Abstract
Sensors are one of the key components in a modern building energy management system (BEMS). Accurate sensors are the prerequisite for the success of any building energy optimization strategy. As sensors are subject to environment disturbance and performance deterioration, their accuracy tends to decrease during their service lives. Sensor calibration is an efficient way to improve measurement accuracy and reliability. However, due to a large number of sensors installed in modern air conditioning (AC) systems, conventional regular calibration may be laborious while not optimal when the system energy performance are concerned. In order to improve measurement accuracy and reliability, a hybrid sensor management strategy is proposed in this article. This strategy integrates a measured variable importance ranking technique with a data fusion technique. Comparison of this strategy with a conventional regular calibration in case studies shows that this strategy improves both the energy and control performance of AC systems.
Nomenclature
b | = | bias growth coefficient |
C | = | cost |
d | = | Moffatt distance |
D | = | performance deviation |
D | = | Moffat distance vector |
E | = | energy consumption |
I | = | dimensionless variable |
i | = | used to denote the sampling time |
K | = | total number of measured variables |
l | = | length of moving window |
m | = | measurement |
N | = | number of sensors |
n | = | number of sampling steps |
R | = | standardized regression coefficient |
r | = | count of normal measurements |
s | = | random variable |
t | = | time |
T | = | total operation period |
U | = | uncertainty index |
Greek
ϵ | = | error |
σ | = | standard deviation |
κ | = | counter |
ξ | = | user defined parameter |
τ | = | calibration interval |
γ | = | ratio of calibration interval of important sensors to that of nonimportant sensors |
Subscript
b | = | bias |
bs | = | base line |
dr | = | sensor drift |
e | = | energy consumption |
f | = | energy consumption under sensor fault |
i, j, k | = | numerical indices |
l | = | limit |
n | = | white noise |
o | = | operation |
r | = | random stress after testing or calibration |
t | = | true value |
max | = | maximum |
mt | = | maintenance |
tr | = | tracking |
un | = | with uncertainty |
BOP | = | testing or calibration |
Funding
The work described in this paper was fully supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. CityU 114011).