- AI might be a buzzword but when it comes to energy management it is currently the only tool that can take huge amounts of data and make meaningful forecasts to optimize the use of renewable energy and storage, especially as EVs proliferate.
Energy startup Lade, based in Mainz, Germany, focuses on optimizing renewable energy consumption across EV charging and energy management. AI is already proving to be a useful tool deployed for customers’ benefits.
Lade founder and chief executive officer (CEO) Dennis Schulmeyer told pv magazine that an internal team of seven dedicated employees is working on AI in combination with the company’s LADEgenius product, that can handle 200 EV chargers, to interface with local data inputs from PV modules, energy storage systems, and EV chargers, along with inputs and outputs to fulfil grid regulations. LADEgenius is basically an on-site load manager and connector that can make decisions with the help of cloud intelligence. That cloud intelligence uses AI and machine learning, via a system the company calls Lana.
“Lana is AI because she is able to forecast the availability of energy,” said Schulmeyer. “Lana can gather data from weather services in Germany and forecast up to five days to ascertain how much renewable energy will be available.
“We also forecast the availability of local renewable energy for the building, for generation, reading inverter data and weather values for the installation, and forecast consumption as well. Our main [unique selling point]is also being able to forecast car arrival and departure times and how much energy the cars will really need up to five days into the future, and [we]calculate the optimal charge plan for that time.”
All of that comes at a “high cost,” said Schulmeyer, as the AI trains on data and runs on models hosted on cloud servers, with Lade adding some additional costs for itself by paying for the use of strictly renewable energy, with offsets for the servers.
“Our internal team developed the AI for the past three years,” said the CEO. “We initially trained it to use open-source data while adding real data from our chargers and, for example, data from customers from their PV generation, and even our own real-world setup here in Mainz.” Schulmeyer confirmed that adding additional customer data to Lana’s training data has improved predictions further.
Schulmeyer said that advanced software-based controls may solve some problems for a single-dwelling situation but standard equipped-load managers and PV surplus charging systems will soon struggle to deliver real advantages when considering multiple EV chargers. “This is the showstopper,” he added.
In larger commercial and industrial situations energy management needs to happen across numerous EV chargers to avoid unnecessarily large demand without coordination, which makes the task increasingly complex. This is made even more complex by adding forecasting generation and consumption via weather data while offering features such as peak shaving. This would be impossible to operate without AI technology, said the Lade founder.
“We do all of this and we’re improving,” he said. “If you connect to our EV chargers for the first time, we say our estimates for the energy the car will need over time will have an accuracy of around 67%, up from a lower starting point. But of course, the more data we have, the better it will be – and the advantage of a startup is that we run many models and AI technologies, and we adapt.”
Schulmeyer was careful to point out advantages for the entire ecosystem that go further than AI. “It’s not only the AI algorithm … it’s how you think as a company,” he said. “We are not alone and we will find ways to include others. Indeed, we’ll add third-party chargers in our cloud, with LADEgenius. But this is important because we are not independent in terms of being the only ones to exist in this area. And our goal, above all, is the energy transition, with the help of electric mobility.”
Author: Tristan Raynor
This article was originally published in pv magazine and is republished with permission.