Validation of virtual sensor-assisted Bayesian inference-based in-situ sensor calibration strategy for building HVAC systems  

作　　者：Guannan Li Jiahao Xiong Shaobo Sun Jian Chen 

机构地区：[1]School of Urban Construction,Wuhan University of Science and Technology,Wuhan 430065,China [2]Department of Building Environment and Energy Engineering,The Hong Kong Polytechnic University,Hong Kong,China

出　　处：《Building Simulation》2023年第2期185-203,共19页建筑模拟（英文）

基　　金：supported by the National Natural Science Foundation of China (51906181);the 2021 Construction Technology Plan Project of Hubei Province (No.2021-83);the Excellent Young and Middle-aged Talent in Universities of Hubei Province,China (Q20181110).

摘　　要：For building heating,ventilation and air-conditioning systems(HVACs),sensor faults significantly affect the operation and control.Sensors with accurate and reliable measurements are critical for ensuring the precise indoor thermal demand.Owing to its high calibration accuracy and in-situ effectiveness,a virtual sensor(VS)-assisted Bayesian inference(VS-BI)sensor calibration strategy has been applied for HVACs.However,the application feasibility of this strategy for wider ranges of different sensor types(within-control-loop and out-of-control-loop)with various sensor bias fault amplitudes,and influencing factors that affect the practical in-situ calibration performance are still remained to be explored.Hence,to further validate its in-situ calibration performance and analyze the influencing factors,this study applied the VS-BI strategy in a HVAC system including a chiller plant with air handle unit(AHU)terminal.Three target sensors including air supply(SAT),chilled water supply(CHS)and cooling water return(CWR)temperatures are investigated using introduced sensor bias faults with eight different amplitudes of[−2℃,+2℃]with a 0.5℃ interval.Calibration performance is evaluated by considering three influencing factors:(1)performance of different data-driven VSs,(2)the influence of prior standard deviationsσon in-situ sensor calibration and(3)the influence of data quality on in-situ sensor calibration from the perspective of energy conservation and data volumes.After comparison,a long short term memory(LSTM)is adopted for VS construction with determination coefficient R-squared of 0.984.Results indicate thatσhas almost no impact on calibration accuracy of CHS but scanty impact on that of SAT and CWR.The potential of using a prior standard deviationσto improve the calibration accuracy is limited,only 8.61%on average.For system within-control-loop sensors like SAT and CHS,VS-BI obtains relatively high in-situ sensor calibration accuracy if the data quality is relatively high.

关 键 词：heating ventilation and air-conditioning(HVAC) in-situ sensor calibration Bayesian inference(BI) virtual sensor(VS) influencing factor energy conservation(EC) 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]