Data-driven technologies for intelligent building energy management and automation have been developed under the energy balance state of systems. The datasets required are typically derived using energy and mass balance principles at each system boundary. Hence, the energy balance of systems significantly affects the system and building performances. However, sensor errors in a building energy sensor network can hinder the identification of balance states, thereby causing energy balance calculation errors in operational systems. In this study, the cooling capacity balance errors between the air and refrigerant sides of two unitary air-conditioner systems, caused by measurement errors, are investigated experimentally. Virtual in-situ sensor calibration (VIC) is performed to determine hidden sensor errors that cause the cooling capacity balance error, as well as to improve the balance and the degraded system performance analysis by calibrating the measurement errors. Experiments reveal the extent of cooling capacity balance errors occurring between the air and refrigerant sides under different operational conditions (0.5%–14.8%). VIC reveals erroneous measurements and the manner by which energy balance errors with measurement errors are solved (below 2%). Finally, the system performance analysis of the two test units shows that the VIC-driven calibration datasets improve the accuracy (0.3%–1.8% errors) of the system energy efficiency ratio even for new test cases after VIC is applied. The primary contributions of this study are as follows: (1) VIC is applied to solve energy balance calculation errors; (2) a strategy is proposed to overcome insufficient datasets to ensure energy balance improvements; and (3) calibrated measurements in real systems are validated.
Read full abstract