PetroChina Co. Ltd. improved production safety, realized greater oil production, and saved energy costs after implementing model-free adaptive (MFA) control on its steam generators in Liaohe oil fields.
Fig. 1 illustrates a typical enhanced oil recovery (EOR) steamflood in which injected high-pressure steam improves recovery from a reservoir containing heavy oil. In this process, the quality of the steam generated is difficult to monitor and control.
For optimal oil recovery, steam dryness should be about 72%. If the steam is too dry, energy is wasted. But on the other hand, oil recovery is inefficient if the steam is not dry enough.
Process, steam dryness
Fig. 2 illustrates a simplified diagram of an oil-recovery steam generator or boiler. In the process, feedwater first enters the economizer to be preheated. It then passes through heating tubes inside the furnace to absorb heat and become steam. Steam generated for oil recovery is wet, as opposed to the 100% dry steam used for power generation.
Natural gas or oil fuels the steam generators. Fuel is burned with a controlled amount of air to ensure complete combustion and minimum pollution. Key process variables that are measured and monitored in the process include water flow (FI), gas pressure (PI1), steam temperature (TI1), steam pressure (PI2), exhaust air temperature (TI2), and furnace temperature.
This article discusses natural gas as the fuel for steam generation, although operators may switch the type of fuel used, depending on the availability.
Steam dryness is the percentage of completely dry steam present in the total steam. The steam becomes wet if suspended water droplets are present in the steam. The droplets carry no specific enthalpy of evaporation.
Steam used for oil recovery does not need to be too dry because generating dry steam is more expensive. Typically, steam dryness for enhanced oil recovery should be around 72% to be most cost effective.
Wet steam generated for oil recovery consists of steam and water droplets. Because the wet steam has the characteristics of both gas and liquid, online measurement of steam dryness becomes a problem.
The industry uses different methods in attempting to measure online steam dryness. For instance, one can measure the water flow, pressure difference, or temperature difference and then calculate the dryness. Another way is the thermal-balance-based method that requires measurement of different temperature zones inside the furnace. All these methods, however, rely on boiler-dependent process models and therefore are not a general-purpose solution.
Dryness control
The critical process variables including steam pressure, temperature, and dryness of the oil-recovery steam generator are difficult to control due to the following:
- Significant changes in the underground reservoir pressure.
- Frequent changes in natural gas pressure.
- Nonlinear process of the steam pressure loop.
- Time variability of the steam temperature loop.
- Large and varying time delays of the steam dryness loop process.
The reservoir pressure causes major disturbances on the steam load. The large disturbances in the natural gas network, frequent changes in gas supply and demand, and the nonlinearity of the pressure and flow loops cause the proportional-integral-derivative (PID) controllers and manual control systems to have various problems resulting in poor steam dryness consistency.
Even during normal operating conditions, most steam generators are under manual control. Steam with inconsistent dryness that is generated and injected into the reservoir results in low oil recovery. Manual control of the furnace with a fixed fuel-air-ratio results in poor combustion efficiency and waste of fuel and electricity.
An automatic system for control of the steam pressure, temperature, and dryness during all operating conditions will increase oil recovery and minimize energy consumption.
Model-free adaptive control
Model-free adaptive (MFA) control developed by CyboSoft, General Cybernation Group Inc., is a patented technology that can be readily embedded into various control equipment. Derivations of the core technology address specific control problems.
Most industrial processes still are controlled manually or by 60-year-old PID controllers. PID is a simple general-purpose automatic controller that is useful for controlling simple processes. PID, however, has major problems in controlling complex systems, and it also requires frequent manual tuning of its parameters when the process dynamics change.
Fig. 3 compares the performance of MFA (top) and PID (bottom) controllers to show how MFA adapts when process dynamics change.
Starting from the same oscillating control condition, the system will continue to oscillate under PID control, while the MFA system will quickly adapt to control conditions.
If both controllers start from a sluggish situation, MFA will control the process faster and better. Better control means improved process stability, higher production efficiency and yield, consistent product quality, and reduced material and energy waste.
MFA controls complex systems without requiring process models. MFA controllers can effectively control tough processes including nonlinear, pH, multivariable, and processes with large time delays. Because there is no model training required, one can launch MFA at any time to immediately control the process.
Once installed, MFA requires no controller tuning and is easy to use and maintain.
MFA controller architecture
The standard MFA controller consists of a nonlinear dynamic block that performs the tasks of a feedback controller. A dynamic block is a dynamic system with inputs and outputs.
The control objective for a controller is to produce an output to minimize the error between the setpoint and the process variable (PV) being controlled. Fig. 4 illustrates the core architecture of a single-input-single-output MFA controller.
The design of the controller uses a multilayer perceptron (MLP) artificial neural network (ANN). The ANN has one input layer, one hidden layer with N neurons, and one output layer with one neuron (Fig. 4).
Within the neural network there is a group of weighting factors (Wij and hi) that can be updated as needed to vary the behavior of the dynamic block. The learning algorithm for updating the weighting factors is based on the goal of minimizing the error between the setpoint and process variable.
Because this effort is the same as the control objective, the adaptation of the weighting factors can assist the controller in minimizing the error while process dynamics are changing.
From another point of view, the artificial neural network based MFA controller remembers a portion of the process data providing valuablnne information for the process dynamics. In comparison, a digital version of the PID controller remembers only the current and previous two samples.
In this regard, PID has almost no memory and MFA possesses the memory that is essential to a smart controller.1 2
Liaohe oil fields
CyboSoft implemented a MFA control and optimization system for steam generators in the PetroChina Liaohe oil fields. The MFA system includes the components as illustrated in Fig. 5.
The basic design includes two high-precision pH sensors for measuring the pH of the feedwater and condensed water. The measurement precision is within ±0.01 pH.
A soft-sensor algorithm calculates the steam dryness using the measured pH values and other process variables. The principle of this method is based on the assumption that the total alkali coming into the system is dissolved in the feedwater, and because dry steam does not include alkali, the total alkali leaving the system should be dissolved in the condensed water.
One can then calculate the steam dryness based on the measured pH values of the feedwater and condensed water.
A multicontroller-based cascade control system controls the steam dryness. The dryness soft-sensor (AC1) calculates the steam dryness online using the outputs from Blocks pH1 and pH2. AC1's output is the process variable (PV) for the steam dryness controllers (AC2 and AC3).
AC3 manipulates the gas flow directly and AC2 provides the setpoint for the inner loop controller FC to manipulate the water flow.
Based on the fuel-air-ratio calculation, the air-flow controller AC4 manipulates the fan speed to provide the proper amount of air for efficient combustion. The combustion is optimized within a special technique developed by CyboSoft.
As shown in Fig. 5, this system uses one soft-sensor and four MFA controllers:
- Steam-dryness soft sensor (AC1).
- Steam-dryness Controller 1 (AC2).
- Steam-dryness Controller 2 (A3C).
- Air-flow controller (AC4).
- Water-flow controller (FC).
The system includes CyboCon MFA control software and CyboMax plant monitoring and optimization software that were installed on a PC running Windows 2000 and connected to the programmable logic controllers (PLCs) through an API interface.
Fig. 6 shows a CyboMax screen that allows the operator to monitor the steam generation operation. The steam dryness soft-sensor is embedded in CyboMax and the MFA controllers are implemented in the CyboCon software.
PetroChina sometimes uses the traditional manual measurement method to validate and calibrate the online steam dryness calculation.
Fig. 6. shows the steam injection monitoring screen.
Results
PetroChina has deployed multiple MFA control and optimization systems for its EOR steam generators in Liaohe oil fields and has seen the following results:
- 3-4% increase in boiler combustion efficiency.
- 17.6% electricity savings.
- Steam dryness measurement accuracy within ±1%.
- Steam dryness control within ±3% in all operating conditions.
- Improved safety and productivity.
Figs. 7a and 7b compare the difference of the control performance when the steam generator is in manual control and in MFA control.
It is interesting to note that when the steam dryness is under good control, variations in furnace, exhaust, and steam temperatures are sharply reduced. The control system design is simplified to focus on steam dryness control so that the furnace temperature controller or steam temperature controller is no longer needed.
In PetroChina's Liaohe oil fields, the first two MFA control and optimization systems have been in service for about 2 years. Another four systems have been in service for several months.
The soft-sensor proved to be accurate and the control systems proved to be effective. PetroChina now is revamping more steam generators with the MFA control and optimization systems. F
References
1.Cheng, G.S., MFA in Control with CyboCon – CyboCon MFA Control Software User Manual, CyboSoft, General Cybernation Group Inc., Rancho Cordova, Calif., 2002.
2.VanDoren, V., et al., Techniques of Adaptive Control, Elsevier Science, Butterworth-Heinemann, Burlington, Mass., 2003.
3.VanDoren, V., "Model Free Adaptive Control - This New Technique for Adaptive Control Addresses a Variety of Technical Challenges," Control Engineering Europe, March 2001.
The authors
George S. Cheng [[email protected]] is chairman and chief technical officer of CyboSoft, General Cybernation Group Inc. He developed MFA control technology. Cheng has a BS, an MS, and a PhD in electrical engineering. He is a member of ISA and IEEE.
Li-Qun Huo [[email protected]] is the deputy general manager of CyboSoft Automation Technology (Beijing) Co. Ltd. Huo has a BS in automation and is a member of the Chinese Instrument Society.