(370e) Hybrid Cutpoint Optimization Using Improved Swing-Cut Method and Data Analytics for Crude-Oil Distillation Towers

A fundamental step in crude-oil refinery optimization is the calculation of distillation unit yields and properties [1]. For such calculation, both rigorous and surrogate representations can be considered for determination of product quantities and qualities of distillation processes. The rigorous or first principles modeling approach considers mass/molar and energy balances and equilibrium equations in the columns. As a result, compositions and flows of internal and external streams with regards to operational conditions, such as pressure and temperature of each stage, can be calculated with process simulation. However, despite the sufficient robustness and accuracy in the formulas, rigorous distillation models demand high computational effort, imposing difficulties for their application in large-scale optimization problems and may also not be directly optimizable. On the other hand, non-rigorous modeling can use surrogate or simplified empirical correlations normally based on crude-oil raw material assays [2] or measured data [3] to represent the modifications promoted by the mass and energy balances over the hydrocarbon components. Due to their simplicity, straightforward application and relatively good accuracy, they are commonly used for process optimization in oil refineries [4]. A review and comparison of some non-rigorous models for cutpoint optimization can be found in Fu and Mahalec [1].

By observations in the refinery site, the bulk quality determination or match of the raw material composition (assay) of crude-oils to be processed typically determines by 80 to 90% of the amounts and properties of distillates. The remaining part (10 to 20%) is determined by variations of operational variables such as material flows of side-strippers and pump-arounds (externally to the towers), initial and final boiling points of distillates (cutpoints), parallel split ratios of pre-heat exchanger trains, feed and tray temperatures in the hydrocarbon fractionator furnace and tower, pressure profile, internal reflux, etc. Hydraulic limits of product distillates may also be included in the mathematical model considering operational conditions inside the distillation towers as well as in downstream unit-operations. Moreover, lower and upper bounds on quality specifications can be considered in the cutpoint model (around 30 distinct types of properties such as sulfur concentration, specific gravity, acidity, residue content). Due to uncertainties in the transformation equations, feed quality, measured data, etc., high-fidelity modeling becomes unrealistic to be included in the online blend scheduling and processing optimization of crude-oil raw materials.

In this work, we address a hybrid cutpoint temperature optimization based on an improved swing-cut method [3] and data analytics in base+delta models for separation of crude-oils in distillation towers. The improved swing-cut method considers crude-oil assays and temperature range spaces of the distillates in the towers. The base+delta models combine reconciliation and regression into an estimation strategy and utilize measured data from the field or laboratory and/or simulated data from rigorous simulation software such as Petro-SIM, AspenPlus, Hysys, UNISIM, and VMGSim, to determine deltas with respect to bases on operational variables and disregards crude-oil assay data. Similarly, the hybrid modeling in Mahalec and Sanchez [5] uses first principles in short-cut distillation methods [6] and simulated results to determine the fractionation inside the columns.

For better predictions on process-shop’s yields and properties, Franzoi et al. [7] use predictive analytics techniques by doing constrained and weighted least squares to fit better base+delta sub-models using data reconciliation and regression techniques. The reconciliation forces the yields to add up to 100% and the regression fits base and delta coefficients simultaneously across all yields. The proposed hybrid cutpoint optimization approach can be applied to online optimization of crude-oil blend scheduling operations in complex industrial-sized refineries to determine the composition-quality feed demands for the amounts and properties of distillates in towers in cascade (as the real process equipment design). This improved model can be integrated into a continuous cycle to provide process measured feedback, which leads to higher precision in the model and aims to reducing the gap between the model and the actual plant values [8].

The cutpoint model is applied in a distillation system [12] to be integrated to a crude-oil blend scheduling optimization so that a mixed integer non-linear programming (MINLP) model is formulated. To solve the MINLP, an MILP-NLP phenomenological decomposition heuristic [9] problem is optimized iteratively until a convergence criteria is achieved. Therefore, the NLP step of decomposed MILP-NLP crude-oil blend scheduling optimization [10], [11] determines the crude-oil final feed quality to be charged to the complex system of towers in cascade (pre-flash, CDUs, VDU and naphtha separation tower) as well as the amounts and properties of the distillate streams. Therefore, this takes into consideration the crude-oil assay to be processed (prescriptive analytics) from the improved swing-cut method [3] and the bases and deltas of the fractionation inside the columns with regards to the operational variables considered in the data analytics step (predictive analytics) of the proposed hybrid model.

[1] Fu G, Mahalec V. (2015). Comparison of Methods for Computing Crude Distillation Product Properties in Production Planning and Scheduling. Industrial and Engineering Chemistry Research, 54 (45), 11371-11382.

[2] Kelly JD, Menezes BC, Grossmann IE. (2014). Distillation Blending and Cutpoint Temperature Optimization Using Monotonic Interpolation. Industrial and Engineering Chemistry Research, 53 (39), 15146-156.

[3] Menezes BC, Kelly JD, Grossmann IE. (2013). Improved Swing-Cut Modeling for Planning and Scheduling of Oil-Refinery Distillation Units. Industrial and Engineering Chemistry Research, 52 (51), 18324-333.

[4] Li W, Hui C, Li A. (2005). Integrating CDU, FCC and product blending models into refinery planning. Computers and chemical engineering, 29 (9), 2010-2028.

[5] Mahalec V, Sanchez Y. (2012). Inferential Monitoring and Optimization of Crude Separation Units via Hybrid Models. Computers and chemical engineering, 45, 15-26.

[6] Thiele EW, Geddes RL. (1933). Computation of distillation of hydrocarbon apparatus for hydrocarbon mixtures. Industrial and Engineering Chemistry, 25 (3), 289–295.

[7] Franzoi RE, Kelly JD, Menezes BC, Gut JW. (2018). Advanced Data Analytics for Process-Shop Base+Delta Sub-Model Estimation in Planning and Scheduling Decision-Making. In: AICHE Annual Meeting, Pittsburgh, PA, United States.

[8] Kelly JD, Zyngier D. (2008). Continuously improve the performance of planning and scheduling models with parameter feedback. In: Proceedings of the foundations of computer-aided process operations (FOCAPO).

[9] Menezes BC, Kelly JD, Grossmann IE. (2015) Phenomenological decomposition heuristic for process design synthesis of oil-refinery Units. Computer Aided Chemical Engineering, 37, 1877-1882.

[10] Kelly JD, Menezes BC, Engineer F, Grossmann IE. (2017). Crude-Oil Blend Scheduling Optimization of an Industrial-Sized Refinery: A Discrete-Time Benchmark. In Foundations of Computer Aided Process Operations, FOCAPO, Tucson, AR, United States, 10-13 January.

[11] Franzoi RE, Menezes BC, Kelly JD, Gut JW. (2018). Effective Scheduling of Complex Process-shops using Online Parameter Feedback in Crude-Oil Refineries. Computer Aided Chemical Engineering, 44, 1279-1284.

[12] Franzoi RE, Menezes BC, Kelly JD, Gut JAW. (2018). Blend Scheduling Optimization Using Factors for Qualities in Cascaded Distillation Towers in Crude-Oil Refineries, In: São Paulo: Blucher. 1233-1236.