Multi-Data Fusion Detection for Lithium Battery Thermal Runaway Based on GRU Neural Network

In current lithium battery thermal runaway detection, traditional neural networks suffer from insufficient temporal feature extraction capability, leading to low accuracy. To address this issue, this paper proposes a multi-data fusion detection method based on Gated Recurrent Unit (GRU) networks, which can effectively extract temporal features for precise detection. First, the thermal runaway process of lithium batteries is simulated using Pyrosim to collect data on five characteristic parameters: CO, H<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>, VOC, temperature, and smoke. Subsequently, a GRU neural network model is designed on the Matlab platform to perform multi-data fusion detection for the five types of data generated during lithium battery thermal runaway. Experimental results demonstrate that the proposed method achieves an accuracy of 98% in thermal runaway detection by leveraging temporal features, representing an improvement of 4.9019% compared to BP neural networks. This study provides a novel technical approach for accurate lithium battery thermal runaway detection.