Remaining useful life prediction of Lithium-ion batteries using spatio-temporal multimodal attention networks

TL;DR

This work tackles the challenge of predicting remaining useful life (RUL) for Lithium-ion batteries in real-world settings where end-of-life (EOL) information is not readily available. It introduces a two-stage framework: first, a health-state (HS) division that identifies the First Prediction Cycle (FPC), and second, a spatio-temporal multimodal attention network (ST-MAN) that predicts the RUL as a percentage from FPC to EOL using multimodal features (e.g., discharge/charge capacities, temperature, internal resistance, charge time). The ST-MAN integrates CNNs, a cross-channel Transformer, and LSTM-based temporal modeling with temporal attention, and optimizes a composite loss combining MAE, RMSE, and MAPE, achieving state-of-the-art accuracy on MIT and HUST battery datasets (e.g., MAE $0.0275$, MSE $0.0014$) without requiring prior EOL knowledge. Ablation studies confirm the importance of channel interactions and temporal modeling, while the method demonstrates smoother, more reliable RUL trajectories than conventional approaches. The work highlights practical potential for battery health management, with considerations for embedded deployment and future reductions in computation (FLOPs) for edge devices.

Abstract

Lithium-ion batteries are widely used in various applications, including electric vehicles and renewable energy storage. The prediction of the remaining useful life (RUL) of batteries is crucial for ensuring reliable and efficient operation, as well as reducing maintenance costs. However, determining the life cycle of batteries in real-world scenarios is challenging, and existing methods have limitations in predicting the number of cycles iteratively. In addition, existing works often oversimplify the datasets, neglecting important features of the batteries such as temperature, internal resistance, and material type. To address these limitations, this paper proposes a two-stage RUL prediction scheme for Lithium-ion batteries using a spatio-temporal multimodal attention network (ST-MAN). The proposed ST-MAN is to capture the complex spatio-temporal dependencies in the battery data, including the features that are often neglected in existing works. Despite operating without prior knowledge of end-of-life (EOL) events, our method consistently achieves lower error rates, boasting mean absolute error (MAE) and mean square error (MSE) of 0.0275 and 0.0014, respectively, compared to existing convolutional neural networks (CNN) and long short-term memory (LSTM)-based methods. The proposed method has the potential to improve the reliability and efficiency of battery operations and is applicable in various industries.

Figures (10)

• Figure 1: Overview of the proposed framework. Stage 1 is the health state (HS) division step using the LSTM classifier and Stage 2 is a remaining useful life (RUL) prediction step using the proposed spatio-temporal multimodal attention network (ST-MAN) architecture.
• Figure 2: HS division classifier network architecture based on the LSTM units.
• Figure 3: Discharge Capacity at initial and last 10% of the total cycles examples from MIT (Upper) and HUST dataset (Lower)
• Figure 4: The architecture overview of the proposed spatial-temporal multimodal attention networks (ST-MAN). The network consists of convolutional operations, a cross-channel Transformer encoder block, a fully connected layer to fuse cross-channel information, an LSTM module to extract temporal information, and a temporal attention module to improve the temporal information extraction.
• Figure 5: Discharge capacity degradation patterns of (a) 124 Lithium-ion battery cells from the MIT dataset severson2019data, (b) 77 Lithium-ion battery cells from HUST dataset ma2022real