Semantic battery data improves health forecasts and insights
Sera: Semantic Representation Aggregation for Reliable and Interpretable Battery Health Forecasting
Machine LearningArtificial Intelligence
Summary
Batteries wear out over time, but predicting exactly how and when can be hard because each battery is different and ages in complex ways. The authors created a new method called Sera that not only looks at the usual battery data over time but also adds expert knowledge about how batteries degrade. This approach combines rules from battery experts and interprets data with large language models to better understand battery health changes. Tests show Sera makes predictions more accurate and easier to understand. This helps battery managers trust forecasts and plan better battery use.
What this means in practice
- •For battery management teams: Provide more accurate and interpretable predictions of battery health for better maintenance scheduling and usage decisions.
- •For energy storage operators: Improve reliability of battery life forecasts to optimize energy storage system operations and prevent unexpected failures.
Authors
Jiawei Li, Fang Liu, Wei Zhang, Zuming Liu, Man-Fai Ng, Zhi Wei Seh
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.