State of health estimation of lithium-ion battery considering sensor uncertainty

The state of health (SOH) is closely related to the stable operation of lithium-ion (Li-ion) battery energy storage systems. Existing methods for SOH estimation assume a reliable multi-source sensor in the battery management system, yet the sensors are prone to failures leading to abnormal measurement data in practical applications. To alleviate the above issue, this paper proposes a detection before fusion multi-model framework (DFMF) considering the uncertainties of multi-source sensor data. Firstly, a density-based spatial clustering of applications with noise is adopted to identify the working states of sensors via detecting the faults from their raw data. Then, the working states of sensors are fed to a state machine to realize the fault isolation. Finally, a data-driven SOH estimation based on multi-source sensors is developed, which includes both the single-sensor mode and the multi-sensor joint mode. The proposed DFMF is validated on two degradation datasets from different aging conditions and chemistries. The experimental results demonstrate the proposed DFMF has a high accuracy and strong robustness in Li-ion SOH estimation. It is noted that the DFMF can provide SOH estimation results with the root mean squared error within 1.68 % and 1.17 % on the two datasets, respectively.

Automated summary

This paper proposes a fault detection method that considers the uncertainties of multi-source sensor data. It uses density-based clustering and a state machine to isolate faults, and develops a data-driven health estimation model based on multi-source sensors. The experimental results demonstrate its high accuracy and robustness.