Chemical Engineering MagazineChementator Briefs
RE-free magnets Tetrataenite (L10-FeNi) is a promising alternative to rare-earth-…

Comment PDF Automation & Control

Dynamic Modeling for Steam System Control

By Ali Bourji, David Ballow and Martha Choroszy, WorleyParsons |

One of the most energy-intensive utilities for many facilities in the chemical process industries (CPI) is the steam system. Traditionally, steam-use optimization has centered on efficient heat transfer and eliminating waste [ 1]. Further optimization can involve a broader look at how steam supply and consumption interact dynamically throughout a large complex. This type of optimization often results in increased interconnectivity and interdependency.

Many CPI facilities have a central steam-production area containing boilers and boiler feedwater treatment, as well as additional steam generators scattered throughout the facility (for example in the petroleum refining sector there are ethylene and catalytic cracking units). If a facility is built in several stages, as is often the case, steam generating systems may be separated by considerable distances. Over these distances, the stability of the integrated steam system could be jeopardized by inappropriate control strategies. How should one go about setting up a control strategy and verifying that it is stable and appropriate for a particular complex?

 Figure 1. An effective control
strategy can be broken down
into four stages

Steady-state modeling and steam balances only show the endpoints of system behavior. Dynamic modeling fills in the space between these endpoints providing a more complete analysis. With potentially billions of dollars in capital investment depending on a reliable supply of steam, employing dynamic modeling during the design development of integrated systems is worth the extra effort. This article breaks down the task of setting up a control strategy into four basic steps (Figure 1).


In order to properly control any system, a thorough understanding of the interactions within the system (the system behavior) is essential. Understanding system behavior begins with gathering as much information as possible about a given process or facility. Ask some fundamental questions, such as those outlined in the box above.

For an existing operational facility, there is no better resource to answer these questions than the senior operations staff. They have direct knowledge of how the system behaves in realtime during real upsets under real conditions. Defining these upsets will become an essential input to dynamic model development.

For new facilities, this investigative exercise may still prove valuable for the complex. It is still necessary to draw on the knowledge of experienced operators who have run similar systems in the past. Supplementing their knowledge and experience with the appropriate process engineering and modeling techniques will allow for sufficiently accurate system emulation.


In the typical workflow of modern process design, a steady-state model is usually developed to facilitate the creation of utility balances and to study various operating cases. A wide array of modeling software has been developed [ 2] and is in use within the CPI. When choosing the platform for the steady-state model, keep in mind the potential for running the model dynamically.

Steady-state modeling is essential, but a plant will never truly achieve steady state. To achieve a reliable and stable steam supply throughout the complex, the fully integrated steam system must be analyzed in a dynamic state to understand the probable interactions between the system components. Operating facilities are generally not able to risk a major shutdown in order to test system responses to the upsets of interest. The next best option is to model the system dynamically. The dynamic model becomes a testing platform on which control concepts can be proven and adjusted if necessary [ 3].

Dynamic process simulations fill the gap between different steady-state operating cases, showing a more complete picture of system behavior. Using the knowledge gained during the investigative process, a model can be constructed that will be useful for testing the system under changing conditions.

Example. Suppose the system to be modeled consisted of three sources and two users of steam. The steady-state model flowsheet may look like Figure 2. In this example system, two sources of steam exist on one end of a main steam header, while a third source sits close to the process user areas.

This same simulation flowsheet can be adapted for dynamic evaluation by adding some basic controls as shown in Figure 3.

Using this source-sink model of a steam distribution header, some of the aspects of the system behavior can be explored. The system may have two design cases that result in different steam balances. The steady-state model gives a snapshot of what is happening when everything is stable. Table 1 shows what these data may look like.

 Figure 2. This example is used to demonstrate a steady-state model flowsheet (Pri used in the figures stands for primary)

 Figure 3. A flowsheet that is ready for dynamic mode is illustrated here
Table 1. Steady-State Results for Two Operating Scenarios
Name Normal operation Alternate operation
  Pressure (psig) Mass Flow (lb/h) Pressure (psig) Mass Flow (lb/h)
Source 1 300.0 125,000 300.0 50,000
Source 2 300.0 75,000 300.0 50,000
Source 3 299.6 50,000 300.3 50,000
Total production 299.1 250,000 299.8 150,000
To users in A 298.0 100,000 298.7 100,000
To users in B 298.7 150,000 299.7 50,000

Switching to a dynamic analysis gives a more complete picture of the system behavior in the time between the two operating cases. At this point, the previous consultation with operators who understand the system comes into play. Using the knowledge gained from the operators, the design engineer must account for the time factors involved in transitioning from normal to alternate operation. For this example, users in Area B are reducing demand to reach the alternate operating mode. Through consultation with the operators, it becomes clear that this demand reduction normally takes place over a 2-min period. Figure 4 is a graph of what this may look like in a dynamic simulation.

Starting from a steady state corresponding to normal operation, the demand reduction begins at 120 seconds. The User B demand is ramped steadily downward for the prescribed two minutes. The source-steam flow controllers initiate a corresponding reduction in steam production to maintain the system balance. This production decrease is typically achieved through some type of master pressure controller. The master pressure controller senses the steam distribution-header pressure and drives the steam producer to increase or decrease production to maintain the desired header pressure.

A major limiting factor in controlling steam header pressure is the response time of the steam generating source. These sources respond very slowly due to the mass of water and steel that must absorb and release energy to affect a change in the system flow. This thermal inertia can cause differing response times on flow increases and decreases at different capacities.

 Figure 4. The dynamic behavior of steam sources during transition, as discussed
in the example, is shown here
 Figure 5. This plot shows the dynamic response of main header pressure as
given in the example

In this example, the Source 1 and 2 characteristics are such that their response is limited to a rate of 10% of total capacity per minute. Figure 5 shows a plot of the pressure at the main sensing point for Sources 1 and 2.

Again starting from steady state and introducing the disturbance at 120 seconds, the header pressure initially rises due to the slow response time of Sources 1 and 2. The sluggish nature of these steam sources also contributes to the overcompensation and severe drop in header pressure. The sources are eventually able to compensate for the change in steam demand, but a large oscillation has been experienced in the interim. These types of oscillations can cause process upsets throughout a large facility. Note that this example is for illustrative purposes only and some of this lag can be attenuated with careful tuning.

A validation step is essential to verify the model’s ability to emulate the system behavior. Typically, a model review is performed involving key personnel from engineering and operations departments. The information gained during the investigative step regarding common upset events is particularly useful at this stage. Ideally, the model is put through a series of known scenarios, and the resulting predicted response is compared to the known response. Any required fine tuning can be implemented, and the model can be used for subsequent analysis with a reasonable degree of confidence. The model can also be used to predict system behavior under new conditions.

Once validated, the model will provide valuable insight into system behavior and interactions. It is the high degree of interconnectivity in facilities that results in greater efficiencies, but can lead to unexpected interactions. A well-constructed dynamic model can lead to the discovery of these interactions and will allow a facility time to develop a plan for controlling the integrated system.

If the model is emulating an existing system, step testing can be used to develop actual system behavior data. Incremental changes of a tolerable magnitude can be made during the operation of a facility. The magnitude of the change need only be greater than the noise band of the target dependent variable. Proper planning and preparation is essential for this type of testing, since there is a risk of upsetting an operating unit. All test parameters must be documented and agreed upon prior to testing.



Using the developed and validated system model, a master control strategy can be developed. Using engineering judgment and insights gained throughout the model development and testing, a preliminary control strategy is assembled. Some key considerations in such a strategy include those shown in the box above.

The control strategy will likely be a combination of traditional proportional-integral-derivative (PID) controllers and logic triggered actions. Steam load shedding is an example of logic triggered actions. Load shedding can be implemented if major steam users need to be shed in order to recover from an upset scenario. The input of experienced operations personnel is essential in developing a ranking of the major steam users that can be shed. This ranking will allow the development of steam shed actions resulting from steam-header pressure loss [ 3].

Once the preliminary control strategy is established, it can be incorporated into the dynamic model. Confidence in the selected controls will be gained by rerunning the previous model cases using the tuned model and planned control strategy. Perturbing this model using upsets from model development will show the effectiveness or ineffectiveness of the proposed control scheme. Initial tuning parameters can be developed along with any adjustments to sensing locations and final control-element characteristics (such as control valve sizes). The dynamic model can then be used to predict reactions to more severe upsets that are not reasonable to attempt in an operating unit.



Implementation of the control scheme is the final step. All of the modeling, checking and rechecking should result in confidence in the new master-control scheme and provide useful predictive data for implementation in a new facility or for navigating an existing facility’s management-of-change (MOC) procedures.

The planned scheme must first be documented in all relevant engineering documents. Piping and instrument diagrams, process flow diagrams, control narratives and instrument loop diagrams are examples of these documents. Once all documentation is in place, a thorough review will take place to ensure nothing has been overlooked in either hazards or operability. This is typically done within the framework of an established plant or project hazard-analysis procedure, such as an MOC procedure or a hazard and operability study (HAZOP).

Prior to activating the new control scheme, all components, including the software components, must be tested to ensure proper functionality.

Edited by Dorothy Lozowski


1. Jaber, David, McCoy, Gilbert A., and Hart, Fred L., Follow these Best Practices in Steam System Management, Chem. Eng. Prog., December 2001.

2. Currie, Jonathan, and others, “Steam Utility Systems are not ‘Business as Usual’ for Chemical Process Simulators”, AIChE Archived Presentations, March 15, 2011.

3. Bourji, Ali, Ballow, David, and Choroszy, Martha, Find Benefits in Automating Boiler Systems, Hydrocarbon Proc., October 2011.


Ali Bourji is a senior technical director at WorleyParsons (6330 West Loop South, Bellaire, TX 77401; Email:; Phone: 713-407-5000). Bourji received his B.S.Ch.E. and M.S.Ch.E. from the University of Houston and his Ph.D. from Lamar University. He is a professional engineer and a member of AIChE and AFPM (formerly NPRA). Bourji is the author of numerous publications and serves on the Chemical Engineering Ph.D. Advisory Council at Lamar University.



David Ballow is a principal process engineer at WorleyParsons (6330 West Loop South, Bellaire, TX 77401; Phone: 713-407-5000) and is a professional engineer. He received a B.S.Ch.E. from Louisiana Tech University and is a member of AIChE.



Martha Choroszy is a chief process engineer at WorleyParsons (6330 West Loop South, Bellaire, TX 77401; Phone: 713-407-5000). She received a B.S.Ch.E. from the Massachusetts Institute of Technology and an M.B.A. from Tulane University. She is a licensed professional engineer in Texas and a member of AIChE and NFPA. She is the author of numerous publications, a recipient of Tulane’s Allen Vorholt Award and has served as a Blue Ribbon Panel Member to define the national agenda for the U.S. Core Combustion Research Program.


Related Content

Chemical Engineering publishes FREE eletters that bring our original content to our readers in an easily accessible email format about once a week.
Subscribe Now
Trinseo Digitizes Control System Migration Projects to Achieve Fast ROI
Purdue University Saves $400,000 Annually with Local Vacuum Networks
Bag filter Housings/Vessels
Innovative Backwashable Media Filter
Automated Vertical Tower Filter Press

View More

Live chat by BoldChat