Implementation and testing of a fault detection software tool for improving control system performance in a large commercial building

This paper describes a model-based, feedforward control scheme that can improve control performance over traditional PID feedback control and also detect faults in the controlled process. We present results from the first phase of tests being carried out in a large commercial building to evaluate the fault detection capability of the control scheme. The control scheme uses static simulation models of the system under control to generate feed-forward control action, which acts as a reference of correct operation. Faults that occur in the system under control cause discrepancies between the feedforward models and the controlled process. By monitoring the level of these discrepancies, faults can be detected in the controlled process. The paper presents results and discusses recent experiences from implementing the control scheme in a real building using the BACnet communication protocol.