Most existing methods for battery state-of-health (SOH) forecasting have been applied to battery cell data from laboratory operation for training and testing. This presentation goes beyond that by using battery pack data from real-world vehicle operation. We aim to provide different feature sets that are accessible to all users of the SOH forecasting model. Two use- cases are exemplary illustrations of how the SOH forecasting model can be applied.

Herein, we establish a battery safety risk classification modeling framework based on a machine learning algorithm that can accurately and rapidly classify the potential safety risk level. The model can identify defective cells, cells with internal short circuits (ISCs), and cells with possible thermal runaway (TR) using a small portion of one-cycle data. Classifiers in the ML model demonstrate satisfactory performance and robustness.

One use case within AVLs Battery Life Cycle Management is the Battery Lifetime Prediction for Vehicle Fleets. Combining battery know-how and machine learning methods such as Long Short-Term Memory and Federated Learning the estimation of the state of health of a battery can be improved and even a prediction of future behavior is possible. The AVL Analytics Platform enables to make use of all data from vehicle fleets.

The novel BMS concept of Smart Battery for transportation and grid storage with improved safety lifetime is introduced. The key feature is the bypass device, capable of cell-level load management with minimized load impact leading to pulsed current operation which is known to slow down degradation processes. An advanced AI-based lifetime optimizer capable of online training is recognizing the early signs of stress and inserts optimized relaxation time.

Reliable and accurate degradation prediction remains 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, we will introduce our current work in battery ageing trajectory prediction with machine learning with case studies of both testing data in the laboratory and large-scale field data from 60 electric vehicles.

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

Increasing the range and fast charging capabilities of EVs are key challenges for the automotive industry. However, lithium-ion batteries with high energy density electrodes and low porosities suffer from increased impedance and higher risks of lithium plating. In this work we use detailed electrochemical simulations to determine the impedance of graphite electrodes using different perforation patterns. This is realized by performing virtual impedance measurements on symmetric cells.

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.

In this work, we will present our work in fast charging algorithm design based on monitoring and controlling the internal physical states using electrochemical models, control algorithms and machine learning.

This study presents a comprehensive review of the latest developments and technologies in battery characterization and their application in Battery Management Systems (BMS) for Electric Vehicles (EV).

Optimization of lifetime performance, early identification of anomalies during production, and description of the health of the fleet are just some of the imperative insights that a battery analytics solution should provide. Enterprise Battery Intelligence (EBI) leverages best-in-class battery data analytics to provide Automotive OEMs with a digital thread across their battery lifecycle. In this talk, we will discuss why this digital thread is business-critical and will provide use cases from the automotive industry.

Using AI and ML for lithium-ion battery modelling while answering questions regarding the predictability and accuracy of the models at various at stages of manufacturing, leave the manufacturer with a black box model hard to explain. The XAI and XML have a significant roll in shedding some light on the model by explaining how the decision is made and how the values are predicted.

Transforming battery data into operational decisions requires not only developing accurate machine learning (ML) models but also understanding how these models make predictions in the context of expert knowledge. In this contribution, we present our recent work on explainable machine learning applied to battery research, with an outlook on how semantic technologies – e.g., ontologies – can help to ingest the deluge of inputs, outputs, models, and patterns into actionable battery intelligence.

Here, we develop a machine learning diagnostic framework that unveils and quantifies rapidly the degradation mechanisms of batteries aged under various conditions in 0.012 s. Rather than performing extensive aging experiments, synthetic aging datasets for network training are generated.

In this lecture I will discuss a digital twin developed in the context of the ERC-funded ARTISTIC project which gives the promise to accelerate the optimization of the manufacturing process of Lithium Ion and Next Generation Batteries. This digital twin combines physics-based process models and artificial intelligence to predict the influence of manufacturing parameters on the battery electrode and cell properties. The digital twin also uses Bayesian Optimization to predict which manufacturing parameters to adopt in order to prepare battery electrodes and cells with optimal properties (inverse design).