Using Model Predictive Control to Reduce Steam Usage in Distillation Columns

A continuing theme to several of these blog posts is how process manufacturers are looking for ways to improve energy efficiency in these times of high energy costs. One way to do this is to optimize the steam required for a distillation process.
I caught up with Pete Sharpe whom you may recall from an earlier post on reducing costs of APC projects using pre-engineered applications. Pete has recently completed some work for a specialty chemical manufacturer that wanted to improve the performance of the distillation columns by decreasing the steam required and decreasing the reflux flows to the columns.
Pete worked with the process engineers to apply model predictive control (MPC) technology found in the SmartProcess Distillation Optimizer. This application is one of the pre-engineered SmartProcess applications Pete described in the earlier post.
The distillation process is a classic multivariable problem with control variables, manipulated variables and constraint variables.

Using model predictive control, the column can be controlled and operated as a unit instead of a collection of loops.
In addition to reduced operator load, the process engineer identified 400 lb/hour savings in steam on one of the columns and close to 900 lb/hr on the first column where the Distillation Optimizer application was implemented. With a cost for 135 psi steam of $5 per klb, this translates into energy savings of more than $50,000 USD for these particular columns. This savings adds up as all of the distillation columns on site are converted over from multi-loop control to MPC-based control. Steam reductions are a result of lower reflux flows that have been reduced by about 20%. While this change increases the average overhead impurities as is expected, it is well within specifications.
Now that the Distillation Optimizer has demonstrated stable results on two of the columns, Pete is working with the process engineers to implement it on the remaining columns over time. Beyond better performance and increased efficiency, the best measure of the success to date has been operators leaving the MPC control on more than 90% of the time. This is one of the true tests according to Pete and the Advanced Automation Services team.