400188 Advanced Control of Wild Loads to or from Multiple Units

Wednesday, April 29, 2015: 4:00 PM
13A (Austin Convention Center)
Dustin Beebe, ProSys, Inc., Baton Rouge, LA and Darwin Logerot, ProSys, Inc., Houston, TX

Advanced Control of Wild Loads to or from Multiple Units

Darwin Logerot, ProSys, Inc., Houston, TX

In many situations in refineries and chemical process units, wild loads must be handled either by common production facilities, utilities generation, vent or waste stream handling systems. These systems must run continuously at a high reliability level in order to avoid environmental releases, lost production and lost efficiency. Too often, changing loads or partial shutdowns can cause these issues because the control systems in place do not handle the rapid changes efficiently. 

This situation is applicable in:

1)      Parallel sulfur recovery units in a refinery

2)      Steam generation facilities: multiple boilers, some steam generated by process heat, with variable overall load

3)      Air compression and drying facilities: multiple compressors, variable load

4)      Hydrogen or Syngas plants

5)      Wastewater treatment

There are usually multiple systems in place to handle the loads. This gives rise to a number of process control challenges, including:

1)      Load balancing among the various handling or generation units

2)      Response to changing loads, sometimes a sudden step change in load (total feed or delivery load cannot be controlled)

3)      Response to variable compositions

4)      Response to a partial system shutdown (rapid load increase for the remaining systems)

5)      Dealing with one or more system at its operating limit (constraint)

This presentation will discuss the ramifications of these challenging control problems along with a successfully implemented advanced control methodology used for each. The control scheme utilizes advanced regulatory control (ARC) without using model-predictive control (MPC).  This means that no external software is required, no process models are needed, and the process response to control action need not be linear.

Extended Abstract: File Uploaded