Thesis - Physics-Informed Machine Learning for Reduced-Order Battery Modelling in Embedded BMS

30 hp – Physics-Informed Machine Learning for Reduced-Order Battery Modelling in Embedded BMS

Introduction

Thesis work is an excellent way to get closer to Scania and build relationships for the future. Many of today's employees began their Scania career with their degree project.

TRATON R&D develops core technologies for the global brands Scania, MAN, and International. This master's thesis offers an opportunity to work with advanced machine learning, physics-based modelling, and embedded systems for heavy-duty electric vehicles.

Background

Advanced electrochemical battery models, such as the Doyle–Fuller–Newman (DFN) model, provide detailed information about internal battery behaviour. However, their computational complexity limits their direct application in production Battery Management Systems (BMS).

Physics-Informed Neural Networks (PINNs) offer a promising approach to developing computationally efficient models by combining physical equations with machine learning. Their accuracy, generalization, and suitability for embedded applications require further investigation.

Objective

Develop and evaluate physics-informed machine learning approaches for computationally efficient electrochemical battery modelling.

The thesis will investigate whether incorporating physical knowledge improves prediction accuracy, robustness, and generalization compared with conventional machine learning, while assessing the feasibility of embedded BMS implementation.

Job description

The thesis work will include the following tasks:

1. Analyze synthetic data generated from electrochemical battery models, including selected internal states and operating conditions.

2. Develop a conventional machine learning baseline for approximating battery dynamics.

3. Investigate physics-informed approaches, including PINNs, using governing equations and physical constraints.

4. Compare approaches in terms of accuracy, robustness, generalization, and computational complexity.

5. Evaluate the most promising approach for reduced-order modelling and embedded BMS feasibility.

Education/program/focus

Master's student (civilingenjör) in computer science, machine learning, applied mathematics, engineering physics, or a related field.

Good knowledge of Python and machine learning is important. Experience with PyTorch or TensorFlow is an advantage. Knowledge of scientific machine learning, numerical optimization, PINNs, or embedded systems is beneficial.

Number of students: 1

Start date for the thesis work: To be agreed

Estimated time required: 20 weeks

Contact persons and supervisors 

Madeleine Ekström Unit Head [email protected]

Björn Bökelund Software Architect [email protected]

Prem Balaji Chinnadurai Industrial PhD [email protected]

Application: Your application must include a CV, personal letter and transcript of grades

A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.