Description: Machine learning (ML) and advanced predictive statistical techniques are gaining widespread use in the field of electrical engineering as a whole, and for state-of-health modelling of Lithium-ion batteries in particular. The introduction of ML and statistics in electrical engineering is a consequence of the field slowly subsidising some of the more expensive laboratory testing by using data collected during real-life operating conditions. The upside of using ML and predictive statistics is that, in many instances, these methods can achieve an acceptable precision using a reduced amount of laboratory testing. However, it comes at the cost of added model complexity and the loss of some of the explanatory power when compared to the physics based state-of-health models.
This two-day course introduces key aspects of machine learning, predictive modelling, and model validation. Focusing on quantitative predictive models for Lithium-ion battery state-of-health modelling. The course will present an end-to-end framework from when data is gathered to a model has been created and put to use for state-of-health estimation. The models will include linear models, support vector regression, Gaussian process regression, and various neural network structures. The general aim of these methods is to predict capacity degradation based on a combination of laboratory and field data.
Exemplifications of some of the discussed topics will be made through exercises in R and Matlab.
Day 1: Lithium-ion batteries and ML-based feature extraction and reduction. Daniel-Ioan Stroe and Søren B. Vilsen; 7.4 hours
- Introduction to lithium-ion batteries and battery performance parameters for SOH
- Overview of machine learning methods, the bias-variance trade-off, and cross-validation.
- Feature extraction (manual extraction).
- Feature reduction through principal components analysis and multi-dimensional scaling.
Day 2: Machine Learning for battery SOH estimation. Daniel-Ioan Stroe and Søren B. Vilsen; 7.4 hours
- Linear models, and shrinkage methods.
- Kernel methods such as support vector regression and Gaussian process regression.
- Neural networks with a short introduction to DNN and RNN.
- Automatic feature extraction and reduction by using neural networks.
Prerequisites: Fundamental understanding of probability and statistics is recommended. Furthermore, basic knowledge of either R, Matlab, or python is strongly recommended.
Form of evaluation: Students are expected to solve several exercises and deliver an individual report with solutions and comments.
Course literature:
- X. Sui, S. He, S. B. Vilsen, J. Meng, R. Teodorescu, D.-I. Stroe, “A review of non-probabilistic machine learning-based state of health estimation techniques for Lithium-ion battery,” Applied Energy, Volume 300, 2021, 117346, https://doi.org/10.1016/j.apenergy.2021.117346.
- S. B. Vilsen and D. -I. Stroe, "Transfer Learning for Adapting Battery State-of-Health Estimation From Laboratory to Field Operation," in IEEE Access, vol. 10, pp. 26514-26528, 2022, doi: 10.1109/ACCESS.2022.3156657.
- Kevin P. Murphy, “Probabilistic Machine Learning: An Introduction,” The MIT Press, 2022
- T. Hastie, R. Tibshirani, J. Friedman, ”The Elements of Statistical Learning,” Springer Series in Statistics, 2nd edition, 2017
A complete list of references will be available one week prior to the course.
Organiser: Associate Professor Daniel-Ioan Stroe, dis@energy.aau.dk
Lecturers: Assistant Prof. Søren B. Vilsen (AAU-MATH)
Associate Professor, Daniel-Ioan Stroe (AAU-Energy)
ECTS: 2
Date/Time: 22-23 May 2023
Deadline 1 May 2023
Place: AAU Energy, Aalborg
Max no. of participants: 30
Price: 6000 DKK for PhD students outside of Denmark and 8000 DKK for the Industry excl. VAT
The Danish universities have entered into an agreement that allows PhD students at a Danish university (except Copenhagen Business School) the opportunity to free of charge take a subject-specific course at another Danish university.
Payment: A Online link will be annonced after deadline for registration
Important information concerning PhD courses: We have over some time experienced problems with no-show for both project and general courses. It has now reached a point where we are forced to take action. Therefore, the Doctoral School has decided to introduce a no-show fee of DKK 3.000 for each course where the student does not show up. Cancellations are accepted no later than 2 weeks before start of the course. Registered illness is of course an acceptable reason for not showing up on those days. Furthermore, all courses open for registration approximately four months before start. This can hopefully also provide new students a chance to register for courses during the year. We look forward to your registrations.
- Teacher: Daniel-Ioan Stroe
- Teacher: Søren Byg Vilsen