(35b) Data-Driven Real-Time Estimation and Optimization for Integrated Pre-Heat Exchange Trains and Distillate Cutpoints in Multiple Fractionation Columns

In crude-oil refineries, the fine adjustment of the operational variables of the hydrocarbon fractionation columns can be optimized in an integrated multi-unit on-line strategy. It considers demands on amounts and properties of intermediate hydrocarbon streams for a) downstream process-shops (e.g., to match quality feed needs in hydrotreating units) and b) blend-shops of final streams as fuels, lubes, asphalts and petrochemical feeds. However, real-time optimization (RTO) strategies in the hydrocarbon refining industry are normally modeled using first principle formulas considering reduced scale problems as for a sole process unit. Nevertheless, due to the convergence difficulties of the solutions, the RTO applications in industry are often found in off-line mode. The lack of high quality techniques of both estimation (data reconciliation, parameter estimation) and statistical analysis (redundancy, observability, variance) is the main reason for that, although difficulties from nonlinearities of rigorous formulation may arise. To enlarge the scale of the RTO from one process unit to a plant or more integrated units, this work uses actual process or operating data with reduced or relaxed-ordered models (ROM) to optimize in real-time initial and final cutpoints [1] of distillates of multiple fractionation columns integrated with their pre-heat exchange trains.

In the integrated data-driven RTO strategy proposed, the setpoints of material flows, 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, are determined considering on-line measurement feedback to incrementally situate the plant/sub-plant to a more profitable operating or processing space. The RTO formulation considers nonlinear programming (NLP) to solve on-line (real-time, closed-loop) and in-line (real-time, open-loop) applications. By adjusting pressure, temperature, flow, separation efficiencies, catalyst activity, etc., slight ranges of separation and conversion of the hydrocarbon molecules can be achieved. Therefore, the integrated RTO manipulates operational variations in physical and chemical processes for production of crude-oil refined streams containing a range of light to heavy hydrocarbon molecules whose extension yields the amounts and properties of bulk fuels such as gasoline, kerosene, jet and diesel [2].

The proposed multi-unit data-driven RTO uses the surrogate approach of distillation blending and temperature cutpoint formulation from Kelly et al. [1] conjugated with rigorous modeling of pre-heat exchanger networks, resulting in a hybrid RTO model as found in Mahalec et al. [3], although their formulation is only applied for a single distillation unit. The surrogate or reduced order model using distillation curve adjustment or shifting technique is based in experimental ASTM / SIMDIST points and measurement of material flows of the column distillates. This is integrated with the heat exchange RTO that uses the real-time estimation (RTE) engine to generate UA heat transfer coefficients continuously and simultaneously with the data reconciliation steps.

The use of RTO is challenging and expensive because of the difficulties in building and adapting accurate models for complex chemical processes [4] and since it involves steady state detection, state/parameter estimation, data reconciliation and solving of nonlinear optimization problem on-line [5].

[1] Kelly, J. D.; Menezes, B. C., Grossmann, I. E. Distillation blending and cutpoint optimization using monotonic interpolation. Industrial & Engineering Chemistry Research, 53, 14146−14156, 2013.

[2] Menezes, B. C.; Grossmann, I. E.; Kelly, J. D. Enterprise-wide optimization for operations of crude-oil refineries: closing the procurement and scheduling gap. Computer Aided Chemical Engineering, 40, 1249–1254, 2017.

[3] Mahalec, V.; Sanchez, Y. Inferential Monitoring and Optimization of Crude Separation Units via Hybrid Models. Computers & Chemical Engineering, 45, 15−26, 2012.

[4] Chachuat, B.; Srinivasan, B.; Bonvin, D. Adaptation strategies for real-time Optimization. Computers & Chemical Engineering, 33 (10), 1557–1567, 2009.

[5] White, D. C. Online optimization: What, where and estimating ROI. Hydrocarbon Processing, 76 (6), 4351, 1997.