Lithium-Ion Battery System Health Monitoring and Resistance-Based Fault Analysis from Field Data Using Recursive Spatiotemporal Gaussian Processes

Joachim Schaeffer, Eric Lenz, Duncan Gulla, Martin Z. Bazant, Richard Braatz, Rolf Findeisen

Keywords: lithium-ion batteries, field data, spatiotemporal Gaussian processes, machine learning, health monitoring

TL;DR: We monitor battery systems using recursive spatiotemporal Gaussian processes and develop probabilistic fault probabilities to analyze battery system faults.

Abstract:

Health monitoring is important for the safe operation of battery systems. We use recursive spatiotemporal Gaussian processes to model the resistance of lithium iron phosphate batteries from field data. These processes scale linearly with the number of data points, allowing online monitoring. The kernels separate the time-dependent and operating-point-dependent resistance contributions.
We develop probabilistic fault probabilities based on time-dependent resistance estimates. The fault analysis underlines that often, only a single cell shows abnormal behavior, consistent with weakest-link failure for cells connected in series, amplified by local resistive heating. The results further the understanding of how battery packs degrade and fail in the field and demonstrate the potential of online monitoring. The data set contains 28 battery systems returned to the manufacturer for warranty, each with eight cells in series, totaling 224 cells and 133 million data rows. The data and code are openly available.

Submission Number: 39