Dynamic Matrix Control (DMC) has proven to be a powerful tool for optimal regulation of chemical processes under constrained conditions. The internal model of this predictive controller is based on step response measurements at an average operating point. As the process moves away from this point, however, control becomes sub-optimal due to process non-linearity. If DMC is made adaptive, it can be expected to perform well even in the presence of uncertainties, non-linearities and time-vary ing process parameters. This project examines modelling and control issues for a complex multivariable industrial operator training plant, and develops and applies a method for adapting the controller on-line to account for non-linearity. A two-input/two-output sub-system of the Training Plant was considered. A special technique had to be developed to deal with the integrating nature of this system - that is, its production of ramp outputs for step inputs. The project included the commissioning of the process equipment and the addition of instrumentation and interfacing to a SCADA system which has been developed in the School of Chemical Engineering. / Thesis (M.Sc.Eng.)-University of Natal, Durban, 2001.

In the thesis, two different model predictive control (MPC) strategies are investigated for linear systems with uncertainty in the presence of constraints: namely robust MPC and stochastic MPC. Firstly, a Youla Parameter is integrated into an efficient robust MPC algorithm. It is demonstrated that even in the constrained cases, the use of the Youla Parameter can desensitize the costs to the effect of uncertainty while not affecting the nominal performance, and hence it strengthens the robustness of the MPC strategy. Since the controller u = K x + c can offer many advantages and is used across the thesis, the work provides two solutions to the problem when the unconstrained nominal LQ-optimal feedback K cannot stabilise the whole class of system models. The work develops two stochastic tube approaches to account for probabilistic constraints. By using a semi closed-loop paradigm, the nominal and the error dynamics are analyzed separately, and this makes it possible to compute the tube scalings offline. First, ellipsoidal tubes are considered. The evolution for the tube scalings is simplified to be affine and using Markov Chain model, the probabilistic tube scalings can be calculated to tighten the constraints on the nominal. The online algorithm can be formulated into a quadratic programming (QP) problem and the MPC strategy is closed-loop stable. Following that, a direct way to compute the tube scalings is studied. It makes use of the information on the distribution of the uncertainty explicitly. The tubes do not take a particular shape but are defined implicitly by tightened constraints. This stochastic MPC strategy leads to a non-conservative performance in the sense that the probability of constraint violation can be as large as is allowed. It also ensures the recursive feasibility and closed-loop stability, and is extended to the output feedback case.

Growing concerns about the environmental impact of fossil fuel energy and improvements in both the cost and performance of wind turbine technologies has spurred a sharp expansion in wind energy generation. However, both the increasing size of wind farms and the increased contribution of wind energy to the overall electricity generation market has created new challenges. As wind farms grow in size and power density, the aerodynamic wake interactions that occur between neighboring turbines become increasingly important in characterizing the unsteady turbine loads and power output of the farm. Turbine wake interactions also impact variability of farm power generation, acting either to increase variability or decrease variability depending on the wind farm control algorithm. In this dissertation, both the unsteady vortex wake loading and the effect of wake interaction on farm power variability are investigated in order to better understand the fundamental physics that govern these processes and to better control wind farm operations to mitigate negative effects of wake interaction. The first part of the dissertation examines the effect of wake interactions between neighboring turbines on the variability in power output of a wind farm, demonstrating that turbine wake interactions can have a beneficial effect on reducing wind farm variability if the farm is properly controlled. In order to balance multiple objectives, such as maximizing farm power generation while reducing power variability, a model predictive control (MPC) technique with a novel farm power variability minimization objective function is utilized. The controller operation is influenced by a number of different time scales, including the MPC time horizon, the delay time between turbines, and the fluctuation time scales inherent in the incident wind. In the current research, a non-linear MPC technique is developed and used to investigate the effect of three time scales on wind farm operation and on variability in farm power output. The goal of the proposed controller is to explore the behavior of an "ideal" farm-level MPC controller with different wind, delay and horizon time scales and to examine the reduction of system power variability that is possible in such a controller by effective use of wake interactions. The second part of the dissertation addresses the unsteady vortex loading on a downstream turbine caused by the interaction of the turbine blades with coherent vortex structures found within the upstream turbine wake. Periodic, stochastic, and transient loads all have an impact on the lifetime of the wind turbine blades and drivetrain. Vortex cutting (or vortex chopping) is a type of stochastic load that is commonly observed when a propeller or blade passes through a vortex structure and the blade width is of the same order of magnitude as the vortex core diameter. A series of Navier-Stokes simulations of vortex cutting with and without axial flow are presented. The goal of this research is to better understand the challenging physics of vortex cutting by the blade rotor, as well as to develop a simple, physics-based, validated expression to characterize the unsteady force induced by vortex

La force de diélectrophorèse (DEP) est utilisée pour manipuler, séparer et positionner différent types des particules (cellules, bactéries, nanotubes de carbone). Dans le but d étudieret de simuler une loi de commande permettant le suivi de la trajectoire d une particule soumise à la force DEP un modèle est nécessaire. Les méthodes utilisées pour simuler la force DEP sont généralement basées soit sur des simulateurs à éléments finis (FEM), soit sur des équations analytiques. Les simulateurs FEM ne permettent pas la variation des paramètres (tensions électriques) lors du calcul de la trajectoire et les équations analytiques sont limitées à des géométries simples des électrodes. Dans ce manuscrit, une méthode hybride basée sur les calculs FEM et analytique est proposée. Cette méthode permet de simuler la trajectoire d une particule en utilisant des géométries complexes et en variant les tensions électriques lors de la simulation. Ce modèle est ensuite validé en le comparant à des relevés expérimentaux. Finalement, une loi de commande, basée sur la commande prédictive généralisée (GPC) est proposée dans le but de contrôler la trajectoire, en profitant de la grande dynamique du déplacement de la particule, et ce malgré les non-linéarités. Cette loi de commande a été validée par des résultats de simulations et une comparaison avec une loi de commande classique. / Micro and nano-particles can be trapped by a non uniform electric field through the effect of dielectrophoretic (DEP) principle. Dielectrophoresis is used to separate, manipulate and detect micro particles in several domains, such as in biological or Carbon Nano-Tubes (CNTs) manipulations. To study and simulate a vision based closed loop control law in order to control the trajectory of micro objects using DEP a numeric model is required. Current methods to simulate the trajectory of micro-particles under a DEP force field are based on finite element modeling (FEM) which requires new simulations when one of its parameters, like the electric voltage, is changed, or on analytic equations which is limited to very simple geometries. In the first section of this manuscript, we propose a hybrid method between analytic and numeric calculation able to simulate complex geometries and to easily change electrode voltage along the trajectory. This numeric model is, then, validated by comparing it with several experimental results. Finally, a control strategy based on the generalized predictive control method is proposed with the aim of controlling the trajectory, taking advantage of the high dynamics despite the non linearity. This control law has been validated by simulation and compared to classical control strategy.

Les installations d'épuration des eaux usées sont des systèmes de grande dimension, non linéaires, sujets à des perturbations importantes en flux et en charge. Une commande prédictive (MPC) a été appliquée au Benchmark BSM1 qui est un environnement de simulation qui définit une installation d'épuration. Une identification en boucle ouverte d'une station d'épuration d'eaux usées a été réalisée pour déterminer un modèle linéaire en se basant sur un ensemble de mesures entrée sortie du Benchmark BSM1. Les réponses indicielles en boucle ouverte ont été obtenues à partir de variation échelon des entrées autour de leurs valeurs stationnaires. Le modèle tient compte des non-linéarités à travers des paramètres variables. Les réponses indicielles obtenues permettent de déterminer par optimisation les fonctions de transfert continues correspondantes. Ces fonctions de transfert peuvent être regroupées en cinq modèles mathématiques. Des fonctions de transfert continues de premier ordre, de premier ordre avec un intégrateur, des réponses inverses, de second ordre et de second ordre avec zéro représentant les réponses indicielles ont été identifiées. Les valeurs numériques des coefficients de chaque modèle choisi ont été calculées par un critère des moindres carrés. La commande prédictive (MPC) utilise le modèle obtenu comme un modèle interne pour commander le procédé. Deux stratégies de la commande prédictive DMC et QDMC d'une station d'épuration avec ou sans compensation par anticipation ont été testées. La commande par anticipation est utilisée pour réduire l'effet de deux perturbations mesurées, le débit entrant et la concentration entrante en ammonium, sur le système / Wastewater treatment processes are large scale, non linear systems, submitted to important disturbances of influent flow rate and load. Model predictive control (MPC) widely used industrial technique for advanced multivariable control, has been applied to the Benchmark Simulation Model 1 (BSM1) simulation benchmark of wastewater treatment process. An open loop identification method has been developed to determine a linear model for a set of input-output measurements of the process. All the step responses have been obtained in open loop from step variations of the manipulated inputs and measured disturbances around their steady state values. The non-linearities of the model are taken into account by variable parameters. The step responses coefficient obtained make it possible to determine by optimization the corresponding transfer functions. That functions are classified by five mathematical models, such as : first order, first order with integrator, inverse response, second order and second order with zero. The numerical values of coefficients of each model selected were calculated using a least squares criterion. Model predictive control (MPC) uses the resulting model as an internal model to control the process. Dynamic matrix control DMC and quadratic dynamic matrix control QDMC predictive control strategies, in the absence and presence of feedforward compensation, have been tested. Two measured disturbances have been used for feedforward control, the influent flow rate and ammonium concentration

Le craquage catalytique à lit fluidisé (FCC) est l'un des procédés les plus importants au sein d'une raffinerie moderne et joue un rôle économique primordial. Le fonctionnement du FCC pose des problèmes d'opération liés à sa complexité. L'étude a porté sur la simulation du FCC, sa commande prédictive multivariable et l'estimation de paramètres-clés. Après une revue de la littérature sur les FCC et les différentes approches de modélisation ainsi que des cinétiques de craquage, un modèle du FCC qui intègre les dynamiques importantes a été choisi pour les besoins de la commande prédictive. La simulation du riser a été effectuée pour différents modèles de craquage et a montré de grandes disparités entre modèles, créant une difficulté à définir un modèle général de riser pour les FCC. Outre le nombre de groupes considérés, les différences concernent la chaleur de réaction globale, les lois de formation de coke sur le catalyseur et la désactivation de ce dernier. Des algorithmes de commande prédictive linéaire et non linéaire basée sur le modèle ont été utilisés pour commander le FCC en tenant compte de sa nature multivariable et des contraintes imposées aux variables manipulées. Les sorties commandées, température en haut du riser et température du régénérateur ont été maintenues proches des consignes, tant en régulation qu'en poursuite, tout en respectant les contraintes portant sur les deux variables manipulées, le débit de catalyseur régénéré et le débit d'air entrant dans le régénérateur. Une commande à trois entrées manipulées, incluant le débit d'alimentation, a également été testée avec succès. La commande prédictive linéaire avec observateur a fourni des résultats encore meilleurs que la commande linéaire quadratique. La commande prédictive non linéaire a été testée mais présente des problèmes pour une implantation en temps réel. L'estimation du coke sur le catalyseur a été réalisée par le filtre de Kalman étendu, mais les erreurs d'estimation sont importantes, probablement à cause du choix insuffisant des mesures effectuées. L'ensemble de l'étude a montré que la commande avancée prédictive du FCC est performante et doit être recommandée, mais peut encore être améliorée en particulier par son réglage et l'estimation des états / Fluid catalytic cracking (FCC) is one of the most important processes in a modern refinery and is of essential economic importance. The FCC operation presents difficulties related to its complexity. The study was related to its simulation, multivariable control and estimation of key parameters. After a litterature review of the FCC, the different approaches of modelling and cracking kinetics, a FCC model that takes into account the important dynamics was chosen for model predictive control purposes. The riser simulation was carried out for different cracking models and shows great differences between these models, which makes it difficult to define a general riser model for the FCC. Besides the number of lumps, differences deal with the global heat of reaction, the coke formation laws and its deactivation functions. Linear and nonlinear model predictive algorithms were used for FCC control taking into account its multivariable nature and the constraints imposed on the manipulated variables. The controlled outputs, temperature at the riser top and temperature in the regenerator were maintained close to their respective set points in regulation and tracking modes while respecting the constraints on the two manipulated variable, the flow rate of regenerated catalyst and the flow rate of air entering the regenerator. A control with three manipulated variables including the feed flow rate was also successfully tested. Linear predictive control with an observer gave better results than linear quadratic control. Nonlinear predictive control was tested but presents problems for real time implementation. The estimation of coke on the catalyst was carried out using extended Kalman filter, but the estimation errors are important, probably due to an insufficient choice of measurements. The overall study showed that advanced predictive control of the FCC is efficient and must be recommended, but it can still be improved upon particularly by its tuning and state estimation

In this thesis, we consider the trajectory planning of an autonomous vehicle to cross an intersection within a given time interval. The vehicle communicates its sensordata to a central coordinator which then computes the trajectory for the given time horizon and sends it back to the vehicle. We consider a realistic scenario in which the communication links are unreliable, the evolution of the state has noise (e.g., due to the model simplification and environmental disturbances), and the observationis noisy (e.g., due to noisy sensing and/or delayed information). The intersection crossing is modeled as a chance constraint problem and the stochastic noise evolution is restricted by a terminal constraint. The communication impairments are modeled as packet drop probabilities and Kalman estimation techniques are used for predicting the states in the presence of state and observation noises. A robust sub-optimalsolution is obtained using convex optimization methods which ensures that the intersection is crossed by the vehicle in the given time interval with very low chance of failure.

The development of Advanced Driving Assistance Systems (ADAS) produces comfort and safety through the application of several control theories. One of these systems is the Adaptive Cruise Control (ACC). In this work, a distribution of two control loops of such system is developed for an embedded application to a vehicle. The vehicle model was estimated using the system identification theory. An outer loop control manages the radar data to compute a suitable cruise speed, and an inner loop control aims for the vehicle to reach the cruise speed given a desired performance. For the inner loop, it is used two different approaches of model predictive control: a finite horizon prediction control, known as MPC, and an infinite horizon prediction control, known as IHMPC. Both controllers were embedded in a microcontroller able to communicate directly with the electronic unit of the vehicle. This work validates its controllers using simulations with varying systems and practical experiments with the aid of a dynamometer. Both predictive controllers had a satisfactory performance, providing safety to the passengers. / A inclusão de sistemas avançados para assistência de direção (ADAS) tem beneficiado o conforto e segurança através da aplicação de diversas teorias de controle. Um destes sistemas é o Sistema de Controle de Cruzeiro Adaptativo. Neste trabalho, é usado uma distribuição de duas malhas de controle para uma implementação embarcada em um carro de um Controle de Cruzeiro Adaptativo. O modelo do veículo foi estimado usando a teoria de identificação de sistemas. O controle da malha externa utiliza dados de um radar para calcular uma velocidade de cruzeiro apropriada, enquanto o controle da malha interna busca o acionamento do veículo para atingir a velocidade de cruzeiro com um desempenho desejado. Para a malha interna, é utilizado duas abordagens do controle preditivo baseado em modelo: um controle com horizonte de predição finito, e um controle com horizonte de predição infinito, conhecido como IHMPC. Ambos controladores foram embarcados em um microcontrolador capaz de comunicar diretamente com a unidade eletrônica do veículo. Este trabalho valida estes controladores através de simulações com sistemas variantes e experimentos práticos com o auxílio de um dinamômetro. Ambos controladores preditivos apresentaram desempenho satisfatório, fornecendo segurança para os passageiros.

Este estudo mostra o desenvolvimento de um controlador da classe MPC Model Predictive Control, ou controle preditivo com modelo, para ser utilizado no reator de amônia da Unidade de Fertilizantes Nitrogenados da Bahia FAFEN-BA, da PETROBRAS, localizada em Camaçari/BA. A estratégia de controle visa manter as temperaturas de saída de cada um dos leitos catalíticos do reator dentro de limites adequados através da manipulação das válvulas de controle instaladas na corrente de alimentação do equipamento. O controlador escolhido foi de horizonte de predição infinito com faixas nas variáveis controladas. Adicionalmente, o controlador contém, em uma única camada, um componente de otimização econômica com o objetivo de maximizar o teor de amônia na saída do reator. A função econômica que dá a direção de otimização consiste em um modelo rigoroso de estado estacionário do reator capaz de calcular a fração molar de amônia na saída do equipamento quando são conhecidas as condições da corrente de alimentação e o valor das variáveis manipuladas do controlador. Os resultados das simulações mostraram que o controlador proposto tem bom desempenho, tanto sob o aspecto de controle, no sentido de controlar o sistema quando este sofre perturbações, quanto sob a ótica de otimização econômica, maximizando a conversão de reagentes em amônia sempre que existem graus de liberdade disponíveis no sistema. Foi verificado que a consideração de um MPC de horizonte de predição infinito elimina a necessidade de considerar o gradiente reduzido da função econômica na função objetivo do controlador. Uma sintonia adequada do controlador permite que se considere o gradiente completo da função econômica sem que haja desvio permanente, ou offset, nas variáveis controladas mesmo quando o ponto ótimo de operação se encontra além da faixa de controle. / This study shows the development of a Model Predictive Control (MPC) to the ammonia reactor of PETROBRAS nitrogen fertilizers unit FAFEN-BA that is located in Camaçari/BA, Brazil. The main goal of the control strategy is to keep the temperature at the outlet of the catalyst beds inside adequate ranges by manipulating the feed flow rates to the reactor beds. It has been chosen an infinite horizon controller with control zones and an economic objective. The control and economic optimization are performed in a single layer structure where the objective is to maximize the ammonia content in the reactor outlet stream. The economic function which provides the optimization direction is based on a steady state rigorous model of the reactor that evaluates the ammonia molar fraction at the outlet stream assuming that the feed stream conditions and the manipulated variables are known. The proposed controller shows satisfactory performance in simulations either controlling the system when it faces external disturbances or optimizing the economic goal by increasing the ammonia conversion when degrees of freedom are available. It is shown that the adoption of the infinite horizon MPC eliminates the need to consider the reduced gradient of the economic function in the cost function of the controller. The proper tuning of the controller allows the consideration of the full gradient of economic function without producing offset in the controlled outputs even when the optimum operating point lays outside the control zones.

Em geral, nos processos de policondensação há formação de um produto volátil, conhecido como condensado. A remoção deste produto é necessária a fim de favorecer o crescimento das cadeias poliméricas. O controle da temperatura do reator deve ser implementado de modo a favorecer a remoção do condensado e de manter a qualidade do produto. A aplicação experimental do controlador preditivo não linear (NMPC) a um reator de policondensação é apresentada neste trabalho. Devido às restrições de tempo de processamento, foi proposto um modelo simplificado, denominado Modelo A, obtido considerando o reator como sendo bem misturado acoplado a um tanque de condensado em que Etilenoglicol é acumulado. Supõe-se também que o metanol que deixa o reator é removido completamente do sistema, como se houvesse uma separação perfeita. O modelo A foi ajustado a dados experimentais obtidos a partir de ensaios com perturbação degrau e PRBS (Pseudo Random Binary Sequence). Em conclusão dos experimentos em malha fechada verificou-se que o controlador NMPC com o modelo A apresenta dificuldades de controle quando há reações químicas no sistema. Sendo assim, foi proposto um novo modelo, denominado modelo B, em que assume-se que os componentes voláteis, etilenoglicol e metanol, acumulam-se ambos em um tanque de condensado, podendo retornar ao reator. Foram realizados estudos de simulação para verificar como os modelos A e B interferem na viabilidade das trajetórias de temperatura. Os estudos mostraram que a presença de Metanol no reator, mesmo em pequenas quantidades, tem um efeito mais acentuado na temperatura do que a presença de Etilenoglicol. No entanto, o desempenho do controlador NMPC não foi muito bom quando comparado ao desempenho de um controlador PID. Foi proposta uma nova estratégia, com retroalimentação simples de estado. Nesta estratégia, a temperatura inicial do reator no modelo de referência é atualizada com o valor medido desta temperatura. O controlador NMPC com retroalimentação de estado apresentou um desempenho superior aos controladores NMPC anteriores e PID. O controlador NMPC com retroalimentação de estado mostrou-se mais robusto que o PID porque tem um desempenho equivalente para diferentes trajetórias, sem para isto ter que sofrer alterações nos seus parâmetros de sintonia. O controlador PID sofre uma degradação do seu desempenho, dependendo da trajetória a ser seguida. Isto torna o controlador NMPC com retroalimentação simples de estado mais flexível, pois evitam-se correções no modelo ou nos parâmetros de sintonia para diferentes bateladas. Esta flexibilidade é importante para a industria de polímeros, que muitas vezes trabalha com reatores multiproduto em batelada. / In general, in polycondensation processes there is formation of a volatile product known as condensate. The removal of this product is necessary in order to favor the growth of the polymer chains. The reactor temperature control should be implemented in order to favor condensate removal and to keep product quality. The experimental application of nonlinear predictive controller (NMPC) to a polycondensation reactor is presented in this work. Due to processing time requirements, a simplified model of the system, called model A, obtained by supposing the reactor is a perfect stirred tank coupled to a condensate tank, where the ethyleneglycol that exits the system is accumulated. It is supposed also that the entire methanol that leaves the reactor leaves the system too, as if there was a perfect separation column. Model A was adjusted to experimental data obtained from step response and PRBS (Pseudo Random Binary Sequence) disturbance experiments. In conclusion of the closed loop experiments, it was verified that the NMPC controller with model A presents difficulties when chemical reactions take place in the system. Therefore, a new model, called model B, was proposed, in which it is assumed that both methanol and ethylene glycol are accumulated in a condensate tank, and are allowed to return to the reactor. Simulation studies were carried out in order to verify how models A and B interfere in the feasibility of trajectories. The studies showed that the presence of Methanol, even in small amounts, has a very large effect in temperature, when compared to Ethyleneglycol. However, the NMPC controller performance was not very better than that of a PID. A strategy implementing a simple state feedback was proposed. In this strategy, the initial reactor temperature in the reference model is replaced by its measured value. NMPC with state feedback has a performance better than that of the previous NMPC and PID controllers. NMPC with state feedback appears to be more robust than PID because it has a constant performance for different temperature trajectories, without need of changing its tuning parameters. PID controller instead, shows a degradation of its performance, depending on the trajectory to be followed. This makes the NMPC controller with state feedback more flexible because there is no need to correct the model or the tuning parameters for different batches. This flexibility is important for the polymer industry, that often works with multiproduct batch reactors.