Energy Reports (Nov 2022)
A fault diagnosis method for VRLA battery in data center
Abstract
The Valve-Regulated Lead–Acid (VRLA) battery is one of the important components of the auxiliary power supply system in the data center. Battery failure in the data center poses a great threat to the smooth operations in the data center. The biggest challenge for the data centers is how to accurately diagnose the faulty battery in real-time in order to ensure the safe operations of the data center. According to the working characteristics of VRLA battery, a battery fault diagnosis algorithm is proposed based on the historical data of battery voltage and internal resistance. The algorithm combines the growth rate of voltage and internal resistance to obtain the indicator that characterizes the performance change of the battery during the working by statistical method. The real-time diagnosis of the faulty battery is estimated by using the Pauta criterion to discriminate the indicator of the battery performance. The effectiveness of the method was verified through a real-time diagnostic test of VRLA battery in a data center. The fault diagnosis method in this paper is compared with the median absolute deviation (MAD) algorithm in the same data set to avoids the missed diagnosis due to voltage value not reaching the fault limits and the false diagnosis caused by power surge or charge–discharge cycle. Compared with the hierarchical clustering based on K-shape algorithm, our algorithm does not need to diagnose the battery in groups. The diagnosis result of the faulty battery is about 7 days faster, and there is no misdiagnosis when multiple batteries fail at the same time. The algorithm can also set the corresponding fault limits for battery fault diagnosis according to the different performance requirements of VRLA battery in data center.
