A Battery Management startup in France and India has today launched its artificial intelligence cloud-based monitoring technology, called Edison.
The Mumbai based start-up has set a $1M revenue goal by 2020, aiming to sell over 50,000 BMSs and have upwards of 1GWh of batteries under management on the platform by that time. ION Energy designs and develops battery technology & software that battery pack makers, OEMs, and electric fleet managers leverage to build world-class batteries. At present the startup exports to 10+ countries including France, Poland, Sweden, Germany, Austria, South Africa, and North America.
It bought French BMS developer Freemans in 2017 to provide hardware and has now launched the Edison AI-based monitoring platform. “We are on the trajectory to become the world’s largest repository of battery data and the way that the licensing is working for us to be at that point in the next two years,” said Aryan.
“My personal background has been in the machine learning and AI space, building one of India’s AI companies that led in conversational AI. We raised $15m and 600 people and acquired by India’s largest telecoms company,” he said. “Since I was a kid I’ve been fascinated by electrons – the way I look at the world, is all the atoms that exist are a function of the electronics around the nucleus. But it was only in 2016 when I had the time and resources to explore business opportunities in the industry.”
“We focus on storage and consumption. The initial idea was to be a software first battery pack maker, but we realised that the software has to be deeply integrate with the brain of the system, the BMS. So we started building out our own BMS. In late 2016 we did a large comparison of other BMS providers and we identified a few companies in Europe that were much better than us.”
“One of these was Freemans in France, and I was able to sell my vision to my co-founder and his shareholders. We spent the next few months packaging the technology into commercialisable products and platforms and what we realised was that the BMS is being seen as the core differentiation in a battery pack. There’s no differentiation at the battery cell level. There’s the packaging of the cells but that’s a packaging s issue, not differentiation.
However battery pack makers not only want to buy a BMS but are developing their own BMS. “So we changed the business model completely to license access to the technology rather than sell the BMS as a product. Even though we do have a business line for battery management selling the BMS we have another that is also significant in licensing. That means we can focus on using software to extend the battery life.”
That battery management data is then used alongside a model of the battery pack, create as a digital twin, to monitor the performance and predict any problems. “This is probably one of the world’s most comprehensive battery pack modelling system so customers can build a virtual clone of the battery on the cloud and the telematics sends data on the entire system,” said Aryan. “This creates intelligent batteries that are configurable.”
“Once we model the cell we ask our customers to share the physical electromechanical assembly to understand the temperature and the internal resistance of the pack so we will allow our customers to create detailed models so they can connect sensors to the platform and feed that data to the model as well to be part of the model, adding in the specific chargers being used and use that information instead of simply relying on the current of the charger.”
“You can look at the state of charge (SoC) and the state of health (SoH) to see how that is performing given the current and voltage, and you can zoom in to diagnose issues. This is really useful for error identification and battery diagnostics. One customer has 4000 systems in India and they can go inside the system and see all the batteries on the road,” said Aryan.
However the battery management system has more intelligence than just the digital twin. “We don’t always follow the model – so if the data deviates that triggers the machine learning to look at the recommend over the air changes in the usage parameters that can change the trajectory to a potential trajectory with a recovery point so that instead of being a diagnostics you can have predictive data,” said Aryan.