Biofuel demand is growing, driven by decarbonisation strategies across hard-to-abate sectors. However, scaling up production presents challenges; feedstock variations are unavoidable, and conventional refineries aren’t built for the chemical diversity of waste inputs, nor the contaminants.

In this episode, we speak to Lewis Sweet, general manager for sustainable fuels and chemicals at Honeywell Process Automation, to understand how automation and AI are preparing the biofuel refinery space for a future of increasing demand.

We consider the current capabilities and future potential of automation and AI, discussing how technology can tackle the challenges of inconsistency, both in feedstock supply and content.

Integrating AI in automation for biofuel refining

“Automation in general is the control of the facility”, explains Sweet, but “when AI enters the picture, we are really talking about mixing some layers together, which is really fun”.

It is the combination of the two which can best prepare the biofuel refining sector for a future of rising demand, but there are questions: “How do you take some analytics and some enterprise knowledge, perhaps even some global market economics, and then make informed decisions all the way back down to the control layer?” asks Sweet.

“We have had advanced analytics and advanced control for a long time, that would sit in between that L3 and L4 layer [of the Purdue model], and that kind of sits on top of your basic control”. What has changed is that “AI can put that whole ecosystem together to make an informed decision.”

He adds: “There is not a technology limitation at this point. We can bring that whole ecosystem together, do the analysis, optimise what is best for the plant, depending on how that company wants to operate, profit, lower carbon, maximise throughput, whatever that optimisation needs to be.”

For biofuel producers then, the question now is where AI offers most value.

Automation and AI in supporting biofuel production

Decarbonisation targets in hard-to-abate industries are driving demand for biofuels; the bottleneck is on the supply side. Using technology to aid this solution is, according to Sweet, “where things get really interesting”.

“Technology and markets and global politics all kind of come together in this exact question. In a conventional refinery we are going to design for a handful of feedstocks, and those feedstocks are consistent and well categorised […] the question for bio feedstocks is where you get it and how you find it.”

With AI, he says: “We can kind of predict what we are expecting the reactor to do, and we can change the set points. We can change the control behaviour before we see it actually change.”

The essential point here is the time saving and, by extension, the cost saving. “Instead of having a six- or 12-hour delay and being off-spec for that long, maybe it is just a two-hour delay being off-spec, and then we are right back to it.”

He adds: “Every hour we can put something back on-spec, the more money you can make, which means the closer to market parity we can run these facilities.”

AI to offer granular insights to biofuel refining industry

Optimising processes first means understanding them. For refineries – especially conventional ones, not built for bio feedstocks – the more insight into operations, utilities, feedstocks and supply chains, the more detailed carbon calculations can be.

Currently, Sweet says carbon intensity is calculated by a soft sensor: “It is an estimation. It is a model based on the supply chain or the custody of that feedstock or that product.

“If you put it on the barge, that is going to have a different footprint than if you put it on a truck or via rail; if you buy used cooking oil (UCO) from China versus buying it from Japan or buying it from Brazil, those are going to have different carbon footprints of just your feedstock.

“Then, as you change your utility consumption, as you change your reactor severity, that will also have an impact.”

AI can improve accuracy of carbon calculations, potentially offering molecular-level insight. This could allow price differentiation as the market matures, enabling renewable projects “to become more financially viable”.

Feedstock composition as a pain point

A granular understanding of the composition of feedstocks will be essential to improving efficiency and identifying a fuel’s carbon profile. However, the scientific study of feedstock composition and contamination is still relatively new, says Sweet. “When you get into bio feedstocks, there are hundreds [of types], and you don’t know where they all are.”

“As scientists and engineers, and a technology provider as well, we [Honeywell] are just learning about some of that.” He points to the vast array of potential contaminants and chemical diversity, from the wash-off of fertiliser in conventional feedstocks to French fry crumbs in UCO.

“In conventional fuels, you have carbon chains, but you are typically measuring up to C15 or C16. With biofuels, you have these super-long organic carbon chains – C60 – plus, you have contaminants that you never even thought of.”

AI and automation will be part of the solution, but there is significant progress to be made in the sector. Sweet comments that the industry will need to figure out how to build a model, a math system that can provide better accuracy.

“The good thing is it is not black or white. Any directional improvement we can make in modelling the responsiveness is money that is being put back on the table, which means it is lowering the risk and improving our chances of seeing more renewable fuels.”