WHITE PAPERMachine Learning for Battery Optimization

The future widespread adoption of electric vehicles will result in reduced harmful emissions andcleaner air, among other social and economic benefits. The battery industry is in need of softwaresolutions for battery manufacturers to reduce fabrication and development costs while improving keybattery metrics. AI is unlocking battery technology that will power the future of clean transport,causing a shift in the automotive industry. However, charging capability, energy density and costs willneed to improve dramatically. AI has the potential to impact battery development and understand therelationship between data and battery parameters. We highlight the current areas where AI will makethe most dramatic changes along with machine learning approaches that are being applied tobattery systems.

Executive Summary

It is no surprise that efforts are being made to optimize battery systems with artificialintelligence (AI). Demand for electrification of transport has surged in recent years, due toincreasing concerns about global warming. The stakes for the global battery market have beenconsistently rising, with some predictions estimating that it will be work over £250 billion peryear from 2025. This will not only lead to an unprecendented demand in batteries, but will alsocreate millions of jobs. The performance, cost and safety of batteries determine the successfuldevelopment of electric vehicles (EVs) and currently, Lithium-ion (Li-ion) batteries are thepreferred choice for EVs due to their cycle life and reasonable energy density, followed by arange of different battery chemistries including Lithium-sulfur (Li-S), Sodium-ion (SIB), andSodium-nickel chloride (Na-NiCl2) batteries. However, further research of battery chemistrieswill result in more complicated battery dynamics, where safety and efficiency will become aconcern. Therefore, an advanced battery management system (BMS) that can optimize andmonitor safety is crucial for the electrification of vehicles.

The Rise of Machine Learning for Battery R&D Machine learning (ML) algorithms have been implemented to predict state of health, state ofcharge and remaining useful life. Data-Driven models have drawn attention in recent years, andcombined with ML techniques, these models appear to be more powerful and able to predictwithout a priori knowledge of the system and have the potential to achieve high accuracy withlow computational cost. With the reduced costs of data storage devices and advancement ofcomputational technologies, data-driven machine learning seems to be the most promisingapproach for advanced battery modelling in the future. The development of new battery technologies is slow, taking many years both in terms of newtechnology development and testing, and, more importantly, the substantial manufacturingchallenges needed to scale up for mass production. The future widespread adoption of electricvehicles will result in reduced harmful emissions and cleaner air, among other social andeconomic benefits. Technological advances will be required to make batteries more efficientand deliver significant environmental benefits enabling the growth and widespread adoption ofelectric cars. Enhancing and optimizing chemical and physical properties of batteries are bothdifficult tasks. Therefore, artificial intelligence will be an essential tool in pushing the boundariesof the current make-up and management systems of batteries, while reducing costs andenvironmental impact at the same time.

Introduction

1

Design of more efficient materials and chemistries within the battery itself

Reduce reliance on toxic materials 

Optimize the flow of current between the cathode and the anode

Efficiently scaling up of lab processes into industrial environments

Battery pack design - understand the trade-offs between size, weight, power,

charge, speed, and lifetime

Control systems - optimization of lithium battery management systems,

including state of charge and safety monitoring

The main aspect of AI currently being applied in the battery industry is machine learning. Thisapproach has been used to solve many high-value problems and the key variables for itssuccessful implementation are both data availability and data quality. Nevertheless, there hasbeen a recent surge in applying machine learning methods to help optimize different aspects ofthe battery industry. In both cases, the usage of data from multiple domains, including datafrom experiments that have failed, play a crucial role in accelerating and optimizing batterydesign, chemistry and management systems. Highlighting how machine learning can accuratelypredict and improve the health and life of a battery will enable manufacturers to embed thissoftware straight into their battery devices and improve their services. The most important uses of both machine learning in battery industry are thefollowing:

1.

2.

3.

4.

5.

6.

Areas for Battery Optimization

2

2 Battery Pack Design/ Geometry

3 Battery Management Systems

3

Machine Learning for Battery Chemistry, Pack Design andManagement Systems

1 Battery Chemistry

Batteries have several key parameters, including voltage, temperature, and state of change.Battery malfunctions are associated with abnormal fluctuations in these parameters, thereforeaccurately predicting them is crucial to ensure that electric vehicles operate reliably and safelyover time. Once in place, predictive models can be used to standardize processes, allowing allstakeholders access to the same knowledge and tools, and reduce costs both in terms of thenumber of experiments that need to be performed and optimizing experiments to minimize theneed for expensive components or processes. This results in reduced environmental impact bydesigning experiments and products that are less dependent on toxic elements or processes.

Intellegens has developed a machine learning tool, Alchemite™, that trains models on allavailable data, no matter how sparse or noisy. We bring all the available data together and useunderlying correlations to accurately predict missing values and generate the most completemodels possible. Applying this novel method to the available historical and simulated data,enables organisations to identify opportunities for reducing costs and downtime, time savingsand overall performance improvements, through predictive maintenance and processoptimization.

Alchemite™ designs the best possible chemical characteristics of molecules that can prove tobe the basis of a next-generation battery material. Reduce reliance on toxic materials andoptimize the flow of current between the cathode and the anode.

Alchemite™ understands how changes in pack design and geometry impact overall batteryperformance and optimize the design to meet target requirements such as thermalmanagement, power density, and charging rates.

Alchemite™ predicts how many cyles a battery can have and how many years it will last beforeit starts degrading. Better lifecycle predictions can help identify why some batteries fail earlierthan others, thereby allowing safety to be monitored.

4

At Intellegens, we use deep learning technology to design better formulations that meet ourcustomers’ needs by running virtual experiments to narrow the search for new materials andreducing experimental costs by over 80%.  We reduce prototype costs and shorten the time tomarket, thereby accelerating discovery and standardizing the design process across anorganization.

Adopting Alchemite™ yields the following benefits: 1. Maximize material performance for multiple target propertiesGenerate new formulations that satisfy multiple targets and understand the effect that altering each material willhave.

2. Reduction of prototype costsA typical new material currently needs 20 prototypes before the final formulation and processing variables areselected. Alchemite™ cuts down the number of prototypes needed to fewer than five. Use confidence levels toguide where you need to test and thus shorten time to market and reduce prototype costs.

3. Reduction in the number of experiments by 90%Virtual experiments extract valuable information about correlations between parameters, saving time and money byeliminating research and development costs.

4. Minimization of expensive properties & reduction of environmental impactDesign a new formulation based on multiple requirements including cost and environmental impact.

5. Standardization of the design process across the companyApply the same process for each new project

The Benefits of Adopting Alchemite™ to Battery Optimization

Intellegens is using Alchemite™ to address a substantial gap in the market by developingsoftware solutions for battery manufacturers and researchers to reduce fabrication anddevelopment costs while improving key batteries metrics. We specialise in providing deeplearning solutions to optimize industrial processes and are now applying this to help create themost efficient batteries. The application of our technology is helping in a number of ways byreducing costs related to battery pack design, chemistry and management systems. We workwith data rather than personal scientific approaches, which reduces the number of prototypesneeded in testing and standardises the design process. Our technology works in three key areas: battery pack design and geometry, battery chemistryand battery management systems. We increase the efficiency of battery systems by efficientlymanaging the state of health and charge, predicting remaining useful life, and avoid complexand costly multi-phys modelling. We reduce the number of prototypes needed in testing, avoidunnecessary experimental costs and optimize battery chemistries using experimental data andguided DFT simulations. Intellegens is currently working on artificial intelligence solutions to predict optimum processparameters for complex interdependencies in the battery manufacturing process. We useartificial intelligence to reduce fabrication and development costs while improving key batterymetrics and for process parameter prediction. By removing technical and commercial barriers tocell manufacturure in the UK, the technology will advance battery metrics and significantlyreduce costs of trials and experimentation. Through systematic information data management,AI will de-risk scaling up innovative technologies across the battery manufacturing value chain,including cell materials and manufacturing processes.

Alchemite™: Real-World Applications

5

Intellegens has developed a unique artificial intelligence engine, Alchemite™ for training neuralnetworks from sparse, and noisy data, typical of real-world data. The technique was firstdeveloped at the University of Cambridge where it has been used to develop several superalloys,guide the design of new drugs and help optimise battery pack design. The tool is now beingused to solve a wide range of real world industrial process problems where rare but valuabledata can be used to improve real world industrial processes leading to reduced costs andenvironmental impact.  For more information on how we can help with batteries, please visithttps://intellegens.ai/batteries/ For more general information, please visit intellegens.ai.

Man-Fai Ng, Jin Zhao, Qingyu Yan, Gareth J. Conduit, Zhi Wei Seh (2020). Predicting the State of Charge and Health ofBatteries using Data-Driven Machine Learning. Nature Machine Intelligence, 2, 161-170. https://doi.org/10.1038/s42256-020-0156-7

1.

About Intellegens

6

References

@intellegensaiinfo@intellegens.ai

intellegens.ai

Click on the link below to subscribe to ourlatest news and upcoming events

https://intellegens.ai/subscribe

Want to learn more about how our AI technologycan be applied to your specific needs? Contactus to learn more at info@intellegens.ai