The grant awarded recently to UM-Dearborn faculty Yi Lu Murphey, Wencong Su and Mengqi Wang isn’t just any old research grant. It comes from a special unit within the U.S. Department of Energy called the Advanced Research Projects Agency (ARPA). As the name suggests, the end-game here is fast-paced research that has the potential to fundamentally transform the energy industry and “dramatically accelerate the pace of energy innovation.”
One of 23 projects funded recently by ARPA, Murphey, Su and Wang’s project is taking on a particularly stubborn problem with electric power converters. Power converters are an unsung component of just about anything that uses electricity, whether it’s a phone charger, an electric car or a sophisticated component of the electric grid. Yet despite their ubiquity, they're hardly an off-the-shelf technology. Typically, anytime someone wants to develop a new device, the power converter has to be designed from scratch to meet the demands of an individual application.
For startups and smaller companies, this can be a huge speed bump in product development.
“You might only have to design one of these converters a year, but it can be extremely expensive,” Su explains. “One of the companies we surveyed estimated the costs of having an experienced engineer, plus overhead, to be upwards of $700,000 a year. And that’s if you can even find a qualified candidate for your project. Right now, there is a big shortage of people who work in power electronics engineering and hardware design.”
The UM-Dearborn team is hoping to solve this issue by building an artificial intelligence-based software to handle the actual power converter design. In essence, a company would just have to define the necessary parameters — like desired power input and output, voltage, size and weight of the power converter, etc. — and the machine would do all the complicated engineering.
If they’re successful, the tool could have wide-ranging impacts, from aiding small-budget consumer electronics start-ups who can’t afford an engineer, to fast-tracking technologies needed to transform the electric grid for the era of renewable energy.
Su says the team will spend the next two years developing an initial version of their software. After that, they plan to deploy it with five partner companies, who will try it out and offer feedback. The end goal is to have a commercial software product that could be distributed widely across industries within just a few years.
###
The UM-Dearborn team is pursuing their project in collaboration with Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, Analog Devices, Inc. and Modelon, Inc. The grant also includes the possibility of $1.2 million in additional funding for Phase II research.
