Thesis Abstract

<p>         <b>ABSTRACT</b></p><p> Fluid catalytic cracking (FCC) is one of the most important processes in the petroleum refining industry for the conversion of heavy gasoil to gasoline and diesel. Furthermore, valuable gases such as ethylene, propylene and isobutylene are produced. The performance of the FCC units plays a major role on the overall economics of refinery plants. Any improvement in operation or control of FCC units will result in dramatic economic benefits. Present studies are concerned with the general behaviour of the industrial FCC plant, and have dealt with the modelling of the FCC units, which are very useful in elucidating the main characteristics of these systems for better design, operation, and control. Traditional control theory is no longer suitable for the increasingly sophisticated operating conditions and product specifications of the FCC unit. Due to the large economic benefits, these trends make the process control more challenging. There is now strong demand for advanced control strategies with higher quality to meet the challenges imposed by the growing technological and market competition. According to these highlights, the thesis objectives were to develop a new mathematical model for the FCC process, which was used to study the dynamic behaviour of the process and to demonstrate the benefits of the advanced control (particularly Model Predictive Control based on the nonlinear process model) for the FCC unit. The model describes the seven main sections of the entire FCC unit&nbsp;</p><p>1) the feed and preheating system,&nbsp;</p><p>(2) reactor,&nbsp;</p><p>(3) regenerator,&nbsp;</p><p>(4) air blower,&nbsp;</p><p>(5) wet gas compressor,&nbsp;</p><p>(6) catalyst circulation lines and&nbsp;</p><p>(7) main fractionators.&nbsp;</p><p>The novelty of the developed model consists in that besides the complex dynamics of the reactorregenerator system, it includes the dynamic model of the fractionator, as well as a new five lump kinetic model for the riser, which incorporates the temperature effect on the reaction kinetics; hence, it is able to predict the final production rate of the main products (gasoline and diesel), and can be used to analyze the effect of changing process conditions on the product distribution. The FCC unit model has been developed incorporating the temperature effect on reactor kinetics reference construction and operation data from an industrial unit.&nbsp;</p><p>The resulting global model of the FCC unit is described by a complex system of partial-differential-equations, which was solved by discretising the kinetic models in the riser and regenerator on a fixed grid along the height of the units, using finite differences. The resulting model is a high order DAE, with 942 ODEs (142 from material and energy balances and 800 resulting from the discretisation of the kinetic models). The model offers the possibility of investigating the way that advanced control strategies can be implemented, while also ensuring that the operation of the unit is environmentally safe. All the investigated disturbances showed considerable influence on the products composition. Taking into account the very high volume production of an industrial FCC unit, these disturbances can have a significant economic impact. The fresh feed coke formation factor is one of the most important disturbances analysed. It shows significant effect on the process variables. The objective regarding the control of the unit has to consider not only to improve productivity by increasing the reaction temperature, but also to assure that the operation of the unit is environmentally safe, by keeping the concentration of CO in the stack gas below a certain limit. <br></p><p> The model was used to investigate different control input-output pairing using classical controllability analysis based on relative gain array (RGA). Several multi-loop control schemes were first investigated by implementing advanced PID control using anti-windup. A tuning approach for the simultaneous tuning of multiple interacting PID controllers was proposed using a genetic algorithm based nonlinear optimisation approach. Linear model predictive control (LMPC) was investigated as a potential multi-variate control scheme applicable for the FCCU, using classical square as well as novel non-square control structures. The analysis of the LMPC control performance highlighted that although the multivariate nature of the MPC approach using manipulated and controlled outputs which satisfy controllability criteria based on RGA analysis can enhance the control performance, by decreasing the coupling between the individual low level control loops operated by the higher level MPC. However the limitations of using the linear model in the MPC scheme were also highlighted and hence a nonlinear model based predictive control scheme was developed and evaluated. Results demonstrate that using modern nonlinear model predictive control (NMPC) approaches based on state-of-the-art optimization algorithms and software the advanced control of complex chemical processes, such as the FCCU can be brought into the realm of possibility.</p><p>&nbsp;The proposed nonlinear control scheme was applied to the control of the highly nonlinear FCC process. Simulations showed that the non-linear model-based MPC is highly effective, and coped well with the complexity and nonlinearities of the FCC process. Both NMPC and LMPC were superior to PID control, and it was shown that the NMPC provided better performance compared to LMPC. A novel economic criteria-based NMPC (ENMPC) was proposed and demonstrated the benefits of controlling the plant based on economic criteria rather than following predetermined setpoint trajectories. The new concept of differentiating between beneficial (good) and harmful (bad) disturbances was introduced and it was shown that regulating the plant at it predetermined operating conditions when beneficial disturbances occur can actually decrease control performance. The novel ENMPC scheme is inherently able to differentiate between good and bad disturbances and exploit potential beneficial disturbances continuously adapting the plant operating conditions to achieve economic optimisation and satisfy environmental and operating constraints. These results have demonstrated that the control structures proposed may be applied in industry in the form of a new scheme for controlling highly complex chemical processes with significant economic benefits. <br></p>

Thesis Overview

<p> <b>1.0 INTRODUCTION&nbsp;</b></p><p><b>1.1 BACKGROUND STUDY&nbsp;&nbsp;</b></p><p>Worldwide, the fluid catalytic cracker (FCC) is the workhorse of the modern refinery. Its function is to convert heavy hydrocarbon petroleum fractions into a slate of more usable range of products, and must adapt to seasonal, environmental and other changing demand patterns, such as for gasoline, LPG and diesel. Furthermore, valuable gases such as ethylene, propylene and isobutylene are produced. The performance of the FCC units plays a major role in the overall economics of refinery plants (Alhumaizi and Elnashaie, 1997). Any improvement in operation or control of FCC units will result in dramatic economic benefits. First commercialised over half a century ago, the FCC is still evolving. Improvements in the technology, as well as changing feed stocks and product requirements continue to drive this evolution. Control of the FCC has been, and continues to be, a challenging and important problem.&nbsp;</p><p>As will be seen, its steady state behaviour is highly nonlinear, leading to multiple steady states and input multiplicities. In earlier years, before the development of zeolite catalysts, the major control problem was one of stabilisation, of just keeping the unit running. Later with zeolite catalysts, the emphasis shifted to increasing production rates in the face of unit constraints and to handling heavier feeds (Arbel et al., 1995). New world trends in product demands, and to meet more severe legislation about fuel compositions raised the significance of controlling FCC product selectivity. The different product slates of the FCC process are the consequence of the complex interplay between reactions, such as cracking, isomerisation, hydrogen transfer, oligomerisation, etc. The complexity of gas oil mixtures, which are the typical FCC feeds, makes it extremely difficult to characterise and describe the inherent kinetics at a molecular level.&nbsp;</p><p>Hence, one is forced to examine generalities rather than the details. One of the methods used to do this, is to consider the behaviour of groups of compounds as a unit. In this way, similar components are grouped into a few “cuts” or “lumps”. Therefore, the study of the reactions involved in the catalytic cracking process has followed the lumping methodology (Serti-Bionda et al., 2010). The FCC unit consists of two interconnected gas-solid fluidised bed reactors. The riser reactor, where almost all the endothermic cracking reactions and coke deposition on the catalyst occur, and the regenerator reactor, where air is used to burn off the coke accumulated on the catalyst. The heat produced is carried by the catalyst from the regenerator to the reactor (Emad and Elnashaie, 1997). Thus, in addition to reactivating the catalyst, the regenerator provides the heat required by the endothermic cracking reactions. The region of economically attractive operational conditions is determined by both the properties of the feed stocks, catalyst and the desired product distribution requirements. In practice, the optimisation of the FCC unit to the desired range of products is usually carried out by trial and error. The disadvantage of this approach is that the transition from one state to the other must be gradual, and is not always successful, because of the complex interactions between the two reactors. As a result, it could lead to loss of production and consequently affect profits.&nbsp;</p><p>Most of the economic gain from FCC control development has come from the optimisation level, with the regulation system simply providing stable, responsive, and safe operation. The problem is to find regulator schemes that are effective, economically justified, related to existing practice, and able to provide an adequate operator interface when desired. Most studies (Kurihara, 1967; Iscoll, 1970; Lee and Kugelman, 1973; Eng et al., 1974; Edwards and Kim, 1988; McFarlane et al., 1990; Krishna and Parking, 1985; Elshishini and Elnashaie, 1990) concerning FCC units have dealt with the process control based on a simplified reactor-regenerator model, which in principle incorporates major observed dynamics. Any FCC control should maintain a suitable reactor temperature distribution so as to achieve good product characteristics. The regenerator temperature profile should also be bounded so as to prevent abnormal combustion and excessive temperatures. At the same time, energy and material balances must be maintained between the two parts of the unit (Lopez-Isunza, 1992). The modelling of complex chemical systems in the simulation of process dynamics and control has been motivated by the economic incentives for improvement of plant operation and plant design. Presently, studies are concerned with the general behaviour of industrial FCC units; these research efforts deal with the modelling of FCC units, which are very useful in elucidating the main characteristics of these units for better design, operation, and control.&nbsp;</p><p>Traditional control theory is no longer suitable for the FCC unit‟s increasingly sophisticated operating conditions and product specifications (Jia et al., 2003). Due to the large economic benefits, these trends make the process control more challenging. There is now a strong demand for advanced control strategies with higher control quality to meet the challenges imposed by the growing technological and market competition (Liao, 2008). <br></p><p> Demand in the world oil markets is primarily for motor gasoline and other high quality fuels. Gasoline and other high products can be sold at highest prices; therefore, heavy oils are less valuable than the light fractions. Crude distillation separates crude oil into more useful fractions. However, distillation alone cannot meet the demand for high quality fuels. Figure 1.1 shows the range of products demanded by the market from a barrel of crude oil, compared with the products produced by distillation alone. An important job of modern refineries is to convert oil from the „bottom of the barrel‟ to gasoline and other marketable products. Since a typical FCC unit can convert a large amount of feedstock into more valuable products, the overall economic benefits of a refinery could be considerably increased if proper control and optimisation strategies are implemented. But, analysis and control of FCC processes have been known as challenging problems due to the following process characteristics;&nbsp;</p><p>(1) very complicated and little known hydrodynamics,&nbsp;</p><p>(2) complex kinetics of both cracking and coke burning reactions,&nbsp;</p><p>(3) strong interactions between the reactor and the regenerator, and&nbsp;</p><p>(4) many operating constraints. However, the large throughput of FCCU, the change in operating conditions, and the substantial economic benefits are the motivation behind this research (Han et al., 2000).<br></p><p> <b>1.2&nbsp;RESEARCH AIM AND OBJECTIVE</b>S</p><p>The aim of this research project is to develop a mathematical model that can simulate the behaviour of the FCC unit, which consists of feed and preheat system, reactor (riser and stripper), regenerator, air blower, wet gas compressor catalyst circulation lines, and the main fractionators. The model will be subsequently used in studies of control and economic optimisation. The developed model deals with the complex dynamics of the reactor-regenerator system, and also includes the dynamic model of the fractionator, as well as a detailed five lump kinetic model for the riser (namely: gas oil, gasoline, diesel, LPG and coke).&nbsp;</p><p>This model is able to predict and describe the compositions of the final production rate, and the distribution of the main components in the final product. This allows the estimation of economic factors, related to the operation of the FCCU. Seven objectives have been identified that lead to a logical progression through the research:&nbsp;</p><p>a) to gain knowledge and understanding of FCC unit behaviour by developing a user-friendly, process simulator using an object-oriented programming environment. The simulator can be used to understand the process dynamics, and perform operator training, control structure design, controller tuning, through a comprehensive literature review;&nbsp;</p><p>b) implement and evaluate the 5-lump kinetic model in the FCC global model, to simulate the dynamic behaviour of open and closed loops using decoupled advanced PID control algorithms;&nbsp;</p><p>c) design and evaluate control schemes for controlled and manipulated variables in order to choose the best control pairs by using an analysis tool, known as relative gain array (RGA);&nbsp;</p><p>d) develop a PID tuning algorithm for a Multi Input Multi Output (MIMO) system by using model based optimisation;&nbsp;</p><p>e) design and implement linear and nonlinear Model Predictive Control (MPC);</p><p>&nbsp;f) derive and evaluate different economic optimisation objectives; and&nbsp;</p><p>g) implement a hierarchical Real Time Optimisation (RTO) algorithm and perform evaluation for different economic objectives. <br></p><p> <b><small></small></b><small><b>1.3</b></small><b><small></small></b> <b>MAIN CONTRIBUTION OF THE THESIS</b></p><p>&nbsp;The main contributions of this thesis are summarised in the following lists:&nbsp;</p><p>a) An original mathematical model for the FCC process was developed based on the momentum, mass, and energy dynamic balances. It incorporates process hydrodynamics, heat transfer, mass transfer and catalytic cracking kinetics based on a lumping strategy, which lumps molecules and reactions according by their boiling point and treated as pseudo-components for a global description of the phenomena taking place in the reactor.&nbsp;</p><p>The process is multivariable, strongly nonlinear, highly interactive, and subject to many operational, safety and environmental constraints, posing challenging control problems. The model was implemented in the C programming language for efficient solution, and compiled in Matlab/Simulink language programming, because it provides a convenient graphical user interface. Moreover, the model has been used to study the dynamic behaviour of the process, control the output variables and economic optimisation. The global model of the FCCU is described by a complex system of partial-differentialequations, which was solved by discretising the kinetic models in the riser and regenerator on a fixed grid along the height of the units, using finite differences.&nbsp;</p><p>The resultant model is described by a complex system of a higher order differential-algebraic-equation (DAE), with 142 ODEs (from material and energy balances and 800 algebraic equations resulting from the discretisation of the kinetic models).&nbsp;</p><p>b) A flexible process simulator was developed that can be used to show how the open and closed loop control systems perform in the case of disturbances and model uncertainty. The developed simulator enables engineering and technical personnel to carry out research on the design, operation, performance, and development of a proper control system for a modern catalytic cracking unit. It could also act as an efficient tool for training operating personnel.&nbsp;</p><p>c) An automatic tool is proposed to tune PID controllers in a MIMO process based on a Genetic Algorithm (GA). The tuning of several interacting controllers in complex industrial plants is a challenge to process engineers and operators. The success of this task depends on complete knowledge of plant behaviour and control requirements, which can present strong interactions among variables, non-linearity and conflicting objectives. An advantage of the proposed scheme is that a coupled MIMO process with several control loops can have all the controllers tuned in a unified way as a full MIMO controller. In the development phase, the problem of tuning n regulatory control loops was modelled as an optimisation problem, where tuning one control loop may damage the performance of the remaining ones. The solution of the multi-objective problem by the weighted sum approach is possible, since each loop is locally evaluated by a function that considers the integral squared error (ISE). It is worthwhile to note that this modelling is not restricted to PID control. The objective function can be used as a basis for the design and tuning of general, linear or non-linear multivariable controllers.&nbsp;</p><p>d) To introduce an analysis method based on the relative gain array (RGA) as a tool to help select variable pairings for decentralised multivariable control structures.&nbsp;</p><p>e) To evaluate the performance of Model Predictive Control (MPC) based on linearised global reactor-regenerator-main fractionator FCCU global model. Various square and non-square control structures are investigated, and the tuning of the MPC is evaluated.&nbsp;</p><p>f) To develop an efficient real-time nonlinear model predictive control (NMPC) strategy, based on efficient multiple shooting optimization and a real-time iteration scheme, and to evaluate the performance of the proposed scheme in the case of the simulated complete FCCU.&nbsp;</p><p>g) To develop a novel NMPC scheme based on the on line optimization of overall process economics related objectives. To demonstrate the benefits of controlling the FCCU based on real time economic decisions, and exploiting the beneficial effects of „disturbances‟ if any, rather than simply tracking set point trajectories determined off line.&nbsp;</p>