Machine learning-based model predictive control for ACMV system using edge device

Abstract

The Building Management System (BMS) can monitor and manage a facility's mechanical, electrical, and electromechanical services. One subset of this system is Air-conditioning and Mechanical Ventilation (ACMV), which regulates interior temperatures. Traditionally, PID controllers have been used in ACMV subsystems to achieve desired temperature control outcomes; however, such controllers are associated with real-time constraints that limit their effectiveness. In an effort to address these limitations inherent in PID-based systems, Model Predictive Control (MPC) has emerged as a promising alternative controller capable of surmounting various issues pertaining to tracking, accuracy, etc. Additionally, thanks to recent improvements in machine learning algorithms, it can now handle large quantities of data efficiently while employing edge devices for operational simplicity. This study thus examines the integration of MPC architecture alongside machine learning algorithms into edge-based HVAC solutions within testbed scenarios. The results will be compared against the existing baseline BMS data on various factors such as power consumption.