Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine Learning Detection of Lithium Plating in Lithium-ion Cells: A Gaussian Process Approach

Ayush Patnaik, Jackson Fogelquist, Adam B Zufall, Stephen K Robinson, Xinfan Lin

TL;DR

This work tackles real-time detection of lithium plating during fast charging by directly modeling the charge–voltage relation $Q(V)$ with a Gaussian Process and analytically deriving the derivative $dQ/dV$. The approach yields noise-aware estimates with closed-form uncertainties, avoiding the noise amplification of finite differencing. It demonstrates that a $dQ/dV$ peak above $4.0$ V serves as a robust plating signature under challenging conditions, and correlates with degraded charge throughput and capacity fade. The method offers a practical path toward online plating diagnostics and integration into battery management systems for safer, more reliable fast charging.

Abstract

Lithium plating during fast charging is a critical degradation mechanism that accelerates capacity fade and can trigger catastrophic safety failures. Recent work has identified a distinctive dQ/dV peak above 4.0 V as a reliable signature of plating onset; however, conventional methods for computing dQ/dV rely on finite differencing with filtering, which amplifies sensor noise and introduces bias in peak location. In this paper, we propose a Gaussian Process (GP) framework for lithium plating detection by directly modeling the charge-voltage relationship Q(V) as a stochastic process with calibrated uncertainty. Leveraging the property that derivatives of GPs remain GPs, we infer dQ/dV analytically and probabilistically from the posterior, enabling robust detection without ad hoc smoothing. The framework provides three key benefits: (i) noise-aware inference with hyperparameters learned from data, (ii) closed-form derivatives with credible intervals for uncertainty quantification, and (iii) scalability to online variants suitable for embedded BMS. Experimental validation on Li-ion coin cells across a range of C-rates (0.2C-1C) and temperatures (0-40°C) demonstrates that the GP-based method reliably detects plating peaks under low-temperature, high-rate charging, while correctly reporting no peaks in baseline cases. The concurrence of GP-identified differential peaks, reduced charge throughput, and capacity fade measured via reference performance tests confirms the method's accuracy and robustness, establishing a practical pathway for real-time lithium plating detection.

Machine Learning Detection of Lithium Plating in Lithium-ion Cells: A Gaussian Process Approach

TL;DR

This work tackles real-time detection of lithium plating during fast charging by directly modeling the charge–voltage relation with a Gaussian Process and analytically deriving the derivative . The approach yields noise-aware estimates with closed-form uncertainties, avoiding the noise amplification of finite differencing. It demonstrates that a peak above V serves as a robust plating signature under challenging conditions, and correlates with degraded charge throughput and capacity fade. The method offers a practical path toward online plating diagnostics and integration into battery management systems for safer, more reliable fast charging.

Abstract

Lithium plating during fast charging is a critical degradation mechanism that accelerates capacity fade and can trigger catastrophic safety failures. Recent work has identified a distinctive dQ/dV peak above 4.0 V as a reliable signature of plating onset; however, conventional methods for computing dQ/dV rely on finite differencing with filtering, which amplifies sensor noise and introduces bias in peak location. In this paper, we propose a Gaussian Process (GP) framework for lithium plating detection by directly modeling the charge-voltage relationship Q(V) as a stochastic process with calibrated uncertainty. Leveraging the property that derivatives of GPs remain GPs, we infer dQ/dV analytically and probabilistically from the posterior, enabling robust detection without ad hoc smoothing. The framework provides three key benefits: (i) noise-aware inference with hyperparameters learned from data, (ii) closed-form derivatives with credible intervals for uncertainty quantification, and (iii) scalability to online variants suitable for embedded BMS. Experimental validation on Li-ion coin cells across a range of C-rates (0.2C-1C) and temperatures (0-40°C) demonstrates that the GP-based method reliably detects plating peaks under low-temperature, high-rate charging, while correctly reporting no peaks in baseline cases. The concurrence of GP-identified differential peaks, reduced charge throughput, and capacity fade measured via reference performance tests confirms the method's accuracy and robustness, establishing a practical pathway for real-time lithium plating detection.

Paper Structure

This paper contains 9 sections, 1 theorem, 17 equations, 3 figures, 1 table.

Table of Contents

  1. INTRODUCTION
  2. Mechanism and Lithium Plating and Electrochemical Features
  3. Gaussian Process-based Lithium Plating Detection Framework
  4. Model Formulation: GP over Q(V)
  5. Inference of $dQ/dV$
  6. Experimental Validation
  7. Testing Setup
  8. Results and Discussion
  9. Conclusions

Key Result

Theorem 1

Let and assume $f$ is mean-square differentiable (equivalently, $k$ is $C^2$ near the diagonal). Then the derivative process is also a Gaussian process with mean and covariance Moreover, $f$ and $f'$ are jointly Gaussian. For any finite sets the random vector is multivariate normal with mean and block covariance where the blocks are defined element-wise by

Figures (3)

  • Figure 1: Representative test sequence (1C at $25\,^{\circ}\mathrm{C}$) showing the complete protocol used for experiments. Degradation cycles consisting of repeated constant-current charge/discharge cycles (left), followed by a Reference Performance Test (RPT) consisting of a low-rate capacity test and Hybrid Pulse Power Characterization (HPPC) pulses (right). Blue = voltage (left axis); red = current (right axis).
  • Figure 2: GP-based plating detection across charge rates and temperatures. (a) V(t) during CC charge for representative cycles (legend shows C-rate and temperature). (b) capacity–voltage trajectories Q(V). (c) GP-derived incremental capacity dQ/dV with inset figures (i), (ii) and (iii) showing the magnified 95% credible bands at the start of the cycle with bounds as % of peak magnitude of the no-plating and plating clusters, respectively.
  • Figure 3: Normalized Charge Throughput values for each CC (constant-current) charge cycle across charging rates and temperatures for the first 10 cycles

Theorems & Definitions (1)

  • Theorem 1: Derivative GP and Joint Gaussianity rasmussen_gaussian_2005