Startup offers AI-based software solution to predict component failure in utility-scale inverters

Share

From pv magazine Global

U.S.-based Infinrel has developed a new software solution that actively predicts component failure in inverters used in large-scale photovoltaic plants.

“Our Energy Kardio Graph (EKG) system operates like a medical EKG but 10,000 times faster,” the company’s founder and CEO, Bert Wank, told pv magazine. “It samples small electrical perturbations on the input and output of the inverter at MHz scale, millions of times per second, and enables insight into the actual operation of the inverter on its switching cycle basis.”

The new solution can reportedly display artifacts of degrading components such as saturating inductors, or capacitors that lose the essential capacity to buffer energy, or high frequencies that can damage step-up transformers.

“Examining those figurative ‘heart rhythms’ of the inverter allows infiniRel’s proprietary software to build causality between the electrical signal shape, its cause, and its effect,” Marco Marazzi, infiniRel‘s VP Software Engineering, added. “Essentially, using a Failure Effect and Method Analysis process, our EKG will extract features from the wavelets that relate to component degradation and eventual failure risk.”

The software uses machine-learning algorithms to process data into features that tag each functional group of components inside the inverter, reducing the amount of high causality data from MB/sec to kB/sec. “The EKG system is independent of any data source as it creates high-resolution data by itself,” Wank said. “It is lock-step with the rate semiconductors experience change of both voltages and currents, on a micro-second level and not reporting every second, but one million times per second.”

The company explained that a utility-scale inverter powered by an insulated gate bipolar transistor (IGBT) switches between 1 kHz and 3 kHz, which requires the actual gate drive to contain up to 1 MHz signals in order to optimize the design for efficiency. More recent inverters which deploy gallium arsenide (GaN) or silicon carbide (SiC) power stacks, switch at 20 kHz to 30 kHz, an order of magnitude faster than its fundamental switching frequency which is the heartbeat of the power conversion system.

“If any of those drive signals is off, even slightly, internal power dissipation dramatically increases, leading to stack failure,” Wank further explained. “The software control loop, geared to match the utilities’ 60 Hz power quality requirements and regulating on a switching cycle basis, does not have the bandwidth to observe and respond to such subtle changes.”

Based in San José, California, Infinirel is currently raising capital to build more EKGs to deploy with its customers in the energy transition.

“We are going to expose our solution at RE+ in Anaheim next week,” said Marazzi. “Infinirel was one of the 10 winning teams of The Set! Contest that will advance to the final phase of the American-Made Solar Prize Round 3.” The U.S. Department of Energy (DoE) developed the competition to incentivize the nation’s domestic solar entrepreneurs. Infinirel and each of the 9 other finalists will receive $100,000 in cash and $75,000 in vouchers at the DOE National Labs or other qualified partner facilities, to further develop their technology.

This content is protected by copyright and may not be reused. If you want to cooperate with us and would like to reuse some of our content, please contact: editors@pv-magazine.com.

Popular content

EV expansion depends on bidirectional charging capability, says report
10 October 2024 Nuvve cites electric school bus fleets as a promising means to improve vehicle-to-grid adoption.