Advanced Process Control in the Plant Engineering and Construction Pha
Vassilis Sakizlis∗, Andy Coward♠, Kumar Vakamudi♣, Ivan Mermans♠.
1. Introduction
The profit in process industries is a function of effective plant operation. This energy and economic bad incentive leads to the development of advanced operation and control strategies (See Qin and Badgwell, 2003; Christofides et al., 2008).吊球
The success of advanced control in the oil and gas industry, in particular, over the last 30 years is widely acknowledged. Successful applications have been reported in olefin plants (Shindo et al., 2004), refineries (Moro and Odloak, 1995; Haloff et al., 2007) and gas plants (Hotblack, 2004; Honeywell 2005). Payback times of a few months have been achieved, in addition to improved economics by 5-10%. The success of the projects is attributed to incread profit margin and reduction in the energy consumption. Currently, the main rearch and development trends in that field focus on centralized co-ordinated control for plant – wide solutions (Cheng et al.,2007), non-linear models (Kumar and Daoutidis, 2002) and multivariable optimizing solutions that can be incorporated in a PLC or DCS hardware (Dua et al., 2008).
Predominately, however, advanced control is developed, installed and commissioned after the plant start-up. Dynamic simulation may be ud to expedite the step testing and plant model development, however, the t-up of the system is still held until at least after the plant commissioning. This conventional strategy (Figure 1) inadvertently introduces a delay in the advanced control implementation and conquently postpones the capitalization of the relevant benefits.
党员民主测评表
∗ Bechtel Ltd. 245 Hammersmith Road, W6 8DP, London, UK.
♠ Honeywell Process Solutions, Lovelace Road, Bracknell, Berkshire, RG12 8WD.
♣ Bechtel Ltd, 3000 Post Oak Boulevard, 77056, Houston, Texas, USA.
张杰新歌
Figure 1. Conventional Approach to Multivariable Control
In the novel work prented here we have minimized this delay by providing advanced process control at the early stages of the process design and engineering. The work concerns the design, testing and installation of advanced control (also termed as multivariable control – MVC) as part of the engineering and construction of an approximately 1300 MMSCFD gas plant in the Middle East.
The MVC scope of work as part of the EPC (Engineering Procurement and Construction) pha included:
Multivariable Predictive Control hardware and software supply, installation and integration.
药膳乌鸡汤Development of a dynamic simulation model for testing the MVC performance.
Supply and configuration of the DCS-OPC interface communication.
This interface compris software, hardware and the necessary graphics
and logics.
The contractor (Eastern Bechtel) managed, supervid and provided information to the vendor (Honeywell) who developed and delivered the MVC.
The next ction explains the prime terms ud in this article. The following paragraph
describes briefly the process while focusing on the NGL recovery system. The
following ctions describe the steps that were followed starting from the dynamic model development for the NGL recovery system and finishing with the MVC training for operators and engineers.
2. Definitions
−Dynamic simulation model: Non-linear model of the process plant bad on first principles and built on a commercial software, e.g.
HYSYS, UniSim, gPROMS.
−Linear dynamic model. Model derived from plant excitation data. The form of this model can be finite time impul respons, transfer家庭剁椒鱼头的做法
functions, state space or AutoRegressive eXogeneous model (ARX).
This model is in general ud to derive a multivariable control solution
for the plant.
−Multivariable Control (MVC) refers to control algorithms that simultaneously compute the optimal value of a t of manipulated
variables bad on a lected t of current plant measurements. The
computation is the on-line solution of an optimization problem
consisting of :
an economic or tracking objective function
the linear dynamic model of the system and
performance and operational constraints
For more information on MVC that is also called APC (Advanced
Process Control) or MPC (Model Predictive Control) refer to Quin and
Badgwell (2003). The Honeywell proprietary MVC controller is the
Profit Controller algorithm and software.
3. Process Description
This gas plant facility includes process units for gas and condensate paration, condensate stabilization, gas treating, gas dehydration, NGL recovery and residue gas injection. The simplified plant flowsheet is shown in Figure 2. The main products from the plant are outlined on Table 1.
Table 1. Products of the gas plant.
No Product Plant Unit
1 Condensates C5+ Condensate Stabilizer bottoms
and Liquid Separation bottoms
2 Natural Gas Liquids
(NGL), mostly C3, C2 NGL recovery unit and Liquid Separation bottoms
3 Sweet natural gas C1,
C2
NGL recovery overhead
Figure 2. Gas Plant flowsheet.
The scope of MVC concerned the three main paration units that are directly linked with the plant production, i.e. the condensate stabilizer and the feed liquid paration unit that remove the heavy components from the gas stream;; and the NGL recovery system that includes a demethanizer distillation column.
The unit which attributes most to the plant production and the one which benefits the most from applying the Multivariable Control (MVC) optimization is the NGL recovery system. Figure 3 is a simplified process flow diagram of the NGL recovery system that omits the details that are not relevant to MVC. The NGL Recovery Unit utilizes UOP/Ortloff’s licend Supplemental Rectification Process (SRP) (US patent number 7,191,617) technology to efficiently recover the ethane from the dehydrated gas while maintaining 99+% recovery of the propane and heavier components for the wide range of ethane recoveries. The resulting residue gas is routed to the residue gas manifold for reinjection after being re-compresd by two compressors in ries that control the pressure in the unit.
The NGL recovery process includes a multi-stage propane refrigeration circuit to
provide the additional cooling duty required to meet the product specifications.
Figure 3. NGL recovery flowsheet and reprentation of MVC manipulated variables.
4. MVC Objectives and Control Scheme
Here, we describe the formulation of the control structure for the NGL unit. The same approach was ud in the rest of the relevant units.
The prime MVC objectives were to
•Minimi the C3+ composition in the overhead product and maximize C3
recovery in the bottoms product.
•Minimi the C2+ concentration in in the NGL stream during C2
recovery and the rever during C2 rejection.
•Maximi plant production rate.
•Minimi plant energy usage.
At the same time operational constraints should not be exceeded. For example, the C1/C2 ratio in the bottoms product should be maintained below a designated t-point, while the pressure in the ov
erhead gas stream should not deviate from its given value. To satisfy the aforementioned constraints and objectives the contractor (Bechtel) compiled the t of manipulated and control variables that constitute the MVC医学生实习小结
structure. The main lection criteria were:
−The manipulated variables (MVs) should have a large impact on the process objectives and constraints. Additionally, the effect of the MVs
on the process objectives should ideally be immediate without large
delays. For example, reducing the temperature of the feed to the
demethanizer at constant pressure, leads to a reduction in the column
temperature, hence increasing the liquid ethane recovery.
−The control variables had to be directly linked to the objectives and constraints of the process. For example the pressure at the column
overhead is a constraint and therefore, was included in the structure.
−The number MVs and CVs should be kept at a relatively low level to avoid complexities during tuning and commissioning and to prevent
negative effects on the NGL or other parts of the plant. For example the
pressure at the NGL inlet was not ud in the structure as it may affect
adverly the downstream units.
小熊陪你玩
The awareness here of the contractor on the plant operation and design was paramount to the determination of the control structure.
5. Configuration and Testing of MVC Controller
Becau the “real” process for this project did not exist at the time of the MVC project, the testing of the developed MVC applications could be handled in one of two ways;
•Testing against a “perfect” inverted linear dynamic model with the Profit Design Studio,
•Testing of the Profit Controller against the control loops and the process model within a dynamic simulation model.
To demonstrate how the Profit Controller would act with unmeasured and process disturbances, the latter option was chon.
5.1. Dynamic Model
清拌黄瓜A dynamic model of the plant was under construction by the contractor (Bechtel) to conduct design studies and to rve as a platform for the operator training plant simulator (OTS). The development of the model was a joint effort by the Bechtel Advanced Simulation Group and the Bechtel Process discipline of the gas plant project. This model is a detailed reprentation of the plant steady state and dynamic behaviour. For instance, it accounted for hydraulics, compressor performance curves and the modelling of the dominant stream components. The NGL portion was extracted from the plant model and after a t of modifications it was handed over to the vendor (Honeywell). The main modifications were:
