Our Connected Battery Management System platform makes use of our Virtual Fleet technology and machine learning to help to predict when faults may occur even before there is actual data available from the physical vehicles.
We can use Virtual Fleet simulations to create input data for our machine learning algorithms for battery health estimation. This data can be used to establish which signals are most significant for monitoring and analytics when no historical fleet data is available.
To start with, we train a model using our Virtual Fleet tool, running simulations over realistic duty cycles. Then, as real data becomes available, we retrain the model based on “over-the-air” fleet data and service interval data.
The goal of the machine learning platform is to identify where batteries are aging excessively within the fleet, enabling online adaptation of the BMS calibration to maximise battery life. It can also predict anomalies that may result in battery failures and coordinate actions with the vehicle owners and service centres before they occur.
When physical fleet data becomes available, this Virtual Fleet can be used as a “digital twin” of the fleet and enables the development and testing of calibration updates, schedules, and strategies to maximise the life, capacity and operational efficiency of the batteries within the fleet. We can also use this technology to improve future battery designs.