While the industry is familiar with the battery and its battery management system (BMS), very few are aware of the critical need for a missing third layer, the Enterprise Battery Intelligence (EBI) system. The EBI is needed to unlock the significant advances in battery yield, energy density, and lifetime that the industry is calling for. Historically, product OEMs have treated batteries like black boxes, building mechanical and electrical interfaces to keep them stable. As batteries now become the make-or-break component in low-cost EVs and long-lived consumer electronics, companies need the EBI to provide a new level of insight and ensure that batteries are performant, reliable, and safe.

One of the most challenging aspects of improving lithium-ion battery lifetime is the presence of "knees", or the sudden loss of capacity, power, or energy with cycling. Here, I summarize the work of an international collaboration to classify observed and proposed knee mechanisms from the literature. We find that some mechanisms can be predicted from electrochemical signals, while others cannot. This work informs strategies for battery lifetime prediction.

Successful commercial innovation in battery technology is uniquely difficult because of interrelated and coupled challenges stemming from the atomic scale all the way through GWh-level manufacturing. Cuberg, as part of Northvolt, has pioneered a vertically integrated approach towards battery innovation that combines excellence in simulation, data science, high-throughput testing, cell design, process development, supply chain, and customer distribution to deliver the world's first lithium metal battery at commercial scale. 

This presentation covers software and data collection for cell formation, aging, and grading on production lines. A tuned solution for formation lines can be based on Industry 4.0 technologies, thus providing a system with flexibility and agility that securely manages processes, data collection, and storage. This data is an important battery intelligence data source to track cell provenance and history as from cell development through manufacturing and eventually into batteries deployed in the field.

AI/ML cycle life prediction is emerging as the broadly transformational capability; initial focus has been on urgent transportation needs.   Argonne’s expanded scope includes diverse stationary markets, where useful life extends beyond 80% capacity and use-protocols are complex and changeable (e.g., stacking or role changes during asset life) and requires alternative research strategies. This talk compares results with various ML approaches and with the use of both experimental and synthetic data.

Growing numbers of electric propulsion for advanced performance (auto)motive applications drive the need for battery condition and health monitoring of temperature, voltage and current in and outside the pack as well as active cell balancing. PhotonFirst develops fiber optic sensing systems based on photonic integrated circuits (PICs) enabling low cost, robust, low weight and scalable solutions. Paul van Wijk will share how the company unlocks this technology for your application.

The field of battery development and manufacturing is full of opportunities for the application of machine learning. Machine learning techniques have accelerated materials discovery at the fundamental atomic scale, and have also impacted the commercial and manufacturing scale, accelerating failure predictions. In this talk, we will be covering some case studies of impactful machine learning applications in the battery field, spanning these time, length, and cost dimensions.

Battery Analytics gives customers control over batteries along the lifecycle: development, for optimized battery design; in-life, for optimized fleet management including maintenance, and in second life scenarios: reselling or repurposing in new use cases. Drawing on customer success stories, we will present arguments for deploying battery analytics in electric vehicle fleets. This holistic approach to lifecycle management enables lower costs, improved quality and a sustainable contribution to renewable energy.

New machine learning (ML) approaches offer a route to accelerated materials discovery by training predictive models on existing experimental data and then using these models to screen databases of candidate materials. In this talk, we present our research in using ML to accelerate electrode and electrolyte discovery, discuss best practices for the application of ML to materials design, and highlight the Aionics materials design software platform.

GBatteries will present an innovative and compelling way to charge unaltered Li-ion batteries, powered by machine learning. One which reduces irreversible chemical reactions and allows for ultra-fast charging. To increase the adoption of EVs globally and drastically reduce greenhouse gas emissions, GBatteries has developed a charging algorithm that can ultra-fast charge lithium-ion batteries without compromising battery lifespan. No need to change the manufacturing process or improve the chemistry. GBatteries’ technology works with off-the-shelf batteries that have been produced by any of the leading manufacturers.

Lithium battery fires are becoming more common. ALGOLiON developed a software solution for the EV and other markets that provides the earliest detection of lithium battery fire hazards, warning of explosions days in advance instead of the seconds available now. AlgoShield provides a better ROI on your battery, and protects your products, the people who use them and your bottom line by lowering warranty costs and exposure to liabilities.

KULR will showcase its latest technology, CellCheck. CellCheck senses and analyzes a battery’s full life history, and adverse incidents including electrical, physical, and environmental events which impact battery health.  CellCheck is able to accurately predict battery health and safety throughout its lifecycle, delivering immediate assessments regarding hazards and risks, degradation of reliability and safety resulting from daily real-world use or abuse of energy dense batteries.

We introduce a data-driven prognostics framework to predict both capacity and power fade simultaneously with multi-task learning. The model is able to predict the degradation trajectory of both capacity and internal resistance together with knee-points and end-of-life points accurately with as little as 100 cycles. The stable prediction ability of the model facing capacity and resistance estimation errors further demonstrate the model’s robustness and generalizability. Compared with single-task learning models for forecasting the capacity and power degradation, the model shows a significant prediction accuracy improvement and reduces 50% of the total computational cost.

Accurate predictions of degradation and lifetime of lithium-ion batteries are essential for reliability, safety, and key to repurposing. Cycle life is a key performance metric for a battery management system since it can dynamically adjust operation limits based on their impact on lifetime 10-20 years ahead. A health-conscious power derating or a temporary stretch of power capability will be continuously adjusted depending on the short-term economics and long-term durability predictions. We show an adaptive inter-cycle extrapolation algorithm that allows us to simulate the entire lifetime of the battery in seconds for a real-time decision. The accelerated simulation allows us also to iteratively tune (learn) degradation parameters to match experimental observations of capacity fade, loss of lithium inventory, and individual electrode capacities (features) from both cycling and calendar aging.

We unveil our new robot, Clio, for optimizing nonaqueous battery electrolytes. A custom-built automated experiment named "Clio" is coupled to Dragonfly - a Bayesian optimization-based experiment planner. Clio autonomously optimizes electrolyte conductivity over a single-salt, ternary solvent design space. Using this workflow, we identify 6 fast-charging electrolytes in 2 work-days and 42 experiments (compared with 60 days using exhaustive search of the 1000 possible candidates, or 6 days assuming only 10% of candidates are evaluated). Our method finds the highest reported conductivity electrolyte in a design space heavily explored by previous literature, converging on a high-conductivity mixture that demonstrates subtle electrolyte chemical physics.

Using model-based control strategies, we have developed optimal charging protocols to minimize the capacity fade due to SEI-layer formation, lithium-plating and intercalation-induced stresses, while controlling internal temperatures inside the batteries. In collaboration with NREL, we have shown increase in cycle-life by more than 100%. We will present further results for faster charging and improved battery life on 2.2kWh battery modules.

Battery diagnosis and prognosis algorithms are critical to increase penetration of electrochemical storage systems. Current data-driven models are often limited by the non-representativity of the training data available. This work will showcase how synthetic big data training datasets can be used for transfer learning to solve this issue. This approach offers the benefits of the broad applicability to various cell chemistries, designs, and operating modes, as well as high fidelity.