Model Predictive Control and Fault Detection and Diagnostics of a Building Heating, Ventilation, and Air Conditioning System

The paper presents the development and application of Model Predictive Control (MPC) and Fault Detection and Diagnostics (FDD) technologies to a large-scale HVAC system, their on-line implementation, and results from several demonstrations. The two technologies are executed at the supervisory level in a hierarchical control architecture as extensions of a baseline Building Management System (BMS). The MPC algorithm generates optimal set points, which minimize energy consumption, for the HVAC actuator loops while meeting equipment operational constraints and occupant thermal-comfort constraints. The MPC algorithm is implemented using a new computational toolbox, the Berkeley Library for Optimization Modeling (BLOM), which generates automatically an efficient optimization formulation directly from a simulation model. The FDD algorithm uses heterogeneous sensor data to detect and classify in real-time potential faults of the HVAC actuators. The performance and limitations of FDD and MPC algorithms are illustrated and discussed based on measurement data recorded from multiple tests. 1. INTRODUCTION The large potential economic impact of advanced technologies underlying modern Building Management Systems (BMS) have led to increased efforts focused on developing, designing, and implementing model-based control and diagnostics technologies for building HVAC systems with the objective to estimate their cost effectiveness. The potential economic impact is apparent both from the high energy-consumption levels of building HVAC systems, estimated currently at 27% (EPA, 2008), and from limitations of existing control technologies for HVAC systems. Model-based paradigms have been employed to integrate in a direct and systematic way sensor data from multiple