(444b) Real-Time Optimization of an Industrial Hydrocracking Plant

Hydrocracking is a catalytic chemical process  which converts high?boiling hydrocarbons to more valuable lighter products like diesel, kerosene, naphtha and LPG. Environmental concerns and increased demand for low-sulphur diesel and high smoke point jet fuel have played a major role in the increased utilization of the hydrocracking technology. Hydrocracking is carried out in the presence of hydrogen at high temperatures (260 ?495°C) and pressures (35 ? 200 bar). The main reactions are cracking and hydrogenation which take place in multiple catalytic beds.  The overall system is highly exothermic and hydrogen quench is used for interstage cooling. Due to hydrogenation, sulfur and nitrogen compounds are removed in the process to yield low?impurity reactor product. In the fractionation part of the plant, the reactor effluent is separated into the final valuable products.

Our research covers the real time plant-wide optimization of the hydrocracking reactors and the fractionation columns of TUPRAS refineries. We have developed a hierarchical control structure which is based on the spatial decomposition of the plant into reactor and fractionation subsystems and temporal decomposition of the control tasks based on the characterization of disturbances, availability of measurements and frequency of control actions. The hierarchical control structure consists of a cascade of MPCs operating in tandem with economic optimization as shown in Fig.1. The highest level in the hierarchy is the management layer which considers the demand for the products by the market and different units in the refinery.  A refinery-wide linear programming is performed to compute the prices of feed, hydrogen and products. In addition product specifications and operating constraints for the hydrocracking plant are also defined. The next layer below in the hierarchy is steady-state economic optimization.  Based on prices, product specifications and operating constraints set by the management, steady?state optimization maximizes profit by calculating the optimal product distribution (i.e. the relative amounts of light ends, light naphta, heavy naphta, diesel, kerosene, and bottoms). The 95% boiling point temperatures (or cut-points) of  the products are also computed.  Next optimal operating conditions found by the steady-state optimizer are realized by a cascade of MPCs. In this cascade, inner MPCs consist of the individual reactor and fractionator MPCs. Reactor MPC controls the bed exit temperatures by changing the set-points of hydrogen quench PID controllers that regulate the bed inlet temperatures. In Fig.1 these PID controllers are within the reactor block.  Fractionator MPC adjusts the set-points of the product withdrawals to control the 95% boiling point temperatures. Similar to the reactor MPC, the PID flow controllers that regulate the product withdrawals are within the fractionator block in Fig.1. The 95% boiling point temperatures are estimated by an on-line soft sensor that uses an empirical model that predicts the 95% BPs from tray temperatures and the column pressure readings. The model has been built using historical data and is updated on-line when the laboratory distillation assay measurements become available.

Outer MPC's task is to control the optimal product disribution at values determined by steady-state optimization. It accomplishes this by manipulating the set-points of inner MPCs. Specifically it adjusts the reactor bed temperature set-points and the 95% boiling point temperature set-points of the fractionater. Reactor MPC uses the dynamic model developed in our earlier work [1]. This model can predict the reactor bed temperatures and the product disribution. Steady-state optimization makes use of the steady-state predictions of this nonlinear model.

Steady-state optimization and control cycle is implemented in a moving horizon pseudo-steady state fashion to handle potential plant-model mismatch and unaccounted disturbances. At the end of each steady-state optimization only a fraction of optimal set-point changes is implemented by MPCs. Using feedback measurements the models are updated and steady-state optimization-control cycle is repeated. 

1. H. Sildir, Y. Arkun, B. Cakal, D. Gokce, E. Kuzu A, ?Dynamic Non-Isothermal Model for a Hydrocracking Reactor: Model Development by the Method of Continuous Lumping and Application to an Industrial Unit? J. of Proc. Control (in review).

Fig.1. Hierarchical Control Structure for the Hydrocracking Plant.