Sign up for NPL updates
Sign up for NPL updates

Receive regular emails from NPL to get a glimpse of our activities and see how our experts are informing and influencing scientific debate

  • Home
  • Case studies
  • Machine learning model trained for better batteries for transport
Case studies

Machine learning model trained for better batteries for transport

In the midst of the pandemic, NPL helped Faraday Battery gather data to train their predictive maintenance model. 

Case study

The challenge

Large passenger vehicles such as buses and trains must meet very high standards of reliability, since the cost of a breakdown is high. To support the move to electrification for mass transit, Faraday Battery is developing an innovative battery pack designed to work reliably for over 10 years. As part of this project, Faraday Battery is building a tool to monitor the battery and predict problems, so they can be fixed before they impact performance.  

The tool will collect and analyse data from sensors on the battery to spot impending problems. A simple example may be identifying an association between temperature increase and cell degradation. But often it will be complex combinations of signals that create a signature of the potential problem.    

Faraday Battery has developed a machine learning model, which can learn to spot problems from their data signatures and use these to predict failures. However, due to COVID-19 there was little opportunity to trial their batteries and gather the data needed to train the model, stalling their product development.  

The solution

Faraday Battery approached NPL due to its expertise in collecting and rationalising experimental data on electrode and cell performance. Faraday Battery needed to gather data on degradation of the cells, which could be used to train the model. However, due to restrictions imposed by COVID-19 it proved too challenging to coordinate suitable onsite experiments to meet a pressing deployment timescale. NPL helped progress the project by reviewing the academic literature available in order to understand what other experiments had discovered about battery cell degradation. This was used to build a library of data signatures for relevant problems, which was presented in a report.  

The impact

Armed with these data, Faraday Battery is working on training and developing their predictive maintenance model. This provides a critical step towards commercialisation at a time when such work would have been very difficult without NPL’s support. They still need to augment this with experimental data from their own batteries – which they hope to explore with NPL next – but they will begin this from a much more advanced position, thanks to this initial project.  

Once work is complete, which they estimate will be by the end of 2021, the resulting tool will allow battery problems to be predicted, with 90% fixed remotely and the remaining 10% addressed manually before they have consequences.    

This will be a vital part of their ‘battery pack’, which they plan to launch by the end of 2021. The result will be a battery for electric trains and buses which guarantees reliable power, and avoids the high cost of taking them out of service for unplanned maintenance. This will allow Faraday Battery to launch into a market which they project to be over $100 Billion worldwide. 

What the customer says

Quality, well-understood data is critical for our needs and NPL is the best place to go for this type of measurement data. To have the project funded was especially valuable when faced with the challenges of the pandemic

Sanjay Gupta, Founder - Faraday Battery

Faraday Battery has created an innovative battery pack for electric trucks, buses, vans and cars. It is unique in that it performs as promised regardless of the weather, and can last more than 10 years, thanks to innovative thermal management and proprietary electronics. They hope it will change transportation as we know it.  

Find out more