Artificial Intelligence in Battery Storage: "The challenge is always the data"

Expert Interview – October 23, 2023

Mandy Schipke, Co-Founder and CEO of the AI-Company NOVUM

Whether ChatGPT or digital art generators - the topic of artificial intelligence (AI) is on everyone's lips throughout the society. It also plays a major role in the energy industry in many areas - for example, in the use of battery storage systems. But what is really behind it and where is the added value? Mandy Schipke, founder and managing director of NOVUM engineering GmbH, reveals.

Artificial intelligence is on everyone's lips in all industries. Is it just hype, or does AI actually bring benefits to the energy storage industry, for example?

Both. It depends on where and how it is used. If all the historical data of a battery storage system is available, AI can be used, for example, to determine how well the battery is functioning, prevent fires, or display predictions about how long the storage system will last under a wide variety of conditions. But if something is wrong with the data or not enough information has been collected, using AI is quite risky, because what comes out is nonsense. Even worse, though, is that you don't know it's useless information because it's a kind of black box. Companies working with artificial intelligence definitely need very good strategies to deal with such data problems.

NOVUM has already been tackling these problems and challenges for nine years. What do you think is already possible today and where will the journey go?

If you have correct data, or know how to properly deal with a lack of data, for example, battery status or a security issue can be determined much better by AI than without it. This is due to the fact that artificial intelligence, unlike the human brain, has a greater capacity to keep track of all the influencing factors. It's a little different when it comes to predicting the life of a battery. While we also rely on AI here, we currently lack data on very old memories. The future will bring this AI input and thus also provide improved results in this area.

In your company, you have set your sights on making batteries more efficient, longer-lasting, safer and more sustainable. In which areas do you come up against limits and what solutions are available for this?

The challenge is always the data. In the best case, we have the "live" data of the batteries. In order to provide a picture of the battery's condition, we would need to know all the facts about, for example, the voltage, temperature and current throughout the life cycle of the storage device. By using artificial intelligence, we determine the status of a battery based on this enormous computing power and can thus make predictions about its remaining capacity. Second life batteries are more difficult to verify because they lack data history. We have now been able to solve this problem. Using a special measurement method, the battery is tested for 90 seconds and then we let an artificial intelligence work with this data. Here, we are moving into the terabyte range.

How does this method work, in which a small time window of 90 seconds replaces the entire lifetime of a data set?

If historical data of a battery can be accessed, a model is trained on this type of memory. It collects all factors around this medium and thus draws conclusions about the internal state of a battery. Our measurement method, on the other hand, looks inside the battery, so it is not necessary to calculate all the factors around it. This method is called electrochemical impedance spectroscopy.

Ideally, batteries and accumulators are recycled. What are the challenges associated with the recycling process and how can artificial intelligence be usefully applied here?

For one, only batteries that have truly reached the end of their useful life should be recycled. Currently, almost 90 percent of recycled batteries could easily be reused in other areas for up to 10 years. This is an immense waste of resources because until now there has been no method to test batteries quickly and easily. The standard process used to take between 5 and 10 hours. Secondly, AI can be used to simulate the use case in which the battery would have the longest life. And it would also be possible to check the second life battery to avoid any nasty surprises. So the challenge in recycling is to prevent most of the recycling.

You have advocated for a battery ID. Is it necessary despite the 90-second quick check?

The idea behind it is that each battery has its own ID document, in which important information can be found. For example, about the material composition or the state of health, which is very important for the Second Life application. Unfortunately, we will very likely not have a common understanding of what this battery passport should look like by 2025, nor is there a common definition of how these values are calculated. Having both would be very useful. The battery ID provides information through data and figures, the quick test provides a look into the inner workings and thus also the safety status of the battery.

What are your company's next innovations and where will your industry go from here?

At the moment, we are able to perfectly monitor stationary battery storage and thus make batteries more sustainable. In the near future, we will have fully automated energy and action decision-making through artificial intelligence. In this process, AI will decide how to operate a battery storage system in terms of its optimal effect between yield, battery life and safety. The trade-off between these three factors remains. Batteries should be used in a much more sustainable way. In a second or third life in another car, or perhaps in a final life as stationary storage. But that can only happen if high-quality information about a battery's condition is available. Novum's AI supports each of these steps in this chain.

This interview is an excerpt from an episode of The smarter E Podcast. You can listen to the full interview here .

You are using an outdated browser

The website cannot be viewed in this browser. Please open the website in an up-to-date browser such as Edge, Chrome, Firefox, or Safari.