Print this article

Comment

01 April 2004

A Matter of Separation

Adaptive control technique clears the way for GLCC separators.

By Vasudevan Sampath, Shoubo Wang, Ram Mohan, and Ovadia Shoham

Separation technology in the petroleum industry has been progressing slowly for several decades. But increasing operational problems and economic pressures have driven petroleum companies to seek alternative solutions to separation processes using conventional separators.

Conventional separators have seen use for several decades, and end users still use them. A new generation of compact separators called gas-liquid cylindrical cyclone (GLCC) has become increasingly popular and an attractive alternative to conventional separators. Significant advantages of a GLCC are its compactness, ease of operation, smaller footprint, lower cost, and lower weight. The total number of GLCCs installed in the field in 1999–2000 was about 150, and current figures show about 400 GLCCs.

A GLCC separator is a vertically installed pipe, mounted with a downward inclined tangential inlet. The two phases of the incoming mixture separate due to the centrifugal/buoyancy forces produced by the swirling motion. The heavier fluid (liquid) is forced radially toward the wall, and collection is from the bottom, while the gas, a much lighter fluid, moves to the center and flows out from the top.

The liquid carryover to the gas stream is LCO; the gas carry under into the liquid stream is GCU. You can control or avoid liquid carryover and gas carry under percentages by implementing appropriate control systems. Thus, GLCC separation efficiency can vastly improve for large flow variations through suitably applying control strategy.

That is where a design of a new optimal control strategy for minimizing the dynamics of the liquid control valve (LCV) and the implementation of an adaptive control algorithm called EXACT control (an embedded design of the hardware controller) come in.

Math model

To analyze a system as complex as the one for a GLCC separator, you need a mathematical model representing the system. This mathematical model allows for the move to transfer functions in the control system block diagrams for analysis and simulation. For this control system study, the controlling parameters are the liquid level and pressure in the GLCC separator and liquid and gas control valve positions.

In this study, the main focus is on feedback control systems. A pressure transducer measures the pressure, and a differential pressure transducer measures the liquid level. A Highway Addressable Remote Transducer (HART) tri-loop sensor measures the position of the gas and liquid control valve. The liquid-level control can occur by the direct operation of an LCV or by the indirect pressurization of the GLCC separator by operating a gas control valve (GCV). You can control the pressure inside a GLCC separator by the direct operation of a GCV. The level transducer senses the liquid level, and this signal goes to the controller as the process variable. The controller compares the actual liquid level to the set point liquid level and operates the LCV/GCV to control the liquid level. Similarly the controller operates on the pressure signal from the pressure transducer to manipulate the GCV for pressure control.

For a gas-dominated system, you can use the liquid-level control by the LCV. For a liquid-dominated system, a gas control valve on the gas leg can control the liquid level. The integrated control strategy can apply to cases with slug flow condition, as it provides an immediate action on liquid level by the LCV and GCV. For bulk separation applications, liquid-level control by the LCV and pressure control by the GCV, as two independent control loops, is desirable. Optimal control strategy takes place under severe slugging flow and where the control valve dynamics will be limited to increase the life of the control valve.

Based on operating conditions and applications, a user can adopt the following control strategies:

1. Liquid-level control by LCV
2. Liquid-level control by GCV
3. Pressure control by GCV
4. Integrated liquid-level control by LCV and GCV
5. Integrated liquid-level control by LCV and pressure control by GCV
6. Optimal control, liquid-level control by LCV and LCV-position control by GCV
7. Optimal control, liquid-level control by GCV and GCV-position control by LCV

The first six of the existing feedback control strategies mentioned above have successfully undergone design, implementation, and testing.

The following are the general steps involved in the development of a control system and dynamic simulator:

Transformation of physical system into block diagram: The GLCC separator consists of electric, pneumatic, and electromechanical devices. Devices such as control valves, actuators, pressure transducers, level transducers, and pneumatic transmitters convert to equivalent block diagram models. A certain degree of approximations apply to obtain a simple yet useful mathematical representation of the real system.

Mathematical model for schematic: Linear model of each subsystem or component occurs by making reasonable and realistic assumptions. Derive the Laplace transform of each subsystem from the linear differential model.

Transfer function representation of complete subsystem: To evaluate the system response, the mathematical model represented as a block diagram in the previous step transforms to transfer functions. A transfer function that relates its input to its respective output represents every single subsystem. In a feedback control system as complex as this, the closed-loop transfer function is too difficult to obtain, but for this study the open-loop transfer function goes into consideration.

Analysis and design: In this step you can perform "root locus analysis" on the open-loop transfer function. This is a very powerful tool for analyzing and designing control systems.

Root locus analysis provides a graphical method of plotting the locus of the roots in the S-plane, as the system gain is varied over a complete range of values. You can locate the roots corresponding to a particular value of system parameters on the locus. Using this root locus technique, you can design the controller parameters, such as P, I, and D, for the desired characteristic response. Once the complete system has a single transfer function, you can analyze and design characteristics, such as transient response, steady-state error, stability, and sensitivity.

Control strategies

The simple control strategies and integrated control strategies listed above primarily consist of simple proportional, integral, and derivative (PID) controllers working on the process variable, which you assume is independent. However, these strategies are not robust enough to handle a wide range of operating conditions. They may fail because of the system dynamics caused by the interaction of pressure and the liquid level inside the GLCC separator. This factor was not a consideration in the design of these controllers. The optimal control strategy is capable of handling this interaction of liquid level and pressure, minimizing the operating pressure and providing unique valve positions for a given flow condition.

There are two optimal control strategies developed: optimal control with an LCV controller as the master controller and optimal control with a GCV controller as the master controller.

GLCC optimal control strategy with LCV

The objective of this optimal control strategy is to minimize the pressure, maximize the liquid flow for any given inflow condition, and optimize the control valve performance. You can achieve this by controlling the liquid valve at an optimum position. This optimum position serves as the set point for the GCV controller. The most important application of this control strategy is that it can operate in a wide range of liquid and gas flow conditions with minimum GLCC pressure.

The optimal control strategy has two different controllers. You can control the liquid level using the LCV controller, and the GCV controller controls the optimum position of the LCV. The main drawback observed and reported on this optimal control strategy is the dynamics in operation of the LCV. Because the LCV is the master valve operating on the liquid level and it also the load-bearing valve, due to its dynamics it has a reduced life expectancy. This is one of the most important reasons for the development of the second optimal control strategy.

GLCC optimal control strategy with GCV

The objective of this control strategy, in addition to minimizing pressure and maximizing flow through the control valve, is to minimize the dynamics of the LCV. The main load-bearing control valve is the liquid control valve and not the gas control valve. Therefore it is preferable to have a GCV with more dynamics than the liquid control valve. This optimal control strategy aims at achieving this.

In this control strategy the liquid level is controlled by the backpressure using the GCV controller as the master controller and the LCV controller, acting on the position of the GCV, as the slave controller. The master controller measures the liquid level with a differential pressure transducer as the process variable. The master controller maintains the liquid level to the user set point liquid level by operating the GCV. The HART tri-loop senses the position of the GCV, and the signal feeds to the slave controller. The slave controller maintains the position of the GCV around the set point (optimum position), thereby minimizing the pressure inside the GLCC separator and at the same time maintaining the liquid level.

The optimum position of a GCV by definition is the position of the GCV that reduces the dynamics of the LCV and results in a very low-pressure drop across the LCV. This optimum position is the set point for the LCV slave controller. If the GCV changes its position from the optimum position, the slave controller operates the LCV to bring the GCV back to its optimum position.

The master control loop consists of the gas control valve on the gas leg, a liquid-level sensor (such as a differential pressure transducer), and a PID controller. The liquid-level signal sent to the PID controller stays around the user set point by correspondingly opening or closing the GCV by this controller.

Thus, the intention of the master control loop using the GCV is to maintain the liquid level around its set point. The slave control loop consists of the liquid control valve in the liquid leg, gas control valve position sensor (the HART tri-loop valve position indicator), and PID controller. The gas control valve position sensor sends a signal to the PID controller, which drives the LCV.

The central concept in this control strategy is to maintain the liquid level with minimal pressure inside the GLCC and by smooth operation of the LCV.

Implementing adaptive control

The performance of GLCC separators can improve by eliminating liquid overflow into the gas leg or gas blow out through the liquid leg, using a suitable control strategy. Most of the hardware controllers manufactured these days are microprocessor-based controllers. One of the most important features of a hardware controller exploited for a GLCC separator control system is adaptive control. The flow variations upstream of the GLCC may cause abrupt changes in liquid level, which might be difficult for a conventional controller with a single PID setting.

To achieve better control of the liquid level or the pro-cess variable to accommodate the different flow conditions, you can use an adaptive control technique. Adaptive control algorithms in the hardware controller continuously monitor the process variable and update the controller PID parameters based on the dynamics of the process variable. The hardware controller chosen for testing on the GLCC separator platform for various control strategies was a Foxboro 762CNA microcontroller with an EXACT adaptive tuning algorithm.

This controller comes from analysis of the transient response of the closed-loop system to set-point changes or load disturbances and traditional tuning methods. The idea behind the algorithm is a pattern recognition approach. This EXACT adaptive control system comes from the determination of dynamic characteristics from a transient, which results in sufficiently large error. If the controller parameters are reasonable, you will obtain a transient error response. Heuristic logic can detect whether a proper disturbance has occurred and can detect the peaks E1, E2, and E3, and period Tp. The estimation process is simple. It is, however, based on the assumption the disturbances are steps or short pulses.

EXACT control requires that the loop closes with a controller that gives a reasonably stable response.

In case of a GLCC separator, the designed PID settings offer reasonable control until the process upset takes place. Even in the worst case scenario the process is still under control, but with larger dynamics of the process variable and control valve positions.

The tuning procedure requires prior user parameter inputs, including controller parameters such as P, I, and D. It also requires information of the time scale of the process. You use this to determine the maximum time the heuristic logic waits for the second peak. You also need some measure of the process noise to determine that a disturbance has occurred and to set the tolerances in the heuristic logic.

The damping and overshoot determined by the algorithm compares with user parameter input, damping, and overshoot. The decision on whether the controller parameter has to be newly calculated comes from the comparison of damping and overshoots from the process with those values from the user parameters.

If the damping ratio and overshoot calculated from the process transients are less than those values that the user inputs, then the algorithm does not calculate new P, I, and D.

If the damping ratio and overshoot values calculated from the process transients are more than the values input by the user, then new P, I, and D values are calculated and used by the controller.

Findings

Based on the results obtained from dynamics simulation of the new optimal control strategy and experimental investigation of control strategies and adaptive control implementation:

• The new optimal control strategy can handle any combination of gas and liquid flow disturbances, including slug flow. However, the set-point position for the gas control valve may not be optimum for all flow conditions. Extreme flow conditions may cause large dynamics to liquid level and control valve positions. You can control the dynamics by appropriately changing the set-point position of the gas control valve.
• The dynamics of the load-bearing control valve, which is the liquid control valve, reduces drastically. For every flow condition there is a unique position for this control valve, and the transition from one position to another for change in liquid flow rate is smooth. Thus, the new optimal control strategy with adaptive control increases the life of the liquid control valve.
• Using adaptive control, the GLCC separator system can now start with a designed PID value and adapt itself to various flow conditions, minimizing the dynamics of the process variable and control valve. DT

Behind the byline

Vasudevan Sampath, M.S.; Shoubo Wang, Ph.D.; Ram Mohan, Ph.D.; and Ovadia Shoham, Ph.D., are members of the departments of petroleum and mechanical engineering at the University of Tulsa.