A method and system for graded early warning of thermal runaway faults in energy storage batteries
By using multi-scale time windows and Bayesian-optimized wavelet transform parameters, combined with a neural network model, the problem of capturing multi-time-scale features of thermal runaway faults in energy storage batteries was solved, achieving high-precision hierarchical early warning and differentiated control, thereby improving the safety and reliability of energy storage systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEFEI GUOXUAN HIGH TECH POWER ENERGY
- Filing Date
- 2026-04-02
- Publication Date
- 2026-05-26
AI Technical Summary
Existing technologies struggle to effectively capture the multi-timescale characteristics of thermal runaway faults in energy storage batteries, leading to the loss or dilution of key precursor information. Furthermore, the lack of a collaborative analysis framework limits the accuracy and generalization capabilities of early warning systems.
By employing a multi-scale time window strategy and overlapping sampling, combined with Bayesian optimization algorithm to select wavelet transform parameters, a neural network model integrating multi-scale features is constructed. Through probabilistic model fusion and multi-task learning, a graded early warning system for thermal runaway is achieved.
It significantly increased the effective sample size, comprehensively captured early signs of failure, improved feature quality and prediction accuracy, realized differentiated control with graded early warning, and enhanced the safety and reliability of energy storage systems.
Smart Images

Figure CN122085129A_ABST