(3e) Improving Hydrotreating Operation Via Dynamic Simulation Modelling | AIChE

(3e) Improving Hydrotreating Operation Via Dynamic Simulation Modelling


Remesat, D. R. - Presenter, Koch-Glitsch Canada LP

The body of work performed by the author (fulfillment of PhD from University of Calgary) entailed gathering a substantial amount of relevant industrial data to validate an optimized rigorous steady state hydrotreater model that is dynamically applied to the author's database. The goal of the work was to develop a reliable model that simulates the process over the length of a hydrotreater's run and then apply the model to determine economic ways to improve the operation. By using industrial data, and creating a model that responds to all disturbances during an entire hydro-treater's operation, refiners/upgraders will now have confidence to introduce increased hydrogen (purity and thus partial pressure) to improve operation. The paper/presentation discusses the development of the model, the validation of the model to industrial data, and the application of the model to improve hydrotreater data.

The author has compiled a data base of 24 hydrotreaters (HT) in various services (distillate, Vacuum gas oil (VGO), naphtha, kerosene). Of the 24, 14 HT units have the necessary information to adequately build a representative model of the hydrotreater unit. For publishing purposes, confidentiality agreements with 6 of the refinery/upgrading operators have been obtained.

Many papers on rigorous steady state models of hydro-treaters are available but are typically based on laboratory data. All dynamic hydro-treater models in the public domain are based on laboratory data and deal with a single, simple controlled disturbance variable. No appreciation for actual operation responses (eg. reduction in crude rates, changes in WABT (Weight average bed temperature)) in the model are considered. There are proprietary models that can provide some cursory indication of performance enhancement during operation but due to cost are not readily available in the public domain. Even if these models were available, it is not certain whether the model would be effective in simulating the process and any disturbances over the run length of the process.

Model Development- Catalyst deactivation is by far the most important consideration/parameter when creating a high fidelity model that involves a catalytic reaction process. The key factors to consider in the catalyst deactivation model are: LHSV (liquid hourly space velocity), hydrogen partial pressure, treat ratio (H2 vs. crude charge) and temperature (WABT).

To accurately model catalyst deactivation, a set of 30 correlations using literature sources was developed and improved.

The model was developed/refined in 8 steps. The first 3 versions were primarily performed in the steady state mode, while the remaining versions were developed in the dynamic mode.

Version 1 ? Initial SOR Catalyst deactivation factor ? coke impact

Version 2 ? EOR Catalyst deactivation factor ? metals impact

Version 3 ? Catalyst deactivation -Add in amine saturation crossover -Obtain basic shape of SOR-EOR temp. curve -R-squared = 0.75 for plant D (for all 6 plants = 0.65-0.75)

Version 4 ? Reaction Kinetics ? Add impact of Nf, Np, Sf, Sp, space velocity, Catalyst

Version 5 ? Optimize specific parameter - temperature

Version 6 - Optimize specific parameter ? H2 PP

Version 7 - Optimize specific parameter ? H2/oil ratio

Version 8 ? Optimize all process, reaction parameters simultaneously R-squared = 0.86 for plant D (0.81-0.86 for 6 operating units.)

The model was checked against all available disturbances to determine the effectiveness of the model to simulate the local operation of the hydro-treater. Once the model could reasonably match a local disturbance, the model was run over the entire run length incorporating all changes to key variables.

Application of the model-


With a good dynamic model of the actual process, various scenarios for Plant D were evaluated to determine what mitigating steps in the process could be used to improve the operation.

Hydrogen was used as the primary variable to impact the hydrogen partial pressure. For Plant D, Hydrogen purity was increased to 99.9% from an average of 90.5%. Access to 3rd party on-the-spot Hydrogen was assumed. Hydrogen partial pressure was increased from an average of 1900 to 2050 psig. The following scenarios were performed

1. Increase hydrogen partial pressure at point of 1st major disturbance and maintain same run length. Result: The crude charge could be increased to 14000 BPD, thereby saving an average of 4000 BPD of crude for 4.5 months. The value to a refiner based on $20/bbl upgrade is $8.1MM. The pre-first disturbance charge rate (15000 BPD) could not be fully recovered due to the impact of metals contamination on the catalyst. Hydrogen partial pressure positively impacts deactivation due to coke formation, but negatively impacts metals contamination (various studies illustrate this), and the model accounts for this catalyst phenomenon.

2. Increase hydrogen partial pressure for entire run length Result- a) The run length could be increased by 20% since the initial coke deposition on the catalyst is mitigated by the increased hydrogen partial pressure (WABT will not be as high) with the increased crude (15000 before 1st disturbance and 14000 BPD after). b) If the crude charge is kept as actually happened during the run (and other variables extrapolated), then the run length could be increased by 45%.

Other application examples of the model can be discussed.

Value of the model

- Using actual plant data to develop model

- No issues with scaling up model from laboratory plant scale data

- Model matches disturbances reasonably well, overall performance very encouraging

- Provide confidence to refiners/upgrader's to justify increasing Hydrogen use to improve operations.

Limitations of the model

- Model dependent on available process and laboratory data

- Need to modify model's parameters to address each refiner's/upgrader's specific crude and catalyst


A high fidelity dynamic model for industrial refinery/upgrader hydro-treaters has been developed based on 14 actual operating hydro-treaters. The model accurately simulates local disturbances, and performs well over the entire run length of the hydrotreater. This robust model can be applied to demonstrate the tangible economic benefit of increasing hydrogen use to improve the operation of a hydro-treater by increasing run length and/or improving crude processing.