Complete electrothermal and lifetime model of 18650 nickel manganese cobalt cell based on artificial neural network

This study presents a comprehensive electrothermal and lifetime model for cylindrical 3 ampere-hours (Ah) lithium-ion cells using artificial neural networks (ANNs) to estimate the cell’s lifespan. The model combines an electrothermal component with an ANN-based lifetime prediction approach, offering a holistic representation of cell behavior over its lifetime by incorporating key parameters, including the state of charge, temperature, current, and cycle life. The ANN is trained offline using extensive experimental data collected from Sony cylindrical 3 Ah cells under various operating conditions. The electrothermal component employs a second-order Thévenin equivalent circuit model topology, enhanced with extended versions of characterization and parameterization procedures. Validation of the coupled model is performed using laboratory tests at different stages of the cells’ life (500, 1000, and 1500 cycles), demonstrating its ability to estimate cell electrical and thermal performance across a broad lifespan range. Results indicate a maximum error of 1% in voltage readings and 3% in temperature evolution during discharge with the complete model. This comprehensive approach not only enhances the understanding of long-term Sony 3 Ah cell dynamics but also provides a computationally efficient tool for battery management systems and control strategies. The model’s capability to predict both electrical and thermal performance simultaneously at different stages of the cell’s lifetime makes it particularly valuable for optimizing battery performance and lifespan in various applications.