Physics-Informed Deep Learning Predicts Battery Thermal Runaway
A recent study published on arXiv introduces a framework called Physics-Informed Long Short-Term Memory (PI-LSTM) designed for anticipating thermal runaway in lithium-ion batteries. This innovative method incorporates fundamental heat transfer equations into a deep learning model through a physics-informed regularization term within the loss function. It analyzes various input sequences, such as state of charge, voltage, current, mechanical stress, and surface temperature. The goal is to reconcile traditional data-driven techniques like LSTM, which may breach thermodynamic laws, with physics-based thermal models that require significant computational resources. This research is crucial for enhancing the safety, efficiency, and dependability of contemporary energy storage systems.
Key facts
- The study is published on arXiv with ID 2604.20175.
- The proposed model is called Physics-Informed Long Short-Term Memory (PI-LSTM).
- It integrates heat transfer equations into the deep learning architecture.
- The physics-based regularization term is added to the loss function.
- Input features include state of charge, voltage, current, mechanical stress, and surface temperature.
- Conventional LSTM networks can produce physically inconsistent predictions.
- Physics-based thermal models are computationally expensive for real-time use.
- The framework targets proactive thermal runaway forecasting in Li-ion batteries.
Entities
Institutions
- arXiv