Preventive approach to determine sensor importance and maintenance requirements Li Zhengwei, Huang Gongsheng ⁎ Department of Civil and Architectural Engineering, City University of Hong Kong, Kowloon, Hong Kong abstract article info Article history: Accepted 14 December 2012 Available online 21 January 2013 Keywords: BEMS Sensor bias Sensor calibration Uncertainty analysis Accurate sensors are the prerequisite for any building control and management operation to succeed. How- ever, due to the lack of awareness of the importance, sensors are often out of calibration. This paper proposes a preventive approach to determine sensor maintenance requirements, which aims to explicitly quantify the importance of sensors and determine a proper calibration frequency for sensors with high importance. The approach takes account of the sensor bias growth mechanism and relates the sensor importance to a user-defined building performance index. The sensor importance is evaluated under a stochastic framework, which essentially is the way in which sensor bias grows. The calibration frequency is identified using a poly- nomial function which describes the degradation rate of the building performance index. The proposed ap- proach is applied to a building located in Hong Kong. Results show that the proposed approach can effectively identify important sensors and their calibration frequency. © 2012 Elsevier B.V. All rights reserved. 1. Introduction A Building Energy Management System (BEMS) is a computer- based centralized system which is used to manage, control and monitor the mechanical and electrical equipment within a building or group of buildings [1]. By improving the performance of the control and management of the energy consumption system, a BEMS can reduce energy costs and labor requirements without sacrificing the thermal comfort required by the occupants [2]. Sensors are the fundamental equipment in a BEMS. Sensors sense the values of the variables of interest and send the measured signals to the corresponding components or devices [3]. Accurate sensors are the prerequisite for any building control and management strategy to succeed. However, due to the lack of awareness of the importance, sen- sors are often out of calibration, which is considered as one of the major causes of the energy waste [4]. Due to the large number of sensors installed in modern buildings, building owners and facility managers seldom have the motivation to examine the accuracy of every sensor when facing the low energy efficiency problem. Frequent calibration of all the sensors requires heavily on labor and is costly. Facing this problem, some approaches have been proposed from the field of sensor fault detection and diagnostics (FDD), in which various techniques and methods have been developed to monitor the perfor- mance of sensors and detect sensor faults [5–7]. In general, these ap- proaches can succeed after a sensor bias becomes abnormally large (larger than a pre-set threshold). Therefore they do not belong to preven- tive approaches. In terms of energy saving, these approaches are not very effective since the existent bias may be large enough to cause energy waste but not yet to be declared as ‘faults’. Preventive approaches have also been developed. For example, Huang et al. proposed using data fusion techniques to calibrate existing sensors to the one with the highest accu- racy [8]. Li et al. have listed various model based methods to be used as ‘ virtual sensors’ [9], which have the potential to calibrate sensors online. This paper proposes a systematic preventive approach to determine sensor maintenance requirements, in which a simulation platform inte- grating the models of the building under consideration and the equipped HVAC systems is constructed as a tool for analysis. The proposed approach takes account of the sensor bias growth mechanism and relates the sensor importance to a user-defined building performance index that can be cus- tomarily determined by stakeholders. Under a stochastic framework which essentially is the way in which the sensor bias grows, the pro- posed approach evaluates the importance of sensors and categorizes the sensors into different types according to the importance. For the sensors with high importance, the calibration frequency is identified using a polynomial function which describes the degradation rate of the building performance index. The rest of the paper is organized as two main parts. The first part introduces the methodology, and the second part applies the methodol- ogy to a hypothetical building located in Hong Kong, followed by the concluding remarks. 2. Methodology 2.1. Overview of the proposed methodology The flow chart of the proposed approach is illustrated in Fig. 1. The ap- proach can be divided into three stages: preparation, evaluation and re- port. In the preparation stage, a simulation platform is constructed; in Automation in Construction 31 (2013) 307–312 ⁎ Corresponding author. Tel.: +852 34427633; fax: +852 34429716. E-mail address: gongsheng.huang@cityu.edu.hk (H. Gongsheng). 0926-5805/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.autcon.2012.12.008 Contents lists available at SciVerse ScienceDirect Automation in Construction journal homepage: www.elsevier.com/locate/autcon