Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Battery GraphNets : Relational Learning for Lithium-ion Batteries(LiBs) Life Estimation

Sakhinana Sagar Srinivas, Rajat Kumar Sarkar, Venkataramana Runkana

TL;DR

This work presents Battery GraphNets (BGN), a framework for lithium-ion battery RUL prognosis that jointly learns time-evolving discrete relationships among battery parameters and a Graph Neural Network backbone to capture nonlinear degradation. By combining Dynamic Graph Inference (DGI), a Grapher with GNN and GRU blocks, and a Graph Readout, the model produces accurate RUL predictions with uncertainty estimates via BGN-UE. Extensive experiments on NASA and UNIBO datasets show state-of-the-art RMSE improvements over diverse baselines, and ablations demonstrate the critical role of the dynamic graph learning and relational encoding. The approach offers a scalable, uncertainty-aware prognostic tool with potential to enhance battery management systems and support hybrid physics–data-driven battery life models.

Abstract

Battery life estimation is critical for optimizing battery performance and guaranteeing minimal degradation for better efficiency and reliability of battery-powered systems. The existing methods to predict the Remaining Useful Life(RUL) of Lithium-ion Batteries (LiBs) neglect the relational dependencies of the battery parameters to model the nonlinear degradation trajectories. We present the Battery GraphNets framework that jointly learns to incorporate a discrete dependency graph structure between battery parameters to capture the complex interactions and the graph-learning algorithm to model the intrinsic battery degradation for RUL prognosis. The proposed method outperforms several popular methods by a significant margin on publicly available battery datasets and achieves SOTA performance. We report the ablation studies to support the efficacy of our approach.

Battery GraphNets : Relational Learning for Lithium-ion Batteries(LiBs) Life Estimation

TL;DR

This work presents Battery GraphNets (BGN), a framework for lithium-ion battery RUL prognosis that jointly learns time-evolving discrete relationships among battery parameters and a Graph Neural Network backbone to capture nonlinear degradation. By combining Dynamic Graph Inference (DGI), a Grapher with GNN and GRU blocks, and a Graph Readout, the model produces accurate RUL predictions with uncertainty estimates via BGN-UE. Extensive experiments on NASA and UNIBO datasets show state-of-the-art RMSE improvements over diverse baselines, and ablations demonstrate the critical role of the dynamic graph learning and relational encoding. The approach offers a scalable, uncertainty-aware prognostic tool with potential to enhance battery management systems and support hybrid physics–data-driven battery life models.

Abstract

Battery life estimation is critical for optimizing battery performance and guaranteeing minimal degradation for better efficiency and reliability of battery-powered systems. The existing methods to predict the Remaining Useful Life(RUL) of Lithium-ion Batteries (LiBs) neglect the relational dependencies of the battery parameters to model the nonlinear degradation trajectories. We present the Battery GraphNets framework that jointly learns to incorporate a discrete dependency graph structure between battery parameters to capture the complex interactions and the graph-learning algorithm to model the intrinsic battery degradation for RUL prognosis. The proposed method outperforms several popular methods by a significant margin on publicly available battery datasets and achieves SOTA performance. We report the ablation studies to support the efficacy of our approach.
Paper Structure (19 sections, 20 equations, 1 figure, 7 tables, 2 algorithms)

This paper contains 19 sections, 20 equations, 1 figure, 7 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Problem statement
  3. Proposed Approach
  4. Dynamic Graph Inference(DGI)
  5. Grapher
  6. Graph Readout(GRo)
  7. Uncertainty Estimation
  8. Experiments and Results
  9. Datasets & Experimental Settings
  10. Benchmarking algorithms
  11. Ablation Studies
  12. Hyperparameters Study
  13. Conclusion
  14. Appendix
  15. Graph Contrastive Algorithms
  16. ...and 4 more sections

Figures (1)

  • Figure 1: The Battery GraphNets(BGN) framework. (a) The DGI module learns the dynamic graphs. (b) The Grapher module consists of GNN and RNN blocks. The GNN block operates on the dynamic graph topology to encode the local-graph neighborhood information in the node-level representations. The RNN block regulates the information flow to capture long-range spatio-temporal dependencies in the node representations. (c) The Graph Readout(GRo) module computes the entire graph representation to preserve the global graph information. $y^{t}_{p}$ denotes the model predictions.