[Paper Review] Latent Function Decomposition for Forecasting Li-ion Battery Cells Capacity: A Multi-Output Convolved Gaussian Process Approach

Abdallah Chehade, Ala A. Hussein|arXiv (Cornell University)|Jul 19, 2019

Advanced Battery Technologies Research34 references4 citations

TL;DR

This paper proposes a Multi-Output Convolved Gaussian Process (MCGP) framework that decomposes capacity trends of multiple lithium-ion battery cells into latent functions, which are then convolved via kernel smoothers to forecast future capacity with high accuracy. The method captures cross-correlations between cells, provides predictive uncertainty, and outperforms benchmarks on experimental data.

ABSTRACT

A latent function decomposition method is proposed for forecasting the capacity of lithium-ion battery cells. The method uses the Multi-Output Gaussian Process, a generative machine learning framework for multi-task and transfer learning. The MCGP decomposes the available capacity trends from multiple battery cells into latent functions. The latent functions are then convolved over kernel smoothers to reconstruct and/or forecast capacity trends of the battery cells. Besides the high prediction accuracy the proposed method possesses, it provides uncertainty information for the predictions and captures nontrivial cross-correlations between capacity trends of different battery cells. These two merits make the proposed MCGP a very reliable and practical solution for applications that use battery cell packs. The MCGP is derived and compared to benchmark methods on an experimental lithium-ion battery cells data. The results show the effectiveness of the proposed method.

Motivation & Objective

• To address the challenge of accurately forecasting the capacity degradation of multiple lithium-ion battery cells in a pack.
• To model complex, non-linear cross-correlations between capacity trends of different cells.
• To provide reliable uncertainty estimates alongside predictions for safety-critical battery applications.
• To develop a generative machine learning framework that enables transfer learning across battery cells.

Proposed method

• The method employs a Multi-Output Gaussian Process (MCGP) to model capacity trends from multiple battery cells as outputs of a shared latent function space.
• Latent functions are inferred from observed capacity data using a probabilistic generative model, capturing underlying degradation patterns.
• These latent functions are convolved with kernel smoothers to reconstruct and forecast individual cell capacity trends.
• The convolution operation allows for flexible, non-stationary modeling of capacity degradation across cells.
• The MCGP framework enables transfer learning by sharing information across cells through the latent function space.
• The model is trained end-to-end using observed capacity data, with uncertainty propagated through the full inference chain.

Experimental results

Research questions

• RQ1 Can a shared latent function space effectively model the degradation trends of multiple lithium-ion battery cells?
• RQ2 How well can the MCGP framework capture nontrivial cross-correlations between capacity trends of different cells?
• RQ3 To what extent does the method improve forecasting accuracy compared to benchmark models?
• RQ4 Can the model provide reliable uncertainty estimates for battery capacity predictions in real-world applications?

Key findings

• The proposed MCGP method achieves higher prediction accuracy than benchmark models on experimental Li-ion battery data.
• The method effectively captures nontrivial cross-correlations between capacity trends of different battery cells.
• Predictive uncertainty estimates are well-calibrated and provide reliable confidence intervals for forecasts.
• The MCGP framework enables transfer learning, improving generalization when data per cell is limited.
• The convolution of latent functions with kernel smoothers enhances reconstruction and forecasting performance.
• The method demonstrates robustness and reliability for practical applications involving battery cell packs.