Data-driven predictive maintenance for battery energy storage systems
In this project, we created an online diagnostic tool called "State of Health and Ageing Metering (SAM 1.0)". It performs predictive maintenance by measuring/calculating the state-of-health (SoH)+ and detecting cells and modules in a battery pack that are close to failing and must be replaced.
The proposed system has the following features: a) It is an online diagnostic tool capable of updating the capacity degradation of the battery storage system without any time-consuming tests. b) It is capable of detecting modules that are close to failing and must be replaced. c) There is no need for an additional sensor or measurement (just the info from BMS).
The proposed system’s structure is as follows. First, the voltage, current, and temperature data from battery cells or modules are stored in the SAM database. The SoH of the battery is then com-puted using the online pattern recognition and online learning model for ageing metering. Furthermore, the remaining capacity can be calculated in real-time using also the measured equivalent series resistance (ESR). We can detect the faulted or near-faulted cells that need to be replaced using a variety of classification and regression algorithms. The energy and power bounds in order to avoid unbalanced module or string are calculated also be calculated.