Thesis: Machine Learning Based Estimation of Battery State of Health

30hp - Machine Learning Based Estimation of Battery State of Health Using Voltage, Temperature, Impedance, and Resistance Measurements Background and motivation Reliable estimation of battery state of health (SOH) is important for assessing battery performance, planning maintenance, and supporting decisions about continued use and replacement. This proposal focuses on lithium-ion batteries tested under controlled laboratory conditions. Battery testing produces electrical and thermal measurements that may contain useful information about degradation. These include voltage and current profiles, temperature measurements, internal resistance, and electrochemical impedance spectra. Machine learning offers a way to identify relationships between these measurements and battery health. Problem statement The proposed thesis will address the challenge of estimating battery SOH from test measurements collected under different operating and diagnostic conditions. A central difficulty is distinguishing degradation from reversible changes associated with temperature, state of charge (SOC), and measurement conditions. A model may perform well on familiar test conditions while producing inaccurate estimates for an unseen battery or temperature range. The study will therefore evaluate both estimation accuracy and generalization to batteries and test conditions excluded from model training. The research contribution will be an experimentally evaluated comparison of measurement combinations and machine learning methods, with particular attention to temperature effects, measurement requirements, and validation on unseen cells. Aim and objectives The aim is to develop and evaluate machine learning models that estimate battery SOH using laboratory test data, with emphasis on voltage, temperature, impedance, and resistance measurements. The Objectives are to:

• Prepare a consistent dataset linking diagnostic measurements, test conditions,

and reference capacity measurements.

• Identify electrical and thermal features or health indicators associated with

battery degradation.

• Compare selected machine learning methods against simple baseline models.
• Quantify the additional value of voltage, temperature, resistance, and impedance

information.

• Evaluate model performance on unseen cells and test conditions.
• Assess prediction uncertainty and the practical measurement requirements of the

resulting models. Research questions

• How accurately can battery SOH be estimated from voltage, temperature,

impedance, and resistance measurements?

• Does combining these measurements improve estimation compared with using

each measurement type separately?

• How do measurement temperature, SOC, and ageing conditions affect model

accuracy?

• How well do the models generalize to unseen batteries and operating conditions?
• Which measurements provide the best balance between estimation, accuracy

and diagnostic effort? Contact persons and supervisors Traton Supervisor: Venu Gopal Prathimala venu-gopal.prathimala@scania.com Application: Your application must include a CV, personal letter and transcript. A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.