Physics-informed gated recurrent unit neural networks model for surface temperature estimation of Lithium-ion batteries

Thermal management is vital for optimizing Lithium-Ion Battery (LIB) performance. Accurately measuring individual cell temperatures using physical sensors poses significant cost and complexity challenges particularly in packs containing hundreds or thousands of cells. State-of-the-art estimation techniques struggle with nonlinear LIB characteristics under varying operating conditions. This thesis introduces a novel surface temperature estimation method that integrates a gated recurrent unit (GRU) network with physics-informed neural networks (PINNs). The GRU processes sequential data voltage, current, and ambient temperature capturing dynamic battery behavior, while the physics-informed layers embed critical physical parameters, including electrical, thermal, and heat generation models, during training. Validated experimentally on a 14-series connected cell module under varying C-rates and ambient temperatures. The model achieves mean absolute errors of 1.32°C, 1.33°C, and 1.91°C for 1C, 1.5C, and 2C rates at 0°C, 10°C, and 25°C. Results demonstrate superior robustness and adaptability compared to traditional machine learning models like LSTM and FNN.

Name thesis:degree_name

Master of Applied Science (MASc)

Discipline thesis:degree_discipline

Electrical and Computer Engineering

Grantor

University of Ontario Institute of Technology

Year dc:date.issued

2025

Author dc:creator

• Sharma, Mohit

Advisor dc:contributor.advisor

• Williamson, Sheldon

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10155/1939

OAI identifier oai:identifier

oai:ontariotechu.scholaris.ca:10155/1939