Big West Oil shares their triumphs from implementing Deep Learning Process Control technology.

By Allison Buenemann, Product Marketing Manager at Imubit

Big West Oil operates a complex, high-conversion refinery in North Salt Lake City, UT. The refinery has a crude capacity 36 kbd and typically sources crudes from Utah, Wyoming, Colorado, and Canada.

In a recent webinar, Big West Oil Operations Technical Specialist Travis Legrande joined Imubit to share how AI is making a big impact on their business. Here’s where they’re delivering value.

1. Increased Capacity in High-value Units

Challenge: Big West Oil’s debutanizer tower (which separates out olefins from the gasoline stream) is undersized, and it was constantly running against its flood point, diminishing separation, and thus throughput capacity. They have an overhead olefins spec limit on C5s and a bottoms vapor pressure spec. They have to strike a delicate balance between violating their 5% max C5s limit on the olefins stream and pushing too much C5 into the bottoms gasoline stream, which raises the vapor pressure and can push that off spec. They have room to push the FCC reactor harder, but when they’re limited by debutanizer capacity they have to cut back FCC conversion, which leaves money on the table.

Solution: Big West Oil implemented Imubit’s Deep Learning Process Control® (DLPC) technology. The closed loop AI model predicts stream properties in real-time, then manipulates reboiler duty tower top temperature simultaneously, and continuously, in order to maximize tower economics. This is a big change from the past, when operators would adjust one parameter, wait a few hours for results, then make another move, all independently.

Results: Big West Oil observed improved tower separation within a day or two of implementing DLPC. They could keep more of the C5s in the higher-value cat gasoline, which allowed them to reduce excess olefin production. Overall throughput capacity of the debutanizer tower increased by 2%. This allowed them to increase FCC conversion and process more gasoline to this tower without hitting the floodpoint.

Figure 1. Yield improvements in cat gas and butane blending show the full picture of Big West Oil’s debutanizer optimization.

2. Increased Liquid Volume Yields

Challenge: The historical operating strategy of Big West Oil’s FCC reactor was to hold riser temperature fairly static, which was leaving significant money on the table. Operators would change riser temperature once a week or so to try to maximize conversions, but this wasn’t often enough to really move the needle on profits.

Solution: Imubit helped Big West Oil build and implement a controller that adjusts that riser temperature continuously, constantly making small moves to maximize conversion. It’s factoring in downstream equipment constraints, and pushing right up against them to ensure they get the most value possible out of the FCC. It optimizes to an objective function consisting of dollar values of products adjusted on a periodic basis as market prices fluctuate. These price deck updates help make sure that DLPC is optimizing the tower in the way that adds the most value to our business.

Results: Big West Oil saw strong initial liquid volume yield improvement of 0.6% on average. They saw that initial improvement slow down after some crude tower changes which impacted the quality of the feed coming to the FCC unit, which prompted the joint Imubit + Big West Oil team to retrain the model on data for the new feed. They’re confident that once the model sees and learns from the data in this new feed quality regime that yield improvements will return to those initially observed.

Figure 2. Increased liquid volume yields versus baseline in months following DLPC implementation.

3. Reduced Conservatism = Reduced Giveaway

Challenge: Due to tower flooding in their diesel stripper tower, Big West Oil was running with the diesel flash point much higher than is optimal, in order to stay on spec. With imprecise lab testing and no real-time certainty around where they stand on flash, they added a 5 degree set point buffer to avoid violating the lower flash limit they must meet to sell their diesel product. This buffer caused them to make more naphtha and less diesel than was possible given their feed.

Solution: DLPC was implemented on the tower, which manipulates the reflux rate and reboiler rate simultaneously. It’s predicting the flash properties in real-time so that they can be less conservative, pushing up against spec limits confident that they aren’t exceeding them.

Results: Big West Oil now has a reliable inferential for the diesel flash property, giving insight into the live flash properties rather than acting on imprecise lab results taken just 3-4 times a day. The inferential gives them a basis to continuously adjust operations rather than reacting to the lab samples. They’ve reduced that buffer on flash from 5 degrees down to 3 degrees, increasing diesel throughput while still keeping diesel on spec.

Figure 3. Reduced diesel flash from baseline.

Big West Oil has delivered big results, fast, with Imubit DLPC. The key to their success? Early operator buy-in. In a world where a controller and optimizer can be turned on or off based on whether that operator trusts the results, trust is a must. Getting all of the operators trained and comfortable with the Imubit technology before implementation led to engagement rates well beyond 85% goals.

Want to join the ranks of the most AI-savvy operations teams? Want to reduce giveaway and increase yields at your refinery? Get a demo.