Northvolt's machine learning R&D pioneers advancements in battery technology, employing sophisticated algorithms to optimise performance and enhance energy efficiency. The team focuses on pushing the boundaries of using software, machine vision, and machine learning applications to drive innovation in sustainable energy storage solutions. The talk will focus on glimpses of Northvolt's playbook to deploy technological advancements in Software and ML in Cell Design, Validation facilities, and on Shopfloor.

The concept of integrating physics-based and data-driven approaches has become popular for modeling energy systems. However, existing literature mainly focuses on data-driven surrogates generated to replace physics-based models. These models often trade accuracy for speed, but lack generalizability, adaptability, and interpretability inherent in physics-based models, which are qualities crucial for optimisation and control purposes. We propose a novel machine learning architecture, termed model-integrated neural networks, capable of learning the physics-based dynamics of general autonomous or non-autonomous systems consisting of partial differential algebraic equations. This architecture is then applied to the modelling of different batteries and electrode materials. 

In this talk, Sam will explain the various microstructural characterisation and analysis methods developed by his team, including some novel machine learning approaches. He will also propose a workflow for optimising the manufacturing parameters of these materials that use generative adversarial networks and Bayesian optimisation.

Here, we give an overview of recent methodological advancements of ML techniques for atomic-scale modeling and materials design. We review applications to materials for solid-state batteries, including electrodes, solid electrolytes, coatings, and the complex interfaces involved.

Reliable and accurate aging estimation and prediction remain challenging due to the nonlinear nature of lithium-ion batteries that stems from internal electrochemical reactions and intrinsic parameter variability across cells. In this talk, I will introduce our current work on battery aging estimation and prediction in field applications based on large datasets from electric vehicles and fast screening approaches for accelerating second-life applications of used batteries.

Smart Battery is a novel BMS concept that brings together cells with power electronics and AI for transportation and grid storage, with significant extended lifetime and high potential for second-lifetime application. The key feature is the bypass device for cell-level load management, allowing complete balancing of SoC, SoT, and SoH, along with square pulse excitation for online impedance measurement and fault-tolerant operation AI-based health and safety management.

Data evaluation of battery aging matrices is very time consuming as tests from different SoC, Temp, DoD and C-rates have to be evaluated. We show how this can be facilitated by online based automatic aging analysis which is also used for phys. & ML aging prediction to judge remaining battery life.

Automotive HV batteries are demanding a focussed effort on safety and failure prevention. Conventional methods for health monitoring fall short due to their supervised nature, relying on historical fault data. This presentation shows an innovative approach involving the implementation of an AI-based digital twin leveraging a graph neural network for unsupervised anomaly detection in fleet data. Furthermore, our approach incorporates domain knowledge to proactively prevent HV battery failure.

• Phase field fatigue modelling for describing particle fracture behaviour
• Continuum modelling explores dynamic stress distribution through an electrode 
• Composite graphite-silicon electrode degradation modelling
• Pack-level degradation effects caused by heterogeneous current distribution 
• Battery digital twins for next-generation control systems
  

Range anxiety is a key reason that consumers are reluctant to embrace electric vehicles (EVs). However, none of today’s EVs allow fast charging in cold or even cool temperatures due to the risk of lithium plating, the formation of metallic lithium that drastically reduces battery life and even results in safety hazards. Here, we present an approach that enables 15-minute fast charging of Li-ion batteries at any temperature (-50 °C).

The high cost and environmental impacts battery production have made the necessity of the application of second-life batteries now clearer than ever. However, one of the main challenges of handling the second-life cells is related to their re-characterisation after the first life, particularly when not enough information is shared from their first life use and storage. Here it will be discussed how the non-invasive experimental techniques such as electrochemical impedance spectroscopy (EIS) combined with machine learning (ML) techniques can support solving this challenge by accurately predicting the second-life battery’s state-of-health (SoH) rapidly.

The integration of data science into battery research opens up new approaches for enhancing battery safety and quality determination. This is demonstrated by current research projects that utilise AI and machine learning for defect detection. A key basis for these advancements is efficient scientific data management. The merging of data science and battery research represents a significant step forward, offering considerable potential for future innovations.

Li-ion batteries are exposed to a variety of degradation effects, causing cell aging over time. One major contributor to capacity and power fade of the cell is growth of the solid electrolyte interphase (SEI). In this talk we will discuss how the long-term growth behaviour of the SEI can be captured with fully coupled electrochemical simulations. Furthermore, we present numerical methods to enable long-term aging studies of such detailed models.

As the battery sector scales to meet rising global demand, companies spanning the battery ecosystem are increasingly understanding the centrality of data analytics to achieving business goals across the product lifecycle. From development to production to battery passports, the imperative to understand battery quality, performance, and health at every stage is clear. Rather than assemble a patchwork of siloed systems to meet these needs, companies that take an integrated, full-lifecycle approach to achieving electrochemical insights at scale will learn faster than the competition, serve customer needs better, and ultimately win in the marketplace.