An AI learns how a battery is actually discharged or charged

02/23/2023

Source: Energy & Management Powernews

The battery company Varta is working with partners on a smart battery management. It aims to use AI to give home storage devices a longer life.

The aging of battery storage devices depends heavily on the number and depth of charge cycles, but also on many other factors, such as charging and discharging power and ambient temperature. Since both the processes in the battery and user behavior are very complex, aging is difficult to estimate using conventional methods and must be considered individually. This is where the "Longer" research project comes in, in which Varta Storage GmbH is collaborating with partners from industry and research.

As part of the project, the partners want to use artificial intelligence (AI) in the form of machine learning to develop models of user behavior and battery aging. To this end, the software programs will be implemented on field test devices of the "Varta.wall" home storage system. "Today, home battery storage typically completes 200 to 300 full charge cycles per year and is used almost exclusively to store solar power," explains Benjamin Achzet, Research Coordinator at Varta Storage. In the future, home storage will additionally act as an "electricity trader" and thus further reduce energy costs and actively relieve the power grid. However, storage systems with higher cycle stability are necessary for this.

The AI is to precisely analyze load profiles as part of the project. It will also learn how the battery should actually be discharged or charged in a given situation in order to operate efficiently over the long term. Jens Haupt, a battery aging specialist at the company Novum Engineering, explains, "In the field test, the AI learns independently how a certain load profile affects the state of the battery, the State of Health. Over time, it can also predict the State of Health. This allows for predictive control in the next step." Such AI-based battery management, he said, can make optimal use of the battery's capacity while maximizing its lifespan.

The "Longer" project will also use simulation models of the storage systems to test and validate the AI-based operating strategy in a targeted and efficient manner. Efficient simulation methods or digital twins allow validation not only with a reasonable amount of time, but also the testing of borderline scenarios that are costly to realize in the laboratory.

Instead of following fixed rules, the AI is to decide on the best strategy depending on the application. The AI already knows typical yield and consumption patterns and continuously refines them through machine learning. The vision of the partners: depending on personal preferences

- maximizes the AI's own consumption,
- minimizes CO2 emissions
- or optimizes the energy and battery management economics of a home with home storage.

Varta intends to use the technology to continuously improve its storage devices, with the perspective of offering a storage device with a lifetime warranty. In addition to Varta and Novum Engineering, TWT GmbH Science & Innovation is also one of the partners. They receive scientific support from the Fraunhofer Institute for Solar Energy Systems (Ise).

Author: Davina Spohn