Challenges and problems with advanced

control and optimization technologies

Campos, M.*, Teixeira, H.*, Liporace, F.** and Gomes, M.**

* PETROBRAS / CENPES / Engenharia Básica Abastecimento e Gás&Energia/ Automação,

Equipamentos Dinâmicos e Confiabilidade, Av. Horácio Macedo, 950 - Cidade Universitária,
Ilha do Fundão, Rio de Janeiro, 21949-915, Brazil
(Tel:55-21-3865-6347; e-mail: mariocampos@petrobras.com.br).
** PETROBRAS / CENPES / P&D de Gás, Energia e Desenvolvimento Sustentável / Gás Natural /
Célula de Otimização e Eficiência Energética, Av. Horácio Macedo, 950 - Cidade Universitária, Ilha do Fundão,
Rio de Janeiro, 21949-915, Brazil.

Abstract: Oil & Gas companies continuously try to create and increase business value of their
installations (platforms, refineries, etc). Particularly the increasing energy consumption on a worldwide
basis and, as a result, the substantial increase in prices volatility is a major drive for better advanced
control and optimization technologies. Advanced control and optimization system can play an important
role to improve the profitability and stability of industrial plants. This paper discusses the problems and
challenges of advanced control and optimization in petroleum industries nowadays. It emphasizes the
importance of control performance assessment technology to maintain a good regulatory control and the
difficulties in using these technologies. It also shows the importance of malfunction detection and
diagnosis advisory system for critical equipment in order to increase the operational reliability. Model
predictive control (MPC) has become a standard multivariable control solution in the continuous process
industries, but there are still many open issues related to accelerate a new implementation and maintain
the controller with a good performance along the years. Real time optimization tools also impose new
challenges for Oil & Gas industries application, which are discussed in this paper.
Keywords: performance assessment, regulatory control, advanced control system, real time optimization

the regulatory and advanced control, and the challenge

1. INTRODUCTION associated with the real time optimizers. In spite of the
several tools in the market that deal with industrial control
The advanced control and optimization systems in oil & gas and optimization solutions, PETROBRAS has decided to
and petrochemical plants are an industrial reality (Qin and invest on the development of its own tools and solutions in
Badgwell, 2003). These advanced systems provide many many situations, usually in association with some Brazilian
advantages for the process units, as improved stability and universities. The goal of this paper is to show some
safety, respect to constraints and higher profitability.
challenges faced, solutions and results obtained in
PETROBRAS has been investing in the development of these PETROBRAS facilities.
systems for several years. Advanced control system is already
a consolidated technology in its refineries with many model
2. REGULATORY CONTROL LEVEL
predictive controllers implemented (Zanin and Moro, 2004).
However, the application of real time optimization (RTO) is Process control aims to maintain certain variables within their
recent, although this technology can bring great economical desirable operational limits and could be visualized as a
earnings, besides to increase the energy efficiency and pyramid. In the base of this pyramid, the first level is the
minimization of emissions. regulatory control, that uses PID controllers (Campos and
Teixeira, 2006; Ogata, 1982) and is configured in the digital
To install and maintain these advanced systems with good systems (DCS - Distributed control system or PLC -
performance is a great challenge. Its performance is Programmable logical controllers). In a second level, we have
influenced by instrumentation problems, bad tuning of the the advanced control systems that use for instance Model
regulatory and advanced control, unreliable process dynamic Predictive Control (MPC). This algorithm considers the
models (Ender, 1993; Kern, 2007), unmeasured disturbances, interaction between control loops, and includes an
etc. optimization layer of the industrial plant. These algorithms
are usually implemented in a process computer that
This article will discuss the problems and challenges of communicates with DCS or PLC systems by the use of OPC
advanced control and optimization in petroleum industries protocol (OPC, 2008). The outputs of this advanced control
nowadays. It discusses some tools for diagnosis and tuning of are usually the set points of the PID controllers. The
architecture is conceived in such a way that if there is a the software "BR-Tuning" (Schmidt et al., 2008; Arruda and
failure in the advanced control level, the plant operation Barros, 2003), which is comprised by a group of techniques
continues with the last PID set points in the DCS. regarding open and close loop identification and the
proposition of new tuning parameters. It communicates
An advanced control system won't reach the expected directly with the process automation system (DCS or PLC)
benefits if is turned off constantly for the operators. using the OPC protocol.
Therefore, the instruments, valves and the regulatory control
loops (PIDs) should operate appropriately. Hence, the As it was said previously, the challenge is to develop an
performance of the regulatory control is fundamental for the "intelligent" layer that helps to make a diagnosis based on
success of the advanced control system. An industrial plant several indexes or indicators. The integration between
usually has hundreds of control loops, and less and less different tools is also an important concern. The use of the
engineers to maintain the system. Therefore, the industries OPC standard for the exchange of information could be an
need tools to perform automatic analysis and diagnoses of the option. So, each tool could make available their indicators to
problems associated with the regulatory control. For example, others tools through OPC. This way, the engineers' work
these tools should be able to detect failures with the would be facilitated, avoiding losses of time and money.
instrumentation (miscalibration, badly sizing, sensor noisy,
out of scale, measurement resolution, etc.), non linear
behavior in the process due to changes in the operational
point, bad PID tuning (oscillation, stability, etc.) and control
strategy problems (coupling between control loops, degrees
of freedom, etc.).

There are several tools in the market that help engineers to

maintain the regulatory control, but most of them require a
well-trained engineers to interpret, analyze and define the
correct actions, for instance: to change a control valve, tune
PID controllers or to implement a new control strategy
(decoupling, feedforward), etc. These engineers should also
know very well the process in order to evaluate the better
actions to be taken.

The great challenge for these tools will be to incorporate

more "intelligence" to help engineers in the definition of the Fig. 1. BR-Tuning interface.
better actions. For instance, in certain case, only PID tuning The challenges in relation to controllers' tuning are associated
could reach 80% of improvement in process variability mainly with the identification of the models, the
reduction, and in some case, the process performance would determination of the process non-linearities, interaction
improve only 10%. A lot of times in industries the engineer between control loops, as well as defining the desired
spends time and money with an action that won't bring great performance for each control loop.
results. So, it is clear the importance of a tool that could
perform the automatic diagnosis and assessment of the There are some processes where the disturbances’ pattern can
regulatory control (Farenzena et al., 2006). The most change with the time, as in some off-shore petroleum
important features of this tool should be to have automatic platform. The slug flow can change its intensity for example
ways to prioritize the actions for each process that might due to changes in the gas-lift. So, we don't have a PID tuning
result in a better performance, and also to provide a parameters that are good for all these different situations. In
standardized metric to compare different actions in different this case, it was developed an "intelligent" system that
processes, even in different scales such as economical, supervises the process plant and changes the PID tuning
environmental or safety (Harris, 1989; Kempf, 2003; automatically when necessary. This control strategy is
Farenzena and Trierweiler, 2008). These features are a great equivalent a "gain-scheduling" where the control
development challenge for these tools. performance (deviation between the process variable and the
setpoint) is evaluated during a time, and the system decides
Despite the several tools in the market, PETROBRAS and what is the best tuning for that moment. All the possible
Federal University of Rio Grande do Sul (UFRGS) have values for the PID tuning are chosen off-line. This system
developed their own tool, the software called “BR-PerfX”. Its was installed in several PETROBRAS' platforms. The figure
main purpose is to compute some universal key performance 2 shows the system changing the PID tuning parameters and
indicators that reduce the subjectivity in the analysis and help the level performance. This project used a tool called MPA,
engineers in their assessments and decisions about problems which was developed by Catholic University of Rio de
affecting the regulatory control. Janeiro (PUC-RJ) to PETROBRAS.
In order to face the PID tuning problem, PETROBRAS and Another challenge is the development of non-linear
Federal University of Campina Grande (UFCG) developed controllers for some special cases, for example to pH control
in certain plants, although PID will continue to be the significant gap between the recent MPC technologies
algorithm more used in this regulatory layer control for development in the academy and those effectively used on
several years. industrial plants. Most industrial MPC applications are based
on the most traditional approaches: linear algorithms based
Researches and developments for the regulatory control level on step-response models obtained through traditional step
are still necessary, and they can bring great economical tests.
earnings. For example, an application of these tools
(evaluation, tuning and changes in control strategy) allows an MPC maintenance
increased of about 9% in the production of LGN (Liquefied
Natural Gas) in a natural gas plant (Campos et al., 2007). MPC performance decay throughout time is a well-known
and widely reported fact (figure 4). If no maintenance work is
done, the operators end up turning them off. There are many
causes for this behaviour:
ƒ Changes in the units operational objectives;
ƒ Equipments efficiency losses (fouling);
ƒ Changes in the feed quality;
ƒ Problems in instruments and in the inferences;
ƒ Lacks of qualified personnel for the controller's
maintenance.

Therefore, the first great challenge associated with MPC

control is to have reliable tools to keep performance and
Fig. 2. Performance of this control strategy in production diagnose problems.
platform (1 day).

3. ADVANCED CONTROL SYSTEM

The multivariable predictive controllers (MPCs) are powerful
tools for the process optimization and are available in many
industrial plants. This system can increase feed and preferred
product rates, reduce energy consumption and waste material.
These benefits are more visible in complex processes where
challenging dynamic responses (significant time delays, non-
minimum phase responses, control loop interaction, etc.) due
to disturbances (feed flow and composition, energy
integration, usefulness, etc.) that must be dealt with while
taking into account process constraints and trying to pursue
the best economic performance. As an example of the Fig. 4. Advanced Control Performance during the time.
benefits achieved, figure 3 shows an increase of about 16% in
Therefore, industry needs better tools to help maintenance
the LPG yield due to the implementation of an Advanced
personnel to answer the following questions:
Process Control (APC) system in a natural gas plant.
ƒ Is advanced control system accomplishing their
objectives?
ƒ What is its performance?
ƒ Is the process optimized?
ƒ What are the benefits?
ƒ How is the level of disturbances?
ƒ What is operational factor of the controller?
ƒ How are the operators adjusting the limits of the
manipulated variables?
ƒ Are manipulated variables very limited?
ƒ What is the variability of the main controlled
variable?
ƒ Is the process operating close to the constraints?

It is necessary a tool not only to answer these questions, but

the system point out the causes of the bad performance: bad
Fig. 3. LPG yield increase in a natural gas plant due to models, bad controller tuning, inference problems, non-
MPC. linearities, frequent changes in the operation point, new
However, even if MPC systems are nowadays seen as a constraints not considered in the design?
commodity, there is still much to be done, due to the
 ƒ Tools for the development of inferences:
Nonlinear models, Identification and Model mismatch o Use of rigorous dynamic simulators, or statistical
methods for better inferences using less laboratory
Many different and even sophisticated approaches have been analysis data.
proposed in order to allow MPC algorithms to cope with ƒ Dynamic models identification:
process nonlinearity. Bequette (2007) presents a recent o Automation of the identification tests,
review on the subject. However, despite all this effort, minimizing problems and loss of data;
industrial Nonlinear MPC (NMPC) applications are relatively o Efficient tools for closed loop identification;
few, and most of these are based on the simplest approaches. o Characterization and identification of the non
linearities of the process.
One possible reason for that might be simply that the ƒ Better tools for tuning the predictive controller:
nonlinear behaviour is not known, and any lack of o How to define the priorities in the several
performance is seen as a typical model mismatch. operating points of the controller and change
automatically the tuning parameters. This activity
Another possibility might be that the nonlinear behaviour is is still done by trial and error in many industrial
known, but can not be easily determined with traditional cases.
plant tests. One way to overcome these problems might be
the use of rigorous dynamic simulators, to improve the New advanced controllers that contemplate these aspects will
understanding of the process behaviour. Information obtained help the users to implement and maintain these industrial
with dynamic simulation could be combined to the existing systems.
linear model in order to provide a reliable nonlinear one.
Dynamic simulation might be useful also to find out the best 4. REAL TIME OPTIMIZATION
way to characterize the observed nonlinearity. Once more,
although there is availability of dynamic simulators, there is Real Time Optimization (RTO) technology is a powerful tool
not much use of them in industrial applications. for the continuous search of the most profitable way to run
petroleum and petrochemical process units. Cutler and Perry
Process identification of complex processes is still a hard (1983) state that despite being a hard and complex task, its
task, where a significant part of the effort on MPC potential benefits are relevant and might provide profit
implementation is spent. increases around 6 to 10% when allied to Advanced Process
Control (APC).
In order to address this problem, some commercial tools have
been conceived in this decade for closed-loop identification. The task of an RTO application is to make the best of an
These tools are based on efficient ways to perform step tests existing process unit, adjusting its process variables for every
allied to modelling strategies for minimization of the model new change of external conditions, like operational variables,
order. While this approach has proved to be useful and feed compositions and process constraints. The RTO benefits
promising, it is still a hard task to apply these techniques to are usually associated with the maximization of products and
complex processes, especially when dealing with noisy data. minimization of the specific energy consumption and other
It seems to be a lot of space for development in this area. resources, depending on the following factors:
ƒ Market availability
Another interesting way to reduce implementation time can ƒ Products prices and feed costs
be the use of algorithms for automation of the plant test. ƒ Safety and environmental constraints
ƒ Product specifications
Tuning
The central figure of an optimization application is the
MPC tuning is another interesting issue, where new mathematical model. It is expected to represent the process
technologies might help to reduce implementation time and behaviour on a wide range of operating conditions with good
also on the maintenance task. accuracy. It should not only guarantee that the predicted
potential profitability matches that of the real process, but
Some interesting ideas have been proposed (Trierweiller and also that when the optimal solution is implemented the
Farina, 2003) that try to combine desired and achievable process constraints must not be violated. Most RTO systems
performances. However, the controller tuning still consume used nowadays are based on rigorous, steady-state, first-
time and is critical points for controller performance. principles mathematical models.
Normally, all MPC tuning methods consider a square
controlled variables x manipulated variables matrix, but, in The good performance of an RTO system depends on a
fact all controller has a rectangular matrix that means reliable mathematical model and on reliable input data. In
different tuning scenarios depending on which constraints is order to obtain that, many procedures must be executed
active. before the economic optimization problem can be solved:
ƒ Gross Error Detection
Another big challenge is to reduce the application time and ƒ Steady-state Detection
maintenance time. For this, it is believed that the main critical ƒ Data Reconciliation
points are: ƒ Parameter estimation
Once that a reconciled data set and a fitted model have been The process model was built using PETROX, a proprietary
obtained, the process optimization can be performed. The sequential-modular process simulator from PETROBRAS.
optimization problem usually consists of the maximization The simulation comprises 53 components and pseudo-
the operational profit (or minimization of operational costs) components and 64 unit operation modules, including the 7
subject to a set of constraints. On most situations the distillation columns and a recycle stream. All modules are
optimization problem is posed as a non-linear programming built with rigorous, first-principles models.
problem (NLP). Most commercial applications are based on
variations of the SQP (Successive Quadratic Programming) For optimization applications, PETROX was linked to
algorithm. This algorithm is also used to solve the previous NPSOL, an SQP optimization algorithm. Procedures for
Data Reconciliation and Parameter Estimation problems. Steady-state and Gross error detection, Data Reconciliation,
Parameter Estimation and Economic Optimization were
Real Time Optimization at PETROBRAS implemented. The economic optimization problem consisted
of the maximization of the operational profit, constrained by
Since 2004, RTO has been classified by PETROBRAS and limits related to product specifications, safety constraints,
its Strategic Downstream Committee as a “High Sustainable” feed rate and performance parameters. The whole
technology. It means that RTO is seen as a key technology to optimization problem involves 19 decision variables and 21
improve PETROBRAS performance and profit, and therefore constraints.
significant effort and resources will be spent on this subject.
Most of the reported problems of optimization based on
PETROBRAS implementations on RTO covered a wide sequential-modular models were observed in this application:
range of alternatives, focusing both on profitability and on ƒ Low computational efficiency, due to slow recycle
the search of the best way to deliver the technology: loops and the numerical derivatives that imply
ƒ Fluid Catalytic Cracking (FCC) and Crude running the SM model several times. These
Distillation Units (CDU); derivatives are also inaccurate, which slows down
ƒ Proprietary and commercial process models and the optimization process even more.
RTO systems; ƒ Lack of reliability: the SM model is computed many
ƒ Sequential Modular (SM) and Equation Oriented times and must converge always. If a single failure
(EO) approaches (Alkaya et al., 2003). happens during the optimization, all the effort is
lost.
The first RTO initiatives were taken using PETROBRAS' in- In order to minimize these problems, a lot of effort must be
house process simulator for FCC, with a small scope spent on the conception, customization and tuning of the SM
covering only the reactor/regenerator section. The proprietary model. However, that is no guarantee of success. When the
process model used is based on a Sequential Modular (SM) Data Reconciliation and Parameter Estimation problems were
approach. Though many difficulties were found (see next implemented, the same problems were observed.
section), this initiative made possible to test the technology as
well as to help our engineers to take a step further.
CONFIGURA TION
ƒ Rigorous Model
ƒ Base Metamodel
Distillation Unit / SM approach (2004) ƒ SAO Parameters
Problem configuration: Variables,,
ƒ
Limits, Initial estimates
This implementation took place at the Crude Distillation Unit
(CDU) and the two Solvents Units of RECAP refinery
(Gomes et al., 2008). Metamodel Adaptation
Rigorous
Model

Optimisation

Update Trust
Check Termination Criteria Region
NO
OK?
YES

END

Fig. 6 - SAO strategy applied to the metamodel-based

optimisation.

Metamodel approach
In order to overcome some of these shortcomings, a
metamodel approach has been studied. Metamodels or
surrogate models (Gomes et al., 2008) are reduced models
whose parameters are obtained with data that is generated
with rigorous, first principles models. In this work, an
Fig. 5 - Scheme of the CDU and the Solvents Units of optimization procedure was developed, combining
RECAP/PETROBRAS.
metamodels and rigorous models with a Sequential ROMeo (Invensys, Inc.) was selected.. The project scope
approximate optimization (SAO) algorithm. The optimization included the Reactor / Regenerator section, Main Fractionator
problem is solved based on the metamodel that is updated and Gas Recovery Plant. Again the unit was fully energy and
with data obtained from the rigorous model throughout the mass integrated modeled.
optimization procedure. The RECAP optimization problem
was addressed with this approach, with kriging models and
neural nets used as metamodels. Accurate results have been
obtained with considerable reduction of the computational
effort on most of the studied cases.

Distillation Unit / EO (2005 to 2006)

This was the first EO RTO project PETROBRAS
implemented. After an International Bid, where 3 well-known
companies were invited to submit their proposals, AspenPlus
Optimizer (Aspentech, Inc.) was selected. The project scope
included all 3 preheat trains as well as Pre-flash, Naphtha Fig. 8 - ROMeo screenshot - Reactor/Regenerator Section.
Stabilizer, Atmospheric, Vacuum and Pre-vacuum The system is running on closed loop (around 8 runs / day)
distillations towers. The unit was fully modeled with the since June/08 with most of the independent variables active.
RTO software, which allowed for instance the understanding On average, around 60% of the successful runs are being
about the implications that changes on the preheat train, like accepted by Operations and targets are being sent to
feed distribution, have on the Atmospheric tower. Or to study Advanced Control. PETROBRAS has evaluated an average
the best pumparound heat removal distribution along this gain of US$ 0.12 / bbl of FCC feed for this application, by
tower and its effects on the preheat train. In order to do that, comparing the unit performance with and without RTO.
all pumparounds were modeled as external streams from the
tower and not as an internal model within its model (see A few comments on both projects:
Figure 7), as it is common on SM simulators. ƒ Lack of instrumentation on preheat train (FCC) –
implied on simplifications, which has impacts on
The system is running on open loop since 2007. A few Main Fractionator heat balance and, thus, must be
closed-loop tests were performed, but the unit had some evaluated from time to time;
operational problems which were solved on this last Oct/08 ƒ Low feed lab analysis frequency – There is a need
turnaround. PETROBRAS intends to close loop in 2009 after for a better way to estimate feed characterization;
making model tuning adjustments in order to incorporate the ƒ Non-convergence problems - Mainly, due to
new atmospheric trays and other unit improvements. instrumentation faulty and/or out of service heat
Nevertheless, by keeping the system running open loop exchanger or other piece of equipment. Although
(around 9 runs / day), we were able to improve our there is a kind of standard procedure to deal with
knowledge of the system itself, how to overcome non them, it is not possible to automate it. So each
convergence problems (feed reconciliation and optimization) problem must be solved on a case to case, hands-on
and attaining expertise on how to maintain such a real time, basis.
strongly data and instrumentation dependent system as well
as evaluate potential benefits (around 13 000,00 dollars / These facts enforce the need for a fully dedicated RTO
day). engineer for each application, not only to assess its results
and make sure they are being implemented, but to keep the
system running despite of the many daily issues the
application faces.

Modelling approach
PETROBRAS experiences showed that the Equation
Oriented (EO) approach is more suitable for RTO, when
compared to the Sequential-modular process models,
especially when process unities of higher complexity are
addressed.
Fig. 7 - Aspen Plus Optimizer Screenshot - Atmospheric
tower. Challenges associated with RTO

FCC Unit / EO (2007 to 2008) Non-convergence tracking

Following the success on the distillation unit implementation, When the optimization process brakes down due to non-
PETROBRAS moved forward to implement an RTO on convergence, it is sometimes a hard task to find out the origin
another very important unit. Again, after an international bid, of the failure, especially when the cause of the problem is not
related to instrumentation or well-known process problems. industries as well as how PETROBRAS is overcoming them.
Therefore, there is a need for better procedures or even an Our vision is that there is still plenty of space for further nd
expert system that might identify the numerical failures and research and development on the improvement of those
provide high-level analysis to support the user on the best technologies. The best accomplishment of this task will come
actions to take. if Industry and Academy work together.

The improvement of the initialization techniques (Fang et al., REFERENCES

2009) might also be useful to avoid convergence problems,
especially for the data reconciliation problem. Alkaya, D. et al., Generalization of a Tailored Approach for
Process Optimization, Ind. Eng. Chem. Res., Vol. 39, pp.
1731-1742, 2003.
Scaling Arruda, G.H.M. and Barros, P.R., Relay based gain and
phase margins PI controller design, IEEE Transactions
Scaling of variables is a subjective issue. Despite the
on Instrumentation and Measurement Technology, Vol.
available heuristic rules provided by the technology licensors,
52, n. 05, pp. 1548-1553, 2003
the users are sometimes required to define scaling factors or
Bequette, B. W, Non-linear Model Predictive Control: A
limits. However, it is possible that a numerical analysis of the
Personal Retrospective, The Canadian Journal of
system of equations to be solved might provide the best
Chemical Engineering, Vol. 85, pp. 408-415, 2007.
scaling factors.
Campos, M. and Teixeira, H., Controles típicos de
Integrating multiple process unities equipamentos e processos industriais, Ed. Edgard
Blücher, São Paulo, 2006.
In order to take the most of process flexibilities, it might be
Campos, M. et al., Ganhos econômicos devidos à melhoria
important to expand the scope of the optimization problem to
no controle de uma planta de processamento de gás
involve more than just one process unit. However, the
natural, IV Congresso Rio Automação 2007, IBP, Rio de
increase of the problem size and the consequent shortcomings
Janeiro, 2007.
can be a challenge to be faced. In this case, the non-converge
Cutler, C. R. and Perry, R. T., Real-time optimization with
tracking procedures would become a key issue.
multivariable control is required to maximize profits,
Steady-State detection Computers and Chemical Engineering, Vol. 7 (5), pp.
663-667, 1983.
The steady-state detection procedures used nowadays in the Ender, D., Process Control Performance: Not as good as you
commercial solutions require the definition of several think, Control Eng., pp. 180, 1993.
parameters, which is a very subjective issue. This task
Fang, X. et al., Mnemonic Enhancement Optimization
demands from the user not only process experience, but also (MEO) for Real-Time Optimization of Industrial
a long time of observation. It would be useful to have Processes, Ind. Eng. Chem. Res.,Vol. 48, pp. 499-509,
procedures that could drive a straightforward choice, 2009.
especially when dealing with multiple-process optimization Farenzena, M. and Trierweiler, J., Fronteiras e desafios em
applications.
gerenciamento de malhas de controle, In: COBEQ 2008 -
Multi-scale optimization Congresso Brasileiro de Engenharia Química, Recife,
2008.
The integration and information exchange between different Farenzena, M. et al., Using the Inference Model Approach to
optimization levels is an issue that requires more attention. Quantify the Loop Performance and Robustness, SICOP
2006 - International Workshop on Solving Industrial
Multi-level optimization concepts could be applied in order Control and Optimization Problems, Gramado, 2006.
that procedures for model re-fitting or tuning and the Gomes, M.V.C. et al.; Using kriging models for real-time
redefinition of search spaces could be done automatically, process optimisation. Proceedings of the 18th European
while the different optimization problems are being solved. Symposium on Computer Aided Process Engineering,
Dynamic RTO pp. 361-366, 2008.
Harris, Assessment of Control Loop Performance, The Can.
Dynamic Real Time Optimization (DRTO) is an open issue. J. of Chemical Engineering, Vol. 67, pp. 856-861, 1989.
The use of rigorous dynamic models for large-scale Kempf, A., Avaliação de Desempenho de Malhas de
applications might allow the simultaneous solution of process Controle, Dissertação de Mestrado, Departamento Eng.
optimization and control problems. Ideally it would also Química, Universidade Federal do Rio Grande do Sul,
avoid the requirement of steady-state detection procedures. UFRGS, 2003.
However, with the present resources, DRTO solutions would Kern, G., Summiting with multivariable predictive control,
demand a significant computational effort and, possibly, Hydrocarbon Processing, 2007
many numerical issues should be addressed before this Ogata, K., Engenharia de Controle Moderno, Ed.
technology can be widely used in industrial applications. Prentice/Hall do Brasil, 1982.
OPC Foundation, 2008, Site:
5. CONCLUSIONS <http://www.opcfoundation.org/>, Accessed in 03/17/08.
This article has discussed some challenges associated with
advanced process control and optimization in petroleum
Qin, S. and Badgwell, T., A survey of industrial model
predictive control technology, Control Engineering
Practice 11, pp. 733-764, 2003.
Schmidt et al., BR-Tuning Ferramenta para sintonia de
controladores PID, Primeiro CICAP – Congresso de
Instrumentação, Controle e Automação da
PETROBRAS, May, Rio de Janeiro, 2008.
Trierweiler, J. and Farina, L., RPN tuning strategy for model
predictive control, Journal of Process Control, Oxford-
Inglaterra - Elsevier, Vol. 13, pp. 591-598, 2003.
Zanin, A. and Moro, L., Gestão da Automação Industrial no
Refino, Rio Oil&Gás 2004, IBP, Rio de Janeiro, 2004.