A REALISTIC DYNAMIC MODELING APPROACH TO SUPPORT LNG PLANT COMPRESSOR OPERATIONS

Authors: Jihong Wu - KBR, USA Jeffrey Feng - KBR, USA Surajit Dasgupta - KBR, USA Ian Keith - Woodside Energy, Australia Publication / Presented: LNG journal Date: October 2007

ENGINEERING FORUM

A realistic dynamic modeling approach to support LNG plant compressor operations
Jihong Wu, Jeffrey Feng, Surajit Dasgupta, KBR, USA, and Ian Keith, Woodside Energy, Australia In LNG plants, refrigeration compressors are among the most critical components in terms of both capital investments and operational costs. Safe, stable and sustained operation of the refrigeration compressors represents a key component of maximizing onstream time and production. Dynamic simulation has been increasingly used in the various stages of the LNG process life cycle, to perform design optimization, identify production limiting constraints, and validating the dynamic, or time-dependent, responses of the process [1],[2]. To accurately and reliably predict the dynamic behavior of real world systems, the based dynamic on model as-built has to be supplemented with accurate input data equipment performance. Once a dynamic model is developed, it should be validated against design and actual operating data to ensure the accuracy of the modeling. While model validation in steady state is quite routine, it is a less frequent practice to validate the model dynamically, mainly because of the challenge of obtaining reliable dynamic data. Also, the controllers used in the dynamic simulation models for general purpose studies are often simplified based on basic control strategies, and thus do not represent the functionality of actual field controllers. The combined challenges of dynamic validation and the lack of full representation of the controls, has been a hurdle for the routine application of dynamic simulation for field and operations support. These challenges also point to the opportunities for applying dynamic simulation to evaluate real world LNG compressor systems that can experience a wide range of operating conditions and whose controls have to be designed to handle both mild disturbances and emergency situations. KBR to provide support for the operation of the refrigeration compressors in an LNG facility operated in Australia by Woodside Energy Ltd. After evaluating the accuracy requirements, Woodside and KBR decided to use a direct controlhardware linked simulation approach instead of conventional software emulation, to accurately simulate the functionality of field-installed controllers. During the study, an integrated software-hardware solution was developed by linking a rigorous plant dynamic model to a control vendor supplied controller emulator. This integrated tool was validated against dynamic data from actual plant events and thus greatly enhanced the ability and precision of the dynamic simulations. In the study, plant data collected actual transient events and was used to validate the rigorous dynamic model, as well as the integrated simulation approach. During the validation process, the simulation results also enabled in-depth analysis of the actual plant events. Several critical operating scenarios were studied to help solve compressor Compressor energy flow surge, reversals can cause internals, which inside damage is the to motor. High pressure vapor from the discharge of the compressor train is chilled provided against by a vaporizing cascading propane Propane characterized as high-speed and highcompressor, compressor

reduced

Refrigeration (PR) system. The chilled MR is then fed to a highpressure separator where heavy MR and light MR are separated before entering the Main Cryogenic Heat Exchanger (MCHE). In the MCHE, the MR is further cooled to cryogenic conditions and used to liquefy the natural gas feed. In the MR compression system, the compressors are subjected to high flow – high-pressure ratio operating conditions. In particular, the Low-Pressure (LP) axial stage is operated with the highest compression ratio and is most vulnerable to surge. Among various operating scenarios, emergency shutdown, or compressor trip, represent the high-risk cases which can expose the compressors to surge conditions. Other major process upsets, such as a trip of the MCHE or propane system, can also cause unwanted shutdowns of the MR system without adequate control strategies, and impose potential risk of surge on the compressors. The objective of the dynamic study was to analyze the risk of surge under these scenarios and improve the anti-

compressor life time, and loss of profit due to downtime and costly repairs. Emergency shutdown, loss of power source and operation at reduced throughput are some of the factors that can lead to the onset of surge.

Surge factors
Although the compressors and anti-surge systems are designed for a range of operational conditions, actual feed variations, changing operating conditions and production demand, and other operational requirements may require the compressors to be operated under less than ideal conditions and thus can lead to the occurrence of undesirable incidents. The focus of the current study is a Mixed Refrigeration (MR) compressor system in the Woodside LNG facility. Figure 1 shows a simplified schematic of the MR circuit in the LNG liquefaction process. The MR compressor train consists of an axial stage and two centrifugal stages. The drive power is supplied by a gas turbine and a variable speed helper

Novel solution
This article describes a novel solution applied in a recent study to address these challenges. A dynamic study was conducted by

operational issues at the plant and improve control system logic to provide adequate protection for the compressors under extreme operating conditions or upsets

LNG journal

October 2007 •

27

ENGINEERING FORUM

such as the axial stage Inlet Guide Vane (IGV) closing speed, control and actuator delay, valve stroke time, and other system dynamics were based on available data from the plant. The dynamic model was further validated dynamically against plant data collected during actual transient events. The dynamic validation of the simulation results will be described in more detail. Using the rigorous dynamic model, compressor operating conditions were generated according to plant operational sequences. Integrated Simulation Approach: To ensure accuracy of the study results, a CCC-supplied control emulator was used to surge control strategies to mitigate the risk. Two critical cases identified by plant operation were studied. The first case was a propane compressor trip scenario, which in several occasions had caused the MR compressor to trip subsequently. The second case was the emergency shutdown operation, where surge events had been observed during compressor coast down. primary anti-surge control responses that are triggered depending on the location of the compressor operating point are: 1) Proportional Integral (PI) control response: The anti-surge PI control response is the normal first level response serving to increase the recycle rate when the operating point is on the left side of the Surge Control Line (SCL), and will reduce the recycle rate when the operating point is on the right side of the SCL. for the MR 2) Recycle Trip response: The antisurge Recycle Trip response provides a 2nd level correction to the normal PI response by step increasing the recycle rate when the operating point is on the left side of the Recycle Trip Line. This is an open loop response that is analog inputs to the implemented to prevent surge if the normal PI response is not sufficient. An illustration of the compressor performance curve and major control lines calculated by the CCC controller is provided in Figure 3. The illustrated compressor operating point is in the stable operating region. In case of an emergency shutdown, a trip signal from the Central Control System will directly initiate the opening of all anti-surge valves. In addition to the anti-surge valves, the LP axial stage is also protected by a hot gas bypass (HGBP) recycle loop. The axial stage inter-stage bleed valve (IBV) will also be opened to provide sufficient flow to the compressor suction. Dynamic Simulation Model: A rigorous dynamic simulation model was developed using Aspen Custom Modeler (ACM), an advanced equation based software licensed by AspenTech. The simulation model covered the entire MR circuit as shown in Figure 1, with input data based on as-built equipment and piping details. The compressors were modeled on tested compressor curves provided by the compressor vendor. The accuracy of the base model was first validated in steady state against plant operating conditions. Dynamic data emulate the actual controllers. files Together with configuration

downloaded from the plant CCC system, the emulator is able to replicate the exact control functions of the field-installed controllers, and thus allow in-depth analysis of the actual compressor and control system in the field. Figure 4 shows a schematic view of the software-hardware assembly used in the dynamic study. The system includes the CCC emulator hardware and a single PC on which the and dynamic emulator simulation operating software

Control system
The control system compressors is supplied by Compressor Control Corp. (CCC). The CCC control system consists of three anti-surge controllers, one for each stage, as shown in Figure 2. The locations for measuring devices that supply controllers are also indicated in the Figure. Based on the analog inputs, each controller determines the position of the compressor operating point and provides one output to the recycle valve. The anti-surge controllers use a combination of closed- and open-loop control responses to provide anti-surge protection for the compressors [3]. The

software were operated. Data link between the simulation model and emulator was accomplished through OPC server interfaces. The values of process variables calculated by the dynamic model, such as flows (as pressure differentials), pressures, temperatures, valve positions and speed, are supplied to the CCC emulator as inputs.

28

LNG journal

• The World’s Leading LNG journal

ENGINEERING FORUM

compressor

conditions

and

control

improve the control strategies to ensure continuous operation of the MR system after a trip of the propane compressor system. The plant operation sequences in case of a propane trip are similar to the MCHE trip described earlier. However, in the event of a loss of the PR system, the reduction of MR flow to the MR compressor train can occur in a faster manner and thus would require more prompt anti-surge control actions.

responses from the emulator. During the process, the dynamic simulation model and CCC emulator were validated separately and combined, and deemed to be highly accurate and reliable for the purpose of the study.

Control issues
Close examination of the MCHE trip event as described in the previous section suggested that, without other preventive measures, the disturbance caused by an MCHE trip could develop into a situation The emulator-calculated controller outputs are sent back to the dynamic model to control the position of the recycle valves. With control parameters downloaded from the plant CCC system, the emulator would function the same way as the field control system. The emulator would also allow the tuning of the parameters so that new control strategies could be evaluated and new tuning parameters could be tested. A Visual Basic (VB) based script was developed to externally control the execution of the dynamic model, which subsequently controls the execution steps of the CCC emulator. Complete synchronization between the dynamic model and emulator was validated before the integrated tool was put in use. closing the MR flow control valves Anti-surge control valves open when compressor approaches SCL With the opening of the anti-surge valves, the MR compressors were expected to remain on line in stable recycle operation. However, in the actual event, the MR compressors became unstable after the MCHE trip, and eventually resulted in a shutdown of the compressor train. Plant data showing the closing of the axial stage IGV, initiated by the MCHE trip signal, and the compressor speed are provided in Figure 5. Speed decay started at approximately 49 seconds into the event, indicating that a train trip was initiated. The dynamic model was first initialized to plant conditions prior to the event. Dynamic tasks were activated based on plant control sequences. To accurately emulate the real plant control system, the simulation time step was set to be the same as the scan frequency of CCC controllers in the actual plant. Figure 6 shows some of the comparisons of dynamic simulation results with plant data. Two of the plots reproduced here are a) Compressor discharge flow (shown as pressure differential) and b) Suction and discharge pressures. Overall, the simulation was able accurately to reproduce the plant event, both in terms of the simulated that was beyond the controllability of the CCC system. Another finding was that the first level PI control response apparently was not tuned to serve such fast disturbances. The anti-surge responses mainly relied on the second level response by stepincreasing the output. However, a sudden increase of flow to one stage could itself magnify the instability of the system as it could deprive flow for the other stages. The CCC system is designed to handle a number of limitations over the compressor operating window. Given the complexity of the multistaged compressor system, plus the extreme operating conditions imposed by the plant operation, the CCC system alone appeared to be ineffective in handling certain scenarios. Adding more safety margin to the surge control line was not considered a practical solution affect as the it would significantly compressor

Recycle valves
The results indicate that the disturbance caused by a propane trip could be more severe as compared to an MCHE trip. Although a series of step-increase control responses unstable were triggered, the showed MR high compressors moving rapidly into the region, and possibility of surge. The simulation also took into account the actuator performance in the field. The recycle valves are shown to have a onesecond delay and a 3 percent dead band before starting to open. To prevent the foreseen surge event from occurring, simulations were performed to combine a feed forward strategy with the CCC control algorithm. The feed forward logic was applied to ramp open the recycle valves at the initiation of a propane trip. To avoid upsetting the compressors, the appropriate ramp target and ramp rate of the recycle valves were tuned using the simulation. The pre-determined recycle valve opening and ramp rate were insufficient to keep the MR compressors in the stable region. The second level control responses from the CCC system were triggered before the end of the feed forward ramp period. The compressor system was unstable and showed the tendency of surge. With the final settings of the feed forward parameters, the feed forward logic was activated by the PR trip signal,

Simulation results
The integrated simulation tool was validated against plant data collected from a previous MCHE trip event, which resulted in a trip of the MR compressor train and surge of the compressors. Dynamic data collected from the actual event was used to validate the simulation results. According to the plant control logic, an MCHE trip will initiate the following operation sequences. Close the axial compressor inlet guide vane (IGV) Stop heavy and light MR flows by

operation flexibility. Therefore, other control improvements became necessary to address these operating scenarios. A trip of the PR compressors was found to be more critical and could drive the MR compressors into the unstable region even faster. The controllability of the MR system in the event of a PR trip is studied in the first case study below. The objective of this case study was to

LNG journal

October 2007 •

29

ENGINEERING FORUM

and the recycle valves were ramped open to 50 percent and 30 percent for the low, medium and high-pressure stage

Table 1: Dynamic Parameters for Compressor Trip-Coast Down
Study Range Delay Opening of anti-surge valves Opening of hot gas bypass valve Closing of axial compressor IGV closing Opening of inter-stage bleed valve 0 – 2 sec 2 – 10 sec Trip of gas turbine driver 0 – 3 sec 5 - 10 sec 2 – 3 sec 0 – 2 sec 1 – 3 sec 0 – 2 sec 1 – 3 sec 0 sec 7 – 35 sec Desired Operating Range As fast as possible As fast as possible No delay, 25 – 35 sec Stroke time based on study

Acknowledgment
The authors wish to express sincere thanks to Nikhil Dukle and Wayne Jacobson of Compressor Control Corp. for technical support and review of control strategies, Martyn Blanchard of AspenTech for valuable comments during the course of the study, and Ming Yan of KBR for technical assistance with the assembly of software-hardware used in the study.

respectively over 10 seconds. With these settings, the recycle flow rates appeared to be sufficient to maintain the MR compressor operating points in the stable region. The adequacy of the feed forward setting was verified for the full compressor operating range. Once the system is stabilized, the imposed feed forward logic can be released and the CCC controller will take over the control. Based on the study results, it was concluded that a feed forward strategy combined with the CCC control system would provide a predictable and reliable way to achieve continuous operation of the MR compressors. The use of the controls emulator in the study allowed fine-tuning of the feed forward settings realistic to the actual plant.

The possible operating range for each dynamic variable and study results are summarized in Table 1. The simulation results revealed two important variables that might have contributed to the surge problem. One was the LP stage IGV closing speed. Field data from actual events indicated that the IGV closing time was in the range of 7 to 10 seconds. The other important variable is the opening time of the LP stage bleed valve. Analysis of plant data indicated that the stroke time of this valve could possibly be as fast as 2 seconds. Conceptually, fast responses of the IGV and inter-stage bleed valve might be considered desirable as they help unload the compressor quickly. However, results from the dynamic simulations indicated that the fast closing of the IGV as observed in the plant could actually drive the compressor toward surge. Based on the simulation results, it is suggested that the IGV stroke time be modified to close the IGV at a moderate rate over 25 to 35 seconds. As for the inter-stage bleed valve, an optimum window for the opening time appeared to exist to avoid surge in either section of the axial stage. A typical logic is to open the bleed valve when speed is reduced down to 95 percent, which in this case is equivalent to a delay of approximately 1 second. Overall, a valve stroke time between 5 to 10 seconds appeared to be the optimum range. The anti-surge recycle valves and the hot gas bypass valve are required to open as quickly as possible to reduce the risk of surge during coast down. Typical requirements for these valves are less than two seconds [4]. In the actual plant, the response time of these valves had limitations for significant improvements. To compensate for any delay in the responsiveness of these valves, the effect of introducing a delay on turbine trip was studied. Based on the simulation results, it was concluded that delaying the turbine trip by 2 to 3 seconds would

provide an additional safety factor to avoid the surge of the compressors during coast down. Jihong Wu is a senior process engineer with KBR in Houston, Texas, specialized in dynamic simulation. Her experience also includes process design and optimization of LNG, olefins and other large-scale processing facilities. She graduated from Tokyo University with a PhD in Chemical Engineering.

Implementation
The feed forward strategy studied in Case Study 1 was incorporated into the plant control logic and the benefits have been observed in several plant events since the implementation. To solve the surge on coast down, which had been previously considered unsolvable by the vendor, field modifications were made based on the findings from Case Study 2. Plant operation compressor has coast demonstrated downs since safe the

Train trip
The objective of this case study was to investigate the existing system and improve anti-surge strategies to protect the compressors from surge during coast down. The plant history has also identified that emergency shutdown, or compressor trip, represents a high-risk case The follows: Close the the axial stage IGV recycle immediately Trip-open anti-surge valves and hot gas bypass valve Open the axial stage bleed valve Stop heavy and light MR flow by closing the MR flow control valves In this case study, the dynamic behavior and effectiveness of anti-surge elements were thoroughly reviewed. In addition to reviewing the size and flow characteristics of the recycle valves, the impact of various dynamic variables such as control signal and actuator delays, valve stroke time, IGV closing speed and the timing for tripping the turbine driver were investigated. In this scenario, the CCC calculation algorithm is bypassed upon the initiation of a train trip. Therefore, the dynamic study was performed without using the CCC control emulator. The LP axial stage was modeled as two stages with flow take-off from between stages to the interstage bleed valve. for possible surge of the of compressors, in particular, the LP stage. plant control sequence operation in the event of a train trip is as

implementation and proved the accuracy of the dynamic simulation results.

Jeffrey Feng is a process leader with KBR, in Houston, Texas. He has been with KBR since 1995 after graduating from Texas A&M University with a PhD in Chemical Engineering. He is responsible for the technical execution of dynamic simulation and other transient analysis in LNG, refining, olefins, ammonia and offshore in domestic and international projects.

Conclusions
It was only through the use of rigorous dynamic simulations and that the of permutations combinations

various system variables could be safely tested to derive the solution. The studies proved the accuracy and effectiveness of such modeling and provided useful results to diagnose and improve field operation. The combination of the dynamic simulation modeling with control vendor supplied hardware significantly enhanced the precision, capability and credibility to develop realistic and reliable solutions for the actual plant system.

Surajit Dasgupta is the Manager of Chemical Engineering Technology and Advanced Process Automation at KBR, in Houston, Texas. He supervises all projects in the area of dynamic simulation, advanced process control, operator-training simulators and realtime optimization. He graduated from Columbia University, NY, in 1977 with a Doctoral degree in Engineering and Science.

Ian Keith is the Chief Process Control Engineer at Karratha Gas plant in Australia. He has been with Woodside and Shell since 1998 and was the Senior Process Engineer for Karratha LNG plant prior to his current position.

References
[1] Omori H., Konishi, H., Ray, S., de la Vega, F. and Durr, C., “A New Tool-Efficient and Accurate for LNG Plant Design and Debottlenecking”, 13th International Conference & Exibition on Liquefied Natural Gas, May, 2001. [2] Valappll J., Mehrotra, V., Messersmith D and Bruner, P., “Virtual Simulation of LNG Plant”, LNG Journal, January/February, 2004. [3] Compressor Control Corporation, Series 5 Antisurge Control Application for Centrifugal and Axial Compressors, Publication UM5411, October 2005. [4] Wilson, J. and Sheldon, A., “Matching Antisurge Control Valve Performance with Integrated Turbomachinery Control Systems”, Hydrocarbon Processing, August 2006

30

LNG journal

• The World’s Leading LNG journal

Sign up to vote on this title
UsefulNot useful