Sustainable aviation fuel, made from bio-based sources instead of petroleum, plays a key role in reducing our reliance on fossil fuels. But developing sustainable aviation fuel, also known as biojet, is not an easy task.
Researchers need to undergo several processes in order to identify a potential biojet fuel and begin successful production in a pilot facility. Often, this means spending time and money in the lab doing trial-and-error basic research and later scaling up production to make enough fuel for testing.
New Berkeley Lab Software Saves Time and Money
Now, two new publicly available web-based software tools developed at Lawrence Berkeley National Laboratory (Berkeley Lab), with funding support from the U.S Department of Energy Bioenergy Technologies Office, aim to help researchers and companies quickly test different scenarios and explore viable bio-based fuels and products without ever stepping foot in the lab.
The first tool, Feedstock to Function, addresses one of the biggest hurdles involved in starting biojet fuel research: knowing whether a molecule could be viable as a fuel.
“During early stage development, we are often missing experimentally measured properties of viable jet fuel molecules,” said Vi Rapp, a research scientist at Berkeley Lab. “Without these properties, it is hard to predict how molecules may perform as a jet fuel.”
Typically, scientists need to produce at least a gallon of fuel in the lab in order to be able to properly examine its properties. Securing funding to support the production of these larger quantities can be a significant challenge.
“This creates a chicken-and-egg problem,” said Corinne Scown, a staff scientist at Berkeley Lab. “Companies want to see testing results for a new fuel molecule or product before they invest time and money in commercializing it, but researchers may struggle to find funding to support this scale-up unless they have already secured a strong industry partner.”
These challenges led Scown and Rapp to realize that machine learning could be a powerful tool to solve this problem.
“Given other scientific machine learning successes, we figured it could help provide an initial understanding of molecule performance, or identify new molecules we may have missed — and that’s what led us to develop this tool,” Rapp said.
Feedstock to Function and BioC2G Work Hand in Hand
Feedstock to Function lets users sort through a database of almost 10,000 potential molecules. Instead of having to measure a molecule’s properties in a lab, the tool uses machine learning to predict these properties based on existing data. Users can search either for a specific molecule or explore multiple molecules that match the properties they’re trying to achieve.
“The tool lets users explore and find molecules that might work for their applications without wasting valuable resources,” Rapp said.
Once a user identifies a potential molecule, they can then use another tool, Bio-Cradle-to-Grave (BioC2G), to explore potential locations and process configurations for scaling up their production. The tool provides valuable information that will help companies decide which locations are optimal based on the potential feedstock availability nearby, and what greenhouse gas emissions and costs are associated with developing a new production facility.
“Scientists sometimes need help figuring out what the cost drivers are and how to prioritize their efforts in terms of further research and process optimization,” Scown said. “Looking at the system as a whole can reveal where the low-hanging fruit are in terms of reducing costs and emissions. With this tool, companies and researchers can go through the exercise of building a kind of virtual test facility before they ever have to break ground on the real thing. It is a great cost-free way to do a trial run and start working the kinks out.”
Knowing this information up front can help scientists and companies figure out how to bring down costs or improve their process so that when they do scale up their product to larger scales, it’s more likely to be successful. Having done this due diligence may also help them find investors.
“That helps them get over that significant hurdle,” Scown said. “When they finally do their first pilot run, it’s a lot more viable. They know where the bottlenecks are and how they can be addressed.”
Access Feedstock to Function and BioC2G
Together, these tools enable scientists and companies to gain the knowledge they need to select feedstocks and products that will make their first pilot runs more successful, speeding up the time to commercialization. The tools are now available online.
“We want to help researchers and industry explore new opportunities rapidly, successfully, and risk-free,” Rapp said. “We hope these tools will help accelerate successful research in biojet fuel development.”