Sera: Semantic Representation Aggregation for Reliable and Interpretable Battery Health Forecasting

2026-10-08

AI Summary

This study addresses the challenge of achieving both reliability and interpretability in battery health prediction, where nonlinear degradation and heterogeneity complicate modeling, and existing methods lack explicit representations of high-level degradation semantics. We propose Sera, a novel framework that integrates structured degradation semantics with temporal modeling for the first time. Specifically, it extracts degradation semantics using domain knowledge, constructs complementary representations via rule engines and large language models (LLMs), and enhances temporal models through a gated aggregation mechanism. Experimental results demonstrate that Sera reduces prediction error by up to 37.3% compared to baselines. Furthermore, counterfactual analysis validates the consistency between key degradation descriptors and predictive responses, confirming that the proposed approach achieves reliable and interpretable health prediction.

Abstract

Battery state of health (SoH) forecasting is important for battery management, but remains challenging due to nonlinear degradation and heterogeneity across batteries. Existing data-driven approaches primarily use temporal models to learn from numerical battery time series, and higher-level degradation characteristics are often not explicitly represented. These characteristics, however, can provide degradation guidance to support reliable forecasting and make the influence of degradation more interpretable. In this paper, we propose \textsc{Sera}, a \underline{se}mantic \underline{r}epresentation \underline{a}ggregation framework that complements temporal modelling with degradation semantics. Guided by battery domain expertise, \textsc{Sera} extracts degradation semantics from time series and constructs two complementary representations using rule-based knowledge and LLM-based interpretation. The representations are independently encoded and integrated with the representation learned by temporal models through gated aggregations. Experiments on the mainstream benchmark across multiple prediction horizons and different temporal models show that \textsc{Sera} consistently improves forecasting performance, achieving up to a 37.3\% reduction in prediction error over the temporal baseline and enhanced generalizability. Counterfactual analysis examines how forecasts respond to changes in degradation semantics to assess interpretability. The results show that prediction responses are consistent with the meanings of key degradation descriptors across tested horizons. Together, these findings demonstrate that structured degradation semantics and effective aggregation can improve forecasting accuracy and support reliable and interpretable battery health forecasting for advanced battery management.

Problem

• Battery State of Health Forecasting
• Nonlinear Degradation
• Degradation Semantics
• Interpretability
• Data-driven Approaches

Innovation

• Semantic Representation Aggregation
• Battery State of Health Forecasting
• Large Language Model
• Gated Aggregation
• Counterfactual Interpretability