Naptha System Yield Optimization

$6-9 MM

Value Generation
per annum

7 months

Project Timeline
kick-off to closed-loop

2 hrs/week

Internal Resources
per client group
The Problem & Deep Learning Process Control® (DLPC) Scope

A client with a 285 kbd refinery identified an opportunity to optimize yield across two reformers, while achieving site-wide octane barrel targets on their gasoline pool during the winter season. The reformers are used as balancing octane producers, requiring frequent adjustment on feed charge and severity to compensate for swings in the overall gasoline pool. Between varying crude feed slates and the fluctuating gasoline component production rates and qualities, the client had difficulty achieving the optimal octane barrel target consistently. If the octane barrel fell below target, the client would have to sell at a discount or blend with high-value material at a loss. The reformers were also operated sub-optimally, bypassing a portion of the feed around the reformers and producing higher octane reformate with an overall loss of yield. To access the untapped value of the naphtha system, Imubit’s DLPC continuously manipulates both reformer feed rates and severities to maximize liquid yields and maintain a consistent desired site-wide pool octane make, while respecting unit constraints.

Naptha system configuration

FIGURE 1 Client’s system configuration around FCCU and Crude with DLPC Design.

Value Generation
octane barrel targets
FIGURE 2 Tighter control to overall octane barrel targets with DLPC.
reformer feed excess
FIGURE 1 Reduction of overall reformer feed bypassed for increased yield through reformers with DLPC.
By manipulating the feed and severity on each reformer and responding to octane barrel disturbances in the rest of the gasoline pool, DLPC demonstrated improved control of overall refinery octane barrels. The average octane barrel deviation from the site-wide target was reduced by over 90 octane- kbd with DLPC closed loop for four months. With tighter control to the octane barrel target, DLPC also captured value by increasing and maintaining higher reformer feed rates. The overall reformer feed charge was increased on average by almost 2 kbd, upgrading the naphtha feed and reducing the amount of excess, reprocessed feed, while still meeting the octane barrel target. DLPC was able to capture an annualized value of $6-9MM/year when compared to the baseline period.
Realtime Performance

Prior to DLPC, the reformer feed charge and severity were manually adjusted using existing Advanced Process Control (APC) external targets, set by the LP and gasoline blenders weekly, to meet the overall octane barrel target for the gasoline pool. However, the reformate octane barrel produced often deviated from the overall octane barrel target, resulting in slow overcorrections and at times as the cost of reformate yield. In addition, disturbances in the gasoline pool have long-time dynamics and non-linear responses, causing large swings to the overall octane barrel production making it very difficult to optimize. When engaged, DLPC manipulates the feed charge and severity on each of the reformers to proactively control the reformate octane barrels produced to the octane barrel target while respecting critical operating limits. With the ability to predict impact and respond to disturbances dynamically, DLPC can adjust the reformers, continuously minute-by-minute, to consistently optimize the reformate octane barrel to the desired octane barrel production with operational stability.

Octane Barrels Produced with DPLC
FIGURE 4 (1) (1) Overall octane barrels produced deviate significantly from the overall octane barrel target when DLPC is not engaged. (2) With DLPC engaged, reformer severity (SCF/bbl.) is adjusted, resulting in steady overall octane barrel production.
Resources & Timeline
hydro case study timeline
FIGURE 1 Avg. client resource requirements for DLPC project
DLPC development, which includes Scoping, Data Cleaning, Model Training, and client approvals throughout the entire process. Client resources during development include defining scope, sending data over to Imubit, and approving DLPC models during various milestones throughout Imubit’s Project Workflow. Commissioning requires the most amount of support as it includes implementation, operations training, and building confidence with the entire organization. Once DLPC is in Closed Loop and performing, Continuous Improvement effort required is minimal and can increase based on the desire of the client to be more involved in the process of maintaining DLPC models.

Unlock your plant’s Untapped Value.

It all starts with Imubit.