Volume 65

Performance of Fully Fault Adaptive Model Predictive Control for Chilled Water Plants under Fault Conditions: A Preliminary Study Huaiyu Yang, Guanjing Lin, Wanpeng Zhang

Abstract

In recent years, model predictive control (MPC) has been widely studied in chilled water (CHW) plants. However, most existing MPC approaches rely on fixed optimization models that fail to account for fault dynamics, leading to significant degradation in performance when faults arise. To address this gap, this preliminary study investigates the fault tolerant performance of fully fault adaptive model predictive control (FFAMPC) for a CHW plant in two fault scenarios – chilled water supply temperature sensor bias and supply air temperature sensor bias. The results demonstrate that for CHW supply temperature sensors, the proposed FFAMPC effectively mitigates negative biases and small positive biases. But large positive biases (e.g., +4°C), lead to a maximum 16.8% increase in energy consumption, attributed to the narrow parameter search range during the optimization process. For air-side supply air temperature sensor biases, both energy and safety performance of the proposed FFAMPC are affected: while positive biases maintain operational safety, cycle energy consumption increases by 38.7% and 73.8% under +2°C and +4°C biases, respectively. Under negative biases, unsafe operating hours account for 100% (168/168) of the operational period, as proposed FFAMPC fails to identify and mitigate supply air temperature sensor bias. In summary, the proposed FFAMPC maintains optimal performance under certain fault conditions, but its performance is still compromised under other fault scenarios. The study findings provide valuable insights for the development and refinement of fault tolerant control strategies for CHW plants in future research.

Keywords Fully fault adaptive, model predict control, chilled water plant, energy and safety