Economic impact of persistent sensor and actuator faults in concrete core activated office buildings

Abstract

Sensor and actuator degradation occurs frequently in office buildings. These can have a large impact on HVAC performance, both on energy use and thermal (dis)comfort. The degree of impact depends on the faulty component(s), fault type(s) and fault severity and has a significant non-linear relation with the control strategy and comfort constraints. Concrete core activated (CCA) office buildings typically have a high thermal inertia, high comfort requirements and they are equipped with low exergy and low capacity production systems. This allows the inclusion of renewables, thermal storage and flexible load shifting, but this also augments the effects of small perturbations in control output.In this paper, the economic fault impact is investigated by dynamic simulations using an emulator model of a CCA office building in combination with four different control strategies. A virtual test-bed is developed, consisting of two emulated office zones and a temperature modulated concrete core activation HVAC system, augmented with persistent faults in temperature sensor and hydronic flow rate actuators. Both the fault free (FF) performance and the fault present (FP) performance are investigated and compared through the relevant, control-associated costs using an economic framework. This methodology is able to determine the fault sensitivity of different supervisory control strategies and assists with the selection of the most economical, fault-robust controller for a certain building type. Also, the most critical sensors and actuators are identified.The evaluated faults are shown to be detrimental for the control performance. The relative economic impact of simultaneous (realistic, randomly distributed and non-correlated) sensor and actuator faults, ranged from +7% to +1000%. By adhering to an appropriate commissioning frequency, this impact can be reduced. The optimal commissioning period for sensors and actuators was determined to be between 2.8 and 5.0 years (case study, controller and assumption dependent). The lowest financial impact due to degradation faults, for this case study and assumptions, is attained by the closed loop model predictive control (CL-MPC) supervisory algorithm, which incurred only a 15% relative increase of total present cost, as opposed to increases above +100% for the other investigated control strategies over the controller lifetime.This study highlights the relevance of taking faults into account when evaluating long term HVAC control performance and quantifies the economic impact of simultaneous persistent sensor and actuator faults on control performance.