Battery Intelligence

In this lecture, I will provide an update on the latest developments in digital models for the accelerated optimization of battery manufacturing processes. These models are based on novel physics-aware AI, able to not only predict properties as a function of manufacturing process parameters but also to explain the physicochemical mechanisms involved in the process. This work gives rise to the first-of-its-kind digital brain of production digital twins.

Join us as we delve into the innovative use of Time-Delayed Recurrent Neural Networks (TD-RNN) for lithium-ion battery state estimation and health prognostics. Our research demonstrates that TD-RNNs significantly enhance SOC accuracy by mitigating overfitting through the identification of "overexcited" neurons. We will also present an AI-driven software demo that empowers non-experts to perform battery analysis, bridging the gap between complex AI technologies and practical applications. Attendees will gain insights into advanced AI techniques and their transformative potential in battery management.

This talk will discuss recent efforts in long-term testing and methodology development for state-of-health (SOH) forecasting, led by a team from the University of Connecticut, Zurich University of Applied Science, and Delft University of Technology. Forecasting methods will be demonstrated using two long-term aging datasets from commercial-grade LCO and LFP cells. Notably, the LFP cells were cycle-aged under second-life application conditions, providing benchmarks for forecasting under repurposing scenarios.

Injectable phase change electrolytes allow for scalable manufacturing of solid and semi-solid state batteries via an in situ liquid to solid transition using standard injection and formation equipment. However, these multi-component systems must have balanced mechanical, electrical, and transport properties to optimize both material level characteristics and cell level performance. Here, an accelerated learning approach leveraging machine-learning driven optimization is presented to showcase how injectable phase change electrolytes can be rapidly optimized to meet commercially relevant battery performance needs.

This presentation explores how modern AI methods - Classical ML, Deep Learning, and Agentic AI - can accelerate battery R&D by shifting from experiment-first to optimization-driven, simulation-based design. By using targeted physics simulations to replace large portions of trial-and-error testing, engineers can rapidly explore design spaces, evaluate constraints, and converge on optimal cell architectures or charge protocols. We will also highlight how Agentic AI systems can autonomously run optimization loops, enabling faster, more efficient development of next-generation battery technologies.

This talk will provide a high-level exploration of some of the key challenges and opportunities in the model-based control of both lithium-ion and lithium-sulfur batteries, including both solid and liquid electrolyte Li-S batteries. Much of the talk will focus on emerging opportunities for test trajectory optimization and machine learning for both battery types, with a focus on Li-S batteries.

A predictive pretraining transformer (PPT) framework is introduced to enable scalable and generalizable battery diagnostics across diverse chemistries and operating conditions. The approach leverages supervised pretraining on large-scale time-series data, followed by task-specific fine-tuning for accurate estimation of state-of-health (SOH). Demonstrated across various battery platforms, the model achieves high accuracy with significantly reduced data and training cost, highlighting its potential for deployment in real-world electric mobility and energy storage applications.

Li-metal batteries offer exceptional energy density but face safety and cycle life challenges. GBatteries’ Intelligent Battery Management System uses adaptive pulse-based control to enhance performance, reduce degradation, and enable prediction, detection, and prevention of safety events. Validated in drone and aerospace applications, it improves runtime by up to 63%. This session explores how intelligent control accelerates the safe adoption of Li-metal batteries for electric mobility.

Advanced BMS plays a critical role in ensuring safe, reliable, and efficient battery operation. While Electrochemical Impedance Spectroscopy (EIS) holds great promise for detailed battery diagnostics, its practical application in real-time systems has not yet been fully realized due to sensitivity to operating conditions. This study presents strategies to overcome these challenges using compensation techniques and AI models for SoH prediction. It also explores core temperature detection for early thermal runaway prevention and accurate SoC estimation under load, demonstrating how EIS can be effectively integrated into intelligent BMS for improved battery monitoring in dynamic environments.

As an MIT study recently highlighted, the hype around AI has failed to materialize measurable returns, with a 95% failure rate. We’ll review a range of the battery industry's ambitious claims and aspirations around AI applications, including proof-of-concepts that are failing to deliver. We’ll then lay out the data building blocks needed to enable scalable AI. Finally we’ll show real-world battery AI delivering measurable returns in production with customers.

This study presents an advanced strategy for state-of-charge (SOC) and state-of-health (SOH) estimation that has achieved errors of less than 1%. This strategy includes combined spectral and temporal characterization of cells. It uses the Smooth Variable Structure Filter together with the Interacting Multiple Model concept for estimation.