Optimizing Comfort and Energy Efficiency: The Impact of Model Accuracy on Multi-Objective MPC

Abstract—Buildings are responsible for ∼30% of primary
energy consumption, mainly because of Heating, Ventilation, and
Air Conditioning (HVAC) systems. The usual ON/OFF controller
tends to react to occupancy presence, causing discomfort and
energy waste. Furthermore, these controllers usually focus on
thermal comfort and disregard other comforts, such as air
quality, visual, etc. due to their inability to handle complex multiobjective
problems. In this context, Model Predictive Controller
(MPC) presents a promising alternative for dynamic control
of HVAC systems. However, the accuracy of the building’s
thermal and air quality models greatly influences the MPC
performance. This paper proposes simple models to develop
a multi-objective MPC to minimize energy consumption while
maintaining occupancy thermal and air quality comfort. Each
models are developed using the limited data available. Since
these models largely depend on occupancy information, a further
study is conducted to analyze the impact of occupancy estimation
accuracy on MPC performance. It is found that occupancy
estimation models reach 98% of the optimum performance with
a 90% accuracy.