Photovoltaic Panels Solar Farm

A kind of synthetic intelligence known as machine studying may help scale up manufacturing of perovskite photo voltaic cells.

Perovskite supplies could be superior to silicon in PV cells, however manufacturing such cells at scale is a big hurdle. Machine studying may help.

Perovskites are a household of supplies which can be at the moment the main contender to switch the silicon-based photo voltaic photovoltaics which can be in broad use at this time. They carry the promise of panels which can be far lighter and thinner, that might be made in giant volumes with ultra-high throughput at room temperature as an alternative of at a whole bunch of levels, and which can be simpler and cheaper to move and set up. But bringing these supplies from small laboratory experiments right into a product that may be manufactured competitively has been a protracted wrestle.

Production of perovskite-based photo voltaic cells includes optimizing no less than a dozen or so variables directly, even inside one explicit manufacturing method amongst many prospects. However, a brand new system primarily based on a novel method to machine studying might pace up the event of optimized manufacturing strategies and assist make the following era of solar energy a actuality.

The system, developed by researchers at MIT and Stanford University over the last few years, makes it possible to integrate data from prior experiments, and information based on personal observations by experienced workers, into the machine learning process. This makes the outcomes more accurate and has already led to the manufacturing of perovskite cells with an energy conversion efficiency of 18.5 percent, which is a competitive level for today’s market.

AI Optimized Production of Perovskite Solar Cells

The optimized production of perovskite solar cells could be sped up thanks to a new machine learning system. Credit: Photo of solar cell by Nicholas Rolston, Stanford, and edited by MIT News. Perovskite illustration by Christine Daniloff, MIT

The research was recently published in the journal Joule, in a paper by MIT professor of mechanical engineering Tonio Buonassisi, Stanford professor of materials science and engineering Reinhold Dauskardt, recent MIT research assistant Zhe Liu, Stanford doctoral graduate Nicholas Rolston, and three others.

Perovskites are a group of layered crystalline compounds defined by the configuration of the atoms in their crystal lattice. There are thousands of such possible compounds and many different ways of making them. While most lab-scale development of perovskite materials uses a spin-coating technique, that’s not practical for larger-scale manufacturing, so companies and labs around the world have been searching for ways of translating these lab materials into a practical, manufacturable product.

“There’s always a big challenge when you’re trying to take a lab-scale process and then transfer it to something like a startup or a manufacturing line,” says Rolston, who is now an assistant professor at Arizona State University. The team looked at a process that they felt had the greatest potential, a method called rapid spray plasma processing, or RSPP.

The manufacturing process would involve a moving roll-to-roll surface, or series of sheets, on which the precursor solutions for the perovskite compound would be sprayed or ink-jetted as the sheet rolled by. The material would then move on to a curing stage, providing a rapid and continuous output “with throughputs that are higher than for any other photovoltaic technology,” Rolston says.

“The real breakthrough with this platform is that it would allow us to scale in a way that no other material has allowed us to do,” he adds. “Even materials like silicon require a much longer timeframe because of the processing that’s done. Whereas you can think of [this approach as more] like spray portray.”

Within that course of, no less than a dozen variables could have an effect on the end result, with a few of them being extra controllable than others. These embrace the composition of the beginning supplies, the temperature, the humidity, the pace of the processing path, the space of the nozzle used to spray the fabric onto a substrate, and the strategies of curing the fabric. Many of those elements can work together with one another, and if the method is within the open air, then humidity, for instance, could also be uncontrolled. Evaluating all attainable combos of those variables via experimentation is not possible, so machine studying was wanted to assist information the experimental course of.

But whereas most machine-learning methods use uncooked knowledge corresponding to measurements of {the electrical} and different properties of check samples, they don’t usually incorporate human expertise corresponding to qualitative observations made by the experimenters of the visible and different properties of the check samples, or data from different experiments reported by different researchers. So, the workforce discovered a method to incorporate such exterior data into the machine studying mannequin, utilizing a likelihood issue primarily based on a mathematical method known as Bayesian Optimization.

Using the system, he says, “having a model that comes from experimental data, we can find out trends that we weren’t able to see before.” For instance, they initially had hassle adjusting for uncontrolled variations in humidity of their ambient setting. But the mannequin confirmed them “that we could overcome our humidity challenges by changing the temperature, for instance, and by changing some of the other knobs.”

The system now permits experimenters to way more quickly information their course of so as to optimize it for a given set of circumstances or required outcomes. In their experiments, the workforce targeted on optimizing the ability output, however the system may be used to concurrently incorporate different standards, corresponding to value and sturdiness — one thing members of the workforce are persevering with to work on, Buonassisi says.

The scientists had been inspired by the Department of Energy, which sponsored the work, to commercialize the expertise, they usually’re at the moment specializing in tech switch to current perovskite producers. “We are reaching out to companies now,” Buonassisi says, and the code they developed has been made freely obtainable via an open-source server. “It’s now on GitHub, anyone can download it, anyone can run it,” he says. “We’re happy to help companies get started in using our code.”

Already, a number of corporations are gearing as much as produce perovskite-based photo voltaic panels, though they’re nonetheless understanding the small print of the best way to produce them, says Liu, who’s now on the Northwestern Polytechnical University in Xi’an, China. He says corporations there should not but doing large-scale manufacturing, however as an alternative beginning with smaller, high-value purposes corresponding to building-integrated photo voltaic tiles the place look is vital. Three of those corporations “are on track or are being pushed by investors to manufacture 1 meter by 2-meter rectangular modules [comparable to today’s most common solar panels], within two years,” he says.

‘The problem is, they don’t have a consensus on what manufacturing expertise to make use of,” Liu says. The RSPP methodology, developed at Stanford, “still has a good chance” to be aggressive, he says. And the machine studying system the workforce developed might show to be vital in guiding the optimization of no matter course of finally ends up getting used.

“The primary goal was to accelerate the process, so it required less time, less experiments, and less human hours to develop something that is usable right away, for free, for industry,” he says.

“Existing work on machine-learning-driven perovskite PV fabrication largely focuses on spin-coating, a lab-scale technique,” says Ted Sargent, University Professor on the University of Toronto, who was not related to this work, which he says demonstrates “a workflow that is readily adapted to the deposition techniques that dominate the thin-film industry. Only a handful of groups have the simultaneous expertise in engineering and computation to drive such advances.” Sargent provides that this method “could be an exciting advance for the manufacture of a broader family of materials” together with LEDs, different PV applied sciences, and graphene, “in short, any industry that uses some form of vapor or vacuum deposition.”

Reference: “Machine learning with knowledge constraints for process optimization of open-air perovskite solar cell manufacturing” by Zhe Liu, Nicholas Rolston, Austin C. Flick, Thomas W. Colburn, Zekun Ren, Reinhold H. Dauskardt and Tonio Buonassisi, 13 April 2022, Joule.
DOI: 10.1016/j.joule.2022.03.003

The team also included Austin Flick and Thomas Colburn at Stanford and Zekun Ren at the Singapore-MIT Alliance for Science and Technology (SMART). In addition to the Department of Energy, the work was supported by a fellowship from the MIT Energy Initiative, the Graduate Research Fellowship Program from the National Science Foundation, and the SMART program.

(function(d,s,id){var js,fjs=d.getElementsByTagName(s)[0];if(d.getElementById(id))return;js=d.createElement(s);js.id=id;js.src=”https://connect.facebook.net/en_US/sdk.js#xfbml=1&version=v2.6″;fjs.parentNode.insertBefore(js,fjs);}(doc,’script’,’facebook-jssdk’));



Sources