Oak Ridge scientist describes Qualas model to predict grid‑scale battery aging

Shriantth Aloo, a computational scientist at Oak Ridge National Laboratory, described Qualas, a new foundational AI model intended to predict how grid‑scale lithium‑ion battery systems age by modeling both individual cell degradation and interactions among cells across the pack.

Aloo said the difference from prior battery‑health systems is procedural: "the previous battery health systems use single cell to make their predictions and based on their single cell predictions they were trying to predict the system level degradations whereas we are trying to focus holistically not just at the cell level on how the cell degrades but also how the cells are connected and how these interactions between cells aggregate towards a system level performance or degradation." The approach is meant to capture how cells affect one another inside a deployed system rather than extrapolating from isolated cells.

He emphasized why that matters for grid applications: large battery systems are used for services such as frequency regulation and energy arbitrage and undergo different load cycles that change how they age. "You can't really do that for systems that cost millions of dollars that are deployed at the grid level," Aloo said, arguing that better system‑level forecasting helps avoid premature, costly replacements.

The interview also tied the work to federal efforts: the Department of Energy's Genesis mission connects all 17 national laboratories to accelerate discovery using AI. Asked how the project contributes, Aloo said the Qualas effort aligns with the Rapid Operational Validation initiative's goal to compress long‑range performance forecasting. "Our mission has been to predict the performance of these battery systems 15 20 years out using one year worth of operational data," he said, adding the foundational model is intended to accelerate degradation predictions that previously took weeks or months so they can be delivered within hours.

The project, as described, aims to produce faster, system‑level degradation forecasts that could inform maintenance, asset management and procurement decisions for large‑scale storage deployments. The interview did not provide details on datasets, commercial partners, specific grid projects using Qualas, or a timeline for public release of the model.

The discussion ended with the team framing the work as a step toward shorter‑turnaround, AI‑driven battery health assessments that could be used across national laboratories and in industry collaborations; no formal policy or procurement decisions were announced during the interview.