Enabling collection of data – from lab to field – to forecast battery ageing

Licentiate thesis, 2024

By 2030, the battery industry is predicted to have an annual output of over 4 TWh of storage capacity, thus supporting the ”electrify everything” paradigm. This comes with societal benefits such as lower CO2-emissions in the end applic- ations, but also with risks of over-consumption and too high life-cycle emissions. Thus, making sure that batteries last longer by mitigating degradation is paramount. Here we explore how mechanistic understanding of degradation can be gained and effectively communicated for use in a battery digital twin (BDT), in it’s turn capable of predicting future battery degradation and suggest mitigating actions.

This is achieved by conducting a campaign of lab-based experiments to identify relevant degradation modes. Simultaneously a battery information format (BIF) suited for field application is developed within the scope of this thesis. The BIF is framed so as to respect both qualitative and quantitat- ive needs existing in battery powered applications and those from a broader informatics perspective. The knowledge about degradation modes are success- ively encoded into a graphical causal model (GCM) that is consequently used to define the binning scheme of the BIF-compatible histograms, enabling a memory-efficient data-collection strategy.

The method is showcased for a Hard Carbon and Na3V2(PO4)2F3 sodium- ion battery, utilizing 1 M NaPF6 as electrolyte, and identifying degradation using electrochemical techniques (e.g. galvanostatic cycling and electrical impedance spectroscopy), coupled gas-chromatography/mass spectrometry, in- frared spectroscopy, and EDX analysis. Subsequently the mentioned translation into a BIF-compatible data collection strategy is carried out. For next steps we suggest doing a field deployment of the strategy, either with the investigated chemistry or with commercial cells of a similar chemistry, so as to validate the BIF concept.