lower energy consumption than the baseline strategy. The framework is implemented in practice in a model predictive control scheme.