Thesis “Machine learning and physical modelling interactions for vehicle batteries optimisation” (GREYDIENT project)¶
Project page: http://www.greydient.eu/
Francesco Cancelliere is supervised by Jean-Marc Bourinet and me.
Funded by the European Union through the ITN program.
Duration: 36 months, starting in September 2021.
Doctoral school enrolment: Ecole Doctorale Sciences pour l’Ingénieur, Clermont Ferrand, France.
Objectives¶
To develop a multi-physics numerical model to represent battery degradation and performance,
to implement this model in combination with real monitoring data in a FMI framework,
to develop a method to iteratively update the numerical model via Bayesian model updating and
to validate this method on a real electric vehicle battery pack case.
Expected results¶
A method for efficiently and effectively monitoring battery degradation and performance using an adaptive grey-box virtual twin approach, based on a FMI integration of a multi-physical model with online monitoring data.
Secondments¶
Politecnico di Milano: Active learning applied to virtual twin models with Piero Baraldi and Francesco Di Maio.
Leibniz University Hannover: Efficient methods for Bayesian data assimilation for battery models with Matteo Broggi and Michael Beer.
Subject¶
ESR 15 will develop an active-learning grey-box virtual twin approach for the monitoring of battery degradation and performance in electric vehicles. Specifically, ESR15 will combine multi-physical (physio-chemical and structural) differential algebraic equation representations of the underlying battery physics (e.g., via 0D/1D simulation in the Modelica language) with data coming from online battery monitoring systems via the Functional Mock-up Interface (FMI) framework. In this way, a highly flexible grey-box virtual twin can be designed. First, a pen-and-paper design of the battery model and monitoring structure is made, based on a mathematical formalisation of the problem. This theoretic development is then implemented in the FMI framework, where the theoretical DAE models are integrated with data. Furthermore, ESR15 will apply advanced sensitivity analysis techniques to actively determine both how the sensor lay-out, as well as the model representation can iteratively be improved. Finally, based on Bayesian data assimilation methods, the developed framework will ensure the accuracy of the DAE representation of the battery model automatically. The developments will be validated via a case study involving a real electric vehicle battery pack equipped with the necessary sensors.
