Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GiNet: Integrating Sequential and Context-Aware Learning for Battery Capacity Prediction

Sara Sameer, Wei Zhang, Xin Lou, Qingyu Yan, Terence Goh, Yulin Gao

TL;DR

GiNet addresses accurate battery SoC prediction under dynamic operating conditions by integrating sequential learning (GRU) with contextual long-range modeling (Informer). The method fuses GRU-derived temporal features with Informer's embedding and attention, enabling efficient processing of long sequences via ProbSparse attention and distillation. Experimental results on the Panasonic 18650PF dataset show GiNet achieves a minimal MAE of $0.11$ and consistently outperforms LSTM, GRU, and Informer, with an average improvement of $27\%$ over the second-best model. The work demonstrates the value of hybrid architectures that couple algorithmic advances with domain battery knowledge, and suggests broader applicability in BMS and industry settings.

Abstract

The surging demand for batteries requires advanced battery management systems, where battery capacity modelling is a key functionality. In this paper, we aim to achieve accurate battery capacity prediction by learning from historical measurements of battery dynamics. We propose GiNet, a gated recurrent units enhanced Informer network, for predicting battery's capacity. The novelty and competitiveness of GiNet lies in its capability of capturing sequential and contextual information from raw battery data and reflecting the battery's complex behaviors with both temporal dynamics and long-term dependencies. We conducted an experimental study based on a publicly available dataset to showcase GiNet's strength of gaining a holistic understanding of battery behavior and predicting battery capacity accurately. GiNet achieves 0.11 mean absolute error for predicting the battery capacity in a sequence of future time slots without knowing the historical battery capacity. It also outperforms the latest algorithms significantly with 27% error reduction on average compared to Informer. The promising results highlight the importance of customized and optimized integration of algorithm and battery knowledge and shed light on other industry applications as well.

GiNet: Integrating Sequential and Context-Aware Learning for Battery Capacity Prediction

TL;DR

GiNet addresses accurate battery SoC prediction under dynamic operating conditions by integrating sequential learning (GRU) with contextual long-range modeling (Informer). The method fuses GRU-derived temporal features with Informer's embedding and attention, enabling efficient processing of long sequences via ProbSparse attention and distillation. Experimental results on the Panasonic 18650PF dataset show GiNet achieves a minimal MAE of and consistently outperforms LSTM, GRU, and Informer, with an average improvement of over the second-best model. The work demonstrates the value of hybrid architectures that couple algorithmic advances with domain battery knowledge, and suggests broader applicability in BMS and industry settings.

Abstract

The surging demand for batteries requires advanced battery management systems, where battery capacity modelling is a key functionality. In this paper, we aim to achieve accurate battery capacity prediction by learning from historical measurements of battery dynamics. We propose GiNet, a gated recurrent units enhanced Informer network, for predicting battery's capacity. The novelty and competitiveness of GiNet lies in its capability of capturing sequential and contextual information from raw battery data and reflecting the battery's complex behaviors with both temporal dynamics and long-term dependencies. We conducted an experimental study based on a publicly available dataset to showcase GiNet's strength of gaining a holistic understanding of battery behavior and predicting battery capacity accurately. GiNet achieves 0.11 mean absolute error for predicting the battery capacity in a sequence of future time slots without knowing the historical battery capacity. It also outperforms the latest algorithms significantly with 27% error reduction on average compared to Informer. The promising results highlight the importance of customized and optimized integration of algorithm and battery knowledge and shed light on other industry applications as well.
Paper Structure (22 sections, 4 equations, 2 figures, 2 tables)

This paper contains 22 sections, 4 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. GiNet Methodology
  3. Model Overview and Technology Recap
  4. GRU
  5. Informer
  6. Building Blocks
  7. Data Pre-processing
  8. GRU-enhanced Features and Feature Fusion
  9. Feature Fusion
  10. Informer-based SoC Prediction
  11. Experimental Study
  12. Data
  13. Model Configuration and Implementation
  14. Comparison Study
  15. Overall Results
  16. ...and 7 more sections

Figures (2)

  • Figure 1: An illustration of GiNet architecture. GiNet performs data pre-processing for the battery time-series data. GRU is used to capture sequential information and extract corresponding features, and Informer is employed to model long-term dependence and extract contextual information. Supporting modules include feature fusion, output mapping to forecast horizon, etc.
  • Figure 2: The GiNet's achieved MAE with different numbers of encoder and decoder layers. The results for input windows 10, 100, and 200 for forecast horizon 25 are reported. E$i$D$j$ in legend means $i$ encoder layers and $j$ decoder layers. GiNet performs the best with two encoder layers and one decoder layer, i.e., E2D1 in the figure.