By Allison Buenemann, Product Marketing Manager at Imubit
At Euro Petroleum Consultants' Energy & Sustainability Forum in Madrid, Imubit Economic Engineer Dennis Rohe presented a session on the role of closed-loop optimization in the energy transition, and joined a panel on how technology is enabling decarbonization in refining. This post picks up the theme of that session.
Objectives beyond margin
The objective of a closed-loop optimization application is usually to maximize profit. It is expressed as an economic objective function that uses the unit mass balance to compare the value of the products with the cost of the feeds. As sustainability goals rise in priority, more has to be built into that objective.
Reducing energy consumption or the overall carbon intensity of a process comes with business trade-offs. Energy use or emissions can enter the objective as a cost term, or as a constraint the controller has to respect. With a well-designed objective function, sound constraint settings and a controller that can act on them, a plant can lower its carbon footprint and still stay in business.
Why advice alone fell short
In energy and chemicals, the words "closed-loop AI" used to cause real concern. To ease it, companies tried AI in open loop, with operators acting on model recommendations. It quickly became clear that an operator cannot adjust dozens of setpoints every few minutes on a model's advice.
To keep the optimization potential while answering the concern, the approach shifted to pre-trained controllers running on site, with the trained model embedded at the plant rather than calculated live in the cloud. That is how Imubit's Deep Learning Process Control® (DLPC®) runs. A Deep Learning Controller, trained by reinforcement learning across millions of simulations of the unit, runs on-premise and writes to the APC or DCS minute by minute, inside limits set by plant engineers.
Results where energy is the target
Hydrocarbon processing is complex and interdependent. A controller trained across millions of operating scenarios can make moves that are sometimes not intuitive to a person at the console, and that keep the unit profitable while it uses less energy.
In one deployment at a processor of refinery byproducts, closed-loop control cut natural gas use while holding product quality and increasing yield. Where energy is the optimization target, Imubit reports a 15 to 30% reduction in energy consumption.
Imubit applies closed-loop control to energy-intensive units in refining, chemicals and polymers, cement, and mining and minerals, on top of the plant's existing APC and DCS.
