(376bu) Optimization of Distillation Processes

Distillation is the primary technology used in industry for liquid separation and purification. This is because it has a proven track record and is often the most cost effective method, particularly in crude separations. Even though it is the economical choice, distillation is very costly both in energy and capital. With new column designs a significant amount of time and effort is consumed with optimizing the tradeoffs between capital and the long term energy costs to get the column positioned for the lowest economic investment for the life of the asset. In existing processes we continue to make empirical adjustments to further push columns to reduce costs and help the company improve the bottom line.

In most cases, column installations are complex and have multiple variables that can be manipulated to optimize the column and each parameter needs to be optimized. Traditional methods of optimization involve performing parametric studies. It is difficult to arrive at the optimal condition of all process parameters that will minimize/maximize the desired objective(s) while fully meeting the design and operational constraints. This makes it a challenge to get to the most optimal set of conditions for the separation. However with the implementation of a rigorous optimization algorithm it is possible to explore all the possible scenarios required to get to the true optimum. The optimization of a distillation column involves both continuous and discrete variables. In this work, we used the Mixed-Integer Non Linear Programming (MINLP) optimization formulation and systematically optimized the performance of different distillation processes using the non-dominated sorting genetic algorithm II (NSGA-II) available in the MATLAB^® global optimization toolbox.

To demonstrate the improvements the MINLP optimization approach can have over parametric methods, several case studies were completed pitting an experienced column designer against the computer. In each case the results indicated that the MINLP would have resulted in a more economical design, especially as the columns became more and more complex. This shows the long term advantages implementing this optimization strategy can have over traditional methods and improvements to the company cost position.