Understanding how lithium ion batteries degrade and predicting when a battery will fail is an open research challenge. This project is addressing that challenge by advancing the latest mathematical models related to battery ageing, combined with state-of-the art statistical learning approaches to evaluate the model predictions and ageing mechanisms. The developed methodology will be invaluable for evaluating battery pack warranties for the UK's automotive sector, including project collaborators Jaguar Land Rover and Aston Martin, and will support the UK's electrification agenda in the longer term.

Header image credit: emirhankaramuk

Explaining the science

The state-of-the-art modelling theory for battery degradation accounts for variations of the electrode porosity, due to a precipitation known as Solid Electrolyte Interface (SEI) and subsequent Lithium-plating on the negative electrode. Under nominal operating conditions, this theoretical approach (SEI and Li-plating) can predict a linear fade of a battery’s utilisable capacity and its sudden decrease during cycling. Understanding and predicting this feature, also known as the ‘knee-point’ or ‘nonlinear capacity fade’ effect, is currently at the forefront of research within the automotive sector. This project will build on this theoretical foundation to combine with statistical learning methodologies.

Project aims

Several challenging objectives are anticipated to be achieved. Understanding and modelling the relevant physics that drives battery degradation is a primary objective. The physics will inform how the remaining energy of a battery gradually diminishes with use followed by a sudden decrease of the energy. Known as the 'knee-point', statistical learning of its underlying mechanisms and predicting when it will occur is another key objective of the project. This will require advances in physics-informed machine learning in order to combine traditional mechanistic approaches with Bayesian statistical learning for an overall semi-parametric framework.

Demonstrating and validating the developed methods on real battery ageing data, that will be performed at WMG University of Warwick, to predict the battery life-time will be a key success. The dissemination of the methodologies, through publications and conferences, and it's uptake from industry and other academic groups will also signify key success of the project.

The theoretical developments and implementations will benefit wider academic groups from both engineering and computer science working in both theory and application of dynamical systems. Combining the strengths of predictive battery modelling together with the latest statistical learning approaches also offers an attractive solution to address a pressing challenge in the automotive sector. Reliably predicting when a battery 'knee-point' can occur will render major opportunities to de-risk the sector's warranty and battery pack design phase.


The immediate application sector is the UK automotive sector. The project is working closely with Jaguar Land Rover and Aston Martin to understand their battery usage and warranty requirements. The developed methodologies will support evaluating the lifetime of batteries. It is anticipated that the approaches can in the long-term be adapted to new battery technologies and support other electrification sectors within the UK's de-carbonisation agenda, notably in relation with the Future of Mobility and Clean Growth Grand Challenges. 

Recent updates

March 2021


Selcuk Atalay et al. 2020, "Theory of battery ageing in a lithium-ion battery: Capacity fade, nonlinear ageing and lifetime prediction"

Daniel Tait et al 2020, "Scalable Multitask Latent Force Models with", Workshop on machine learning for engineering modeling, simulation and design, NeurIPS 2020

Daniel’s talk at NeurIPS 2020

May 2019

Developed initial electrochemical model with non-linear capacity fade.

April 2019

Meeting with Prof. Tony Hey (Chief Data Scientist at STFC) and presentation at STFC of the general approach.


Researchers and collaborators

Dr Liuying Li

Research Fellow, Warwick Manufacturing Group, University of Warwick

Previous contributors

Contact info

[email protected]



Support from HVM Catapult, European Union’s Horizon 2020 research and innovation programme grant agreement No 685716 and the Faraday Institution [EP/S003053/1] grant number FIRG003


Catapult EU Faraday