Thermal Runaway Early Warning Algorithm for Lithium-ion Batteries Using NGO-LSTM Neural Network

To address the technical limitations of insufficient prediction accuracy and high false alarm rates in traditional fire alarm systems for monitoring thermal runaway incidents in lithium-ion batteries, this study innovatively proposes a hybrid early warning model integrating Northern Goshawk Optimization Algorithm (NGO) and the Long short-term Memory neural network (LSTM). By incorporating the bio-inspired intelligent optimization mechanism of the NGO, the model Effective balancing exploration and exploitation during the global search in parameter space. The experimental results show that the LSTM model based on the Northern Goshawk optimization algorithm is successfully implemented thermal runaway prediction accuracies of 96.83% and 93.57% on training and test sets respectively, surpassing the baseline LSTM model by 9.69% and 6.34% improvements. This research provides a novel algorithmic framework for early-stage thermal runaway warning in lithium- ion batteries, offering significant engineering applications for enhancing intelligent fire prevention in energy storage systems.