Blending gasoline is a critical refinery operation. Texaco implemented the system, OMEGA, in 1980 on personal and large computer systems. Computers were being installed at Texaco refineries in 1960’s and were mostly used for accounting purposes, data acquisition, process control, and refinery modeling. At this time, compositions of gasoline blending were being developed by trial and error, experience, and average response tables.
Computers eventually became easier to acquire. Mathematical models were developed that helped to figure out the characteristics of each blend. It did so based on the properties of available stocks and blend proportions suggested by the blender. Then, by the 60’s computers advanced quit a bit. Linear programming models were being solved by linear programming introduced by the refiner engineers. A process optimization program for nonlinear optimization that solved nonlinear programming problems was developed by IBM, called POP II. Then, immediately after, the GOP blending optimization system was developed that used the POP II system.
In the 1980’s it was found that the GOP system was not being used by all the refineries routinely and many problems inhibited its use. The blends rarely met their correct specifications, due to that fact that the blending model in GOP was not accurate. The POP II algorithm frequently stopped at solutions that were not feasible, and it was also very slow. The inaccurate results stopped at different values which showed the starting points were often different as well, which decreased the confidence in the results.
In 1982 OMEGA was beginning to be developed by Texaco. First thing they did was, replace the optimizer that was POP II with GRG2. They then had to improve the mathematical blending model, which calculates the output blend properties, other constraint values, and the value of the objective function. In this step they also...