Pace: Physics-Aware Attentive Temporal Convolutional Network for Battery Health Estimation

2025-12-12

AI Summary

To address the low accuracy and poor deployability of battery State of Health (SOH) estimation, this paper proposes a lightweight physics-informed time-series modeling framework. Methodologically, it introduces a novel battery-specific three-module architecture: (i) dilated temporal blocks to capture multi-scale degradation dynamics; (ii) block-sparse attention to model long-range temporal dependencies; and (iii) a dual-head output structure jointly predicting capacity and internal resistance—two key SOH indicators. Physical features derived from an equivalent circuit model are embedded, and efficient inference is enabled via temporal convolution. Evaluated on a large public benchmark dataset, the method achieves 6.5× and 2.0× reductions in mean absolute error for capacity and internal resistance estimation, respectively, over the best baseline. Furthermore, the model is successfully deployed on a Raspberry Pi, enabling millisecond-level real-time edge inference.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Intelligent Robots: State Estimation
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Batteries are critical components in modern energy systems such as electric vehicles and power grid energy storage. Effective battery health management is essential for battery system safety, cost-efficiency, and sustainability. In this paper, we propose Pace, a physics-aware attentive temporal convolutional network for battery health estimation. Pace integrates raw sensor measurements with battery physics features derived from the equivalent circuit model. We develop three battery-specific modules, including dilated temporal blocks for efficient temporal encoding, chunked attention blocks for context modeling, and a dual-head output block for fusing short- and long-term battery degradation patterns. Together, the modules enable Pace to predict battery health accurately and efficiently in various battery usage conditions. In a large public dataset, Pace performs much better than existing models, achieving an average performance improvement of 6.5 and 2.0x compared to two best-performing baseline models. We further demonstrate its practical viability with a real-time edge deployment on a Raspberry Pi. These results establish Pace as a practical and high-performance solution for battery health analytics.

Problem

• Estimates battery health using physics-aware neural network
• Integrates sensor data with battery physics features
• Improves accuracy and efficiency in various usage conditions

Innovation

• Integrates raw sensor data with battery physics features
• Uses dilated temporal blocks and chunked attention modules
• Fuses short- and long-term degradation patterns via dual-head output