November 2020

Special Focus: Process Controls, Instrumentation and Automation

Advanced process control for unsteady-state operation: A lesser-known technique

A goal of all process industries is to run the plant in a steady-state zone of operation, improving optimization and control, and producing more benefits.

A goal of all process industries is to run the plant in a steady-state zone of operation, improving optimization and control, and producing more benefits. When steady-state operation is achieved, process plant operational outcomes remain at their desired levels or in predicted zones. Unfortunately, disturbances that try to deviate a process from its desired steady state can appear continually, and some may be caused by ambient changes or input conditions, which may vary due to changes in output demands.

Such disturbances can create a driving force to divert a process from its steady state of operation to an unsteady state, increasing losses in process value. Often, human interference is required to return operating conditions to steady state. However, some processes perform better in an unsteady state. In literature, Matros, et al. have demonstrated the advantage of deliberately operating catalytic reactors under unsteady-state conditions to achieve improved performance.

Handling unsteady-state process operation

Advanced process control (APC) applications have become normal for refining units. From regulatory control to advanced regulatory control (ARC), to conventional APC and multivariable predictive control (MVPC), MVPC technology has become the main workhorse of refinery process control and optimization. Conventional APC software and models are designed for steady-state operating conditions. In theory, a system or process is in steady state if the variables (state variables) that define the behavior of the system or process—process gain, time to steady state, process dead time, etc.—do not change over time. This article describes a superior and less-known method of APC at the hydrogen generation unit in the Train III complex in the Mumbai Refinery of Bharat Petroleum Corporation Ltd. (BPCL) that can handle the unsteady state of process operation.

APC was installed for hydrogen management, as basic APC was unsuccessful according to plant requirements. Hydrogen management was unsuccessful using a conventional APC philosophy due to a lack of fixed gain, dead time and time to steady state. The system became unsteady intermittently, leading to flaring. In the BPCL Mumbai Refinery, APC was implemented in the hydrogen generation unit (HGU) with the prime objective of hydrogen management between producer and consumer plants. The conventional design of the HGU APC was insufficient in determining the hydrogen requirement, subsequently leading to cyclic flaring of hydrogen dioxide (H2) and a drop in HCU system pressure.

An innovative method has been applied to the traditional APC based on a root cause analysis of its ineffectiveness for hydrogen management. This method has added a new dimension in APC and proven an effective tool for unsteady-state operation and where process gain, dead time, time to steady state, etc., are changing. Correct steady-state prediction is the key to APC performance. Since unmeasured disturbance variables cannot be considered in APC, the prediction of the steady state of controlled variables would be incorrect. Conventional APC acts on manipulated variables (MVs) only when the steady-state value violates the low and high limits of control variables.

Case study

Over last 60 yr, the BPCL Mumbai Refinery has undergone many changes in configuration, processing several types of crudes, upgrading technologies and continuously modernizing and improving performance. APC has been implemented in all process plants, including the sulfur recovery unit (SRU), steam boiler unit (boiler house) and nitrogen generation unit. APC controllers were implemented on the HGU in 2012 to manipulate hydrogen production to meet demand and stop hydrogen flaring from the grid. FIG. 1 shows the network of hydrogen grid system with producer and consumer plants.

FIG. 1. The network of hydrogen grid system with producer and consumer plants.

Environmental restrictions, updated transportation fuel specifications and increased processing of heavier, more-sour crudes were leading to substantial increases in refinery hydrogen consumption. The APC controller installed for the HGU in 2012 was very effective in controlling all parameters, such as steam/carbon ratio, reformer outlet temperature, etc., but it failed to stop the hydrogen flaring problem from the hydrogen grid. The main reason for the failure was that the system can become unsteady intermittently and conventional APC is incapable of handling an unsteady state of operation.

The system state variables change with time. Due to some system disturbances, the time between producer plant (HGU) feed change and the effect of HGU feed change on consumer hydrocracker unit’s (HCU) system pressure is not a fixed variable. In the trends shown in FIG. 2, the red color indicates the HGU feedrate, and the HCU system pressure is represented in black. It can be seen that the effect of HGU feed change on HCU system pressure varies in different situations, meaning that process gain is not fixed, but changes with time.

FIG. 2. The red color indicates the HGU feedrate, and the HCU system pressure is represented in black.

It has been observed that the model relation parameters (process gain, dead time, time for stabilization, etc.) between the manipulated variable (HGU feed) and the control variables [(CV), HCU system pressure and hydrogen grid pressure] change with time and situation. A conventional steady-state APC model does not work for this system.

While the HCU system becomes unsteady due to external or internal disturbances (continuous changes in hydrogen demand, continuous changes in crude composition in upstream CDU plant, hydrogen grid pressure fluctuation due to changes in demand in other consumer plants, etc.), a significant pressure fluctuation in the HCU system pressure has been observed and leads to intermittent hydrogen flaring from the makeup gas compressor suction—an average of 3 tpd–4 tpd of pure hydrogen was flared from the makeup gas compressor suction knockout drum (KOD). From the process study, it is required in APC to maintain slow MV/CV relation for normal operation, but very fast action is required during flaring or unsteady-state operation. How much action is needed by the MV or how long the MV requires to move could not be determined due to the unsteadiness of the system behavior. So, since the system is unsteady at the time of flaring and traditional APC does not work or is not designed for unsteady-state operation, running hydrogen management in the APC became a significant challenge. Additionally, it was not possible to manually conduct hydrogen management without APC.

Finding a solution

After brainstorming and trial and error, one nontraditional method was created to control the unsteady-state situation. From the feed vs. hydrogen production trends, it was clear that MV/CV gain, dead time and time to steady state are all variables for that system. APC was taking unnecessary (or very slow) MV moves to control the HCU system pressure (CV) since the models are not correct. To control the HCU system pressure, HGU feed is fluctuating unnecessarily and increasing the HCU system pressure fluctuation, which can be seen in FIG. 3.

FIG. 3. The HGU feed is fluctuating unnecessarily and increasing the HCU system pressure fluctuation.

A change in methodology for MV move calculations was required to reduce the fluctuation of hydrogen plant feed. A methodology was invented in-house to restrict the unnecessary movement of hydrogen plant feed MV to control HCU system pressure. In this calculation, MV moves according to the slope of the HCU system pressure trend. It was observed that the HCU system pressure does not stay steady for long; when it is steady, an MV move should be immediately stopped, otherwise it will create further disturbance on the CV. In conventional APC, an MV move is calculated according to the steady-state value of the CV, low and high operator limits of the CV and the model parameters (gain, steady-state time, dead time, etc.). Since the correct model cannot be derived, MV movement has been restricted by the output of a calculation, which is calculating the slope of HCU pressure trend every minute (FIG. 4).

FIG. 4. Calculating the slope of HCU pressure trend every minute.

If the HCU system pressure slope is changing from negative to positive or nearer to zero, NHGU feed will not increase further, and vice-versa. That means that APC action on the MV (i.e., HGU feed) has been sufficient to achieve the desired HCU system pressure. So, in every point where the slope is changing, MV moves stop if the CV desired value is not achieved after the allowable time. This slope-based calculation does not allow an MV to take continuous movement until the CV steady-state value moves within the CV low and high limits.


How can the MV movement be stopped when the slope is nearer or equal to zero? This is achieved by a simple trick, by providing a pseudo or fake information to the APC controller. As discussed, if the steady-state value of the CV is within the CV high and low limits, APC does not take any action on the MV to control the CV (unless no optimization has been provided on the CV). The calculation output of the slope calculation automatically provides low and high CV limits far from the CV steady-state value to stop the MV movement when the slope is nearer or equal to zero. A set-off calculated dynamic CV limits system has been generated rather than an operator-provided fixed CV limits system. By incorporating this slope of CV trend-based MV movement APC, we have successfully run APC during unsteady state of operation and the zero hydrogen flaring from hydrogen grid. This resulted in an approximately 3 tpd–4 tpd reduction of pure hydrogen flaring from the system. This method is appropriate not only for hydrogen management, but also can be implemented to handle any unsteady state of operation where state parameters are not fixed and varied with time, and where conventional APC has failed to control the system. HP

The Authors

Related Articles

From the Archive



{{ error }}
{{ comment.comment.Name }} • {{ comment.timeAgo }}
{{ comment.comment.Text }}