How AI is helping with battery development

Batteries are becoming increasingly important as energy storage devices. Without them, there would be no electromobility, no smartphones and no pacemakers. However, conventional lithium batteries will no longer be able to fulfil the efficiency and capacity requirements in the future. Researchers are therefore relying on artificial intelligence (AI) to accelerate battery development.

Renewable energy from the sun, wind and water is available in almost unlimited quantities. They enable the global energy transition as well as electromobility and the transition to a more sustainable economy overall. However, as energy generation from natural sources depends on factors that cannot be reliably controlled, the development of efficient and powerful batteries as energy storage systems is of particular importance.

Rapidly growing demand requires new solutions

The importance of developing new energy storage systems is demonstrated by an analysis conducted by RWTH Aachen University in autumn 2023, which puts the rapidly growing global demand for batteries at 5,000 gigawatt hours by 2030, which corresponds to annual growth of 34 percent. The main drivers of rising demand are electric vehicles, stationary energy storage systems and portable electronics.

Lithium-based batteries have mainly been used for mobile applications to date. However, this raw material could become the limiting factor in electrification, as demand is growing faster than supply. In a market forecast, the German Mineral Resources Agency (DERA) concludes that ‘there could be massive supply bottlenecks in 2030’.

Even apart from availability bottlenecks, lithium is problematic. For example, extraction from salt lakes in South America leads to regional water shortages and the resulting social conflicts. In addition, the recovery of lithium from the recycling of used lithium-ion batteries is still at the development stage. Although the first promising processes are available, they do not yet recover any significant quantities.

The predicted increase in demand for lithium and simultaneous shortage of supply will drive up prices both for the raw material and for all products that use lithium batteries. This in turn would jeopardise the transition to electromobility, for example, which is already faltering because electric cars are more expensive than combustion engines.

The scientists hope that accelerating the development of new generations of batteries - with and without lithium - will provide a solution. They are utilising the possibilities of artificial intelligence in a variety of ways.

Which technology will inherit lithium-ion batteries?

The development of new batteries has been in full swing for years, but new lithium-free batteries for mobile use will still take some time before they are ready for the market. Researchers are therefore also working on optimising the production of established lithium-ion batteries. Although these batteries have been in commercial use since 1991, there is still a lot of potential for optimisation in their production.

Researchers at the Chair of Production Engineering of E-Mobility Components at RWTH Aachen University have identified around 2,100 cause-and-effect relationships that can reduce cell quality in the production of lithium batteries. Even small deviations in electrode production, cell construction and cell finalisation can have a massive impact and lead to rejection rates of over 10 percent.

With the help of AI data analyses, the error rate should be significantly reduced in the future. Specialised AI applications carry out automated root cause analyses to identify the causes of quality deviations. In addition, AI applications monitor the condition of the machines in the production line. This allows problems to be recognised at an early stage and avoided through intelligent, proactive maintenance planning. Overall, these AI approaches promise significant increases in efficiency, quality improvements and cost savings that can be realised in the short term.

At the Pacific Northwest National Laboratory (PNNL), a research facility of the US Department of Energy, a battery research project recently caused a stir. Together with the software company Microsoft, the researchers used AI to search for suitable materials for new batteries that use significantly less lithium without losing the advantages of the ultra-light metal. A total of 32 million potential materials were analysed.

The enormous capabilities of the AI were evident not only in the amount of data to be processed, but also in the speed: after just 80 hours, 18 materials were identified that can now be used for further research. Normally, this process in research takes several years or even decades. At PNNL, however, an initial prototype has already been developed that uses 70 per cent less lithium.

Another starting point for the sensible use of AI in battery development is analysing usage to improve service life. Prof Ralf Herbrich is researching this at the Hasso Plattner Institute (HPI) in the field of ‘AI and Sustainability’ in cooperation with a Berlin start-up. Together, they are developing algorithms that can record various wear factors in batteries and draw conclusions for optimal utilisation. In the case of batteries in electric cars, for example, these factors include driving style, the characteristics of charging processes and temperature windows during charging.

The aim of this work is to physically understand ageing processes without having to open the battery. The start-up ‘betteries’ then wants to use the findings to give used batteries a second life after they have been removed from the electric car. This would also improve the carbon footprint of the batteries, as around 500 charging cycles with renewable energy are currently required to save as much CO2 as was previously released during production. A longer service life would therefore be desirable. In any case, the research project is optimistic that the number of charging cycles can be at least doubled, tripled or even increased tenfold with AI support.

Science will not become superfluous - it will be accelerated

In the medium term, however, new generations of batteries will be needed that do not require lithium and other scarce raw materials. Until now, however, the research and development of such batteries has been expensive and time-consuming. The sharp rise in demand is now pushing several boundaries at once. On the one hand, lithium prices will continue to rise, making alternative solutions more attractive. At the same time, it is already foreseeable that the demand for energy storage systems will no longer be covered by lithium in the coming years.

For research, this will most likely also mean a shift in priorities if industry contributes to the funding. So far, this has only been insufficiently the case. Success stories from the USA about the involvement of AI in research could provide additional persuasion here.

Conclusion

The combination of different AI systems can trigger the urgently needed development boost for new batteries. Their rapid development to market maturity is essential for solving the numerous challenges associated with energy issues. In future, the coupling of AI systems with quantum computers could provide further acceleration: Proactively recognising and avoiding problems instead of having to react reactively to challenges.

Text: Falk Hedemann