This doctoral thesis presents new ideas and research results on control of marine electric power system. The main motivation for this work is the development of a control system, power management system (PMS) capable to improve the system robustness to blackout, handle major power system faults, minimize the operational cost and keep the power system machinery components under minimal stress in all operational conditions. Today, the electric marine power system tends to have more system functionality implemented in integrated automation systems. The present state of the art type of tools and methods for analyzing marine power systems do only to a limited extent utilize the increased knowledge available within each of the mechanical and electrical engineering disciplines. As the propulsion system is typically consisted of the largest consumers on the vessel, important interactions exists between the PMS and vessel propulsion system. These are interacted through the dynamic positioning (DP) controller, thrust allocation algorithm, local thruster controllers, generators' local frequency and voltage controllers. The PMS interacts with the propulsion system through the following main functions: available power static load control, load rate limiting control and blackout prevention control (i.e. fast load reduction). These functions serve to prevent the blackout and to ensure that the vessel will always have enough power. The PMS interacts with other control systems in order to prevent a blackout and to minimize operational costs. The possibilities to maximize the performance of the vessel, increase the robustness to faults and decrease a component wear-out rate are mainly addressed locally for the individual control systems. The solutions are mainly implicative (for e.g. local thruster control, or DP thrust allocation), and attention has not been given on the interaction between these systems, the power system and PMS. Some of the questions that may arise regarding the system interactions, are as follows: how the PMS functionality may affect a local thruster control, how the local thruster control may affect the power system performance, how some consumers may affect the power system performance in normal operations and thus affect other consumers, how the power system operation may affect the susceptibility to faults and blackout, how various operating and weather conditions may affect the power system performance and thus propulsion performance though the PMS power limiting control, how propulsion performance may affect the overall vessel performance, which kind of faults can be avoided if the control system is re-structured, how to minimize the operational costs and to deal with the conflicting goals. This PhD thesis aims to provide answers to such questions. The main contributions of this PhD thesis are: − A new observer-based fast load reduction system for the blackout prevention control has been proposed. When compared to the existing fast load reduction systems, the proposed controller gives much faster blackout detection rate, high reliability in the detection and faster and more precise load reduction (within 150 miliseconds). − New advanced energy management control strategies for reductions in the operational costs and improved fuel economy of the vessel. − Load limiting controllers for the reduction of thruster wear-out rate. These controllers are based on the probability of torque loss, real-time torque loss and the thruster shaft accelerations. The controllers provide means of redistributing thrust from load fluctuating thrusters to less load fluctuating ones, and may operate independently of the thrust allocation system. Another solution is also proposed where the load limiting controller based on thrust losses is an integrated part of DP thrust allocation algorithm. − A new concept of totally integrated thrust allocation system, local thruster control and power system. These systems are integrated through PMS functionality which is contained within each thruster PLC, thereby distributed among individual controllers, and independent of the communications and dedicated controllers. − Observer-based inertial controller and direct torque-loss controller (soft anti-spin controller) with particular attention to the control of machine wear-out rate. These controller contribute to general shaft speed control of electrical thrusters, generators and main propulsion prime movers. The proposed controllers, estimators and concepts are demonstrated through time-domain simulations performed in MATLAB/SIMULINK. The selected data are typical for the required applications and may differ slightly for the presented cases.

Ce travail propose de nouvelles structures de contrôle pour la téléopération bilatérale à travers des réseaux de communication non dédiés. L’enjeu est donc de concevoir et calculer des structures de commande garantissant la stabilisation et un bon degré de performance en termes de synchronisation (suivi des positions et vitesse) et de transparence (ressenti des forces) sous les retards variables et asymétriques.Nous faisons tout d’abord un tour d’horizon des recherches récentes dans le domaine des systèmes de téléopération et de leurs caractéristiques. Puis, nous considérons des modèles linéaires à plusieurs retards variables pour lesquels nous proposons une approche d’analyse de stabilité par fonctionnelles de Lyapunov-Krasovskii et contrôle robuste H [infinity symbol . Ensuite, trois structures de téléopération seront proposées en temps continu, la comparaison de ces architectures montre que, pour un retard de réseau maximum donné ou calculé, toutes garantissent un suivi de position et vitesse. Les deux dernières, qui utilisent les forces mesurées ou estimées de l’opérateur humain et de l’environnement, garantissent de plus un suivi en force. Au final, la troisième structure (avec proxy) présente la meilleure performance, même si elle demande un peu plus de calcul. Puis, afin d’analyser et d’améliorer les performances de la troisième structure pour des modèles encore plus réalistes, une étude est menée en temps discret, mais aussi sur un modèle non linéaire ou non stationnaire sous perturbations bornées en norme. L’implantation sur la plate-forme est décrite dans un quatrième et dernier chapitre, et puis l’analyse des résultats expérimentaux est alors menée / This PhD thesis is dedicated to the control scheme design of the bilateral teleoperation under asymmetric communication channels: the stabilization and a high-level performance under asymmetric time-varying delays and the perturbations of the human operator and environment. After a review of the recent researches and their features in the field of teleoperation system, a less conservative Lyapunov-Krasovskii functional together with H [infinity symbol] control theory has been applied to linear time delay systems, and then the LMI theorems have been obtained in order to calculate the controllers in the control schemes.Firstly, three novel teleoperation control schemes have been presented. Comparing three architectures, all of them guaranteed the stability and the position tracking thanks to the position/velocity information. Force-reflecting control scheme without or with proxy, in addition, ensured the force tracking by using the estimated/measured force of the human operator and the environment. Here, the control scheme with the proxy got a better performance. Secondly, a discrete-time approach has been developed to analyze the force-reflecting control scheme with proxy and obtain a better system performance. Besides, more general systems with time-varying uncertainties (the polytopic-type uncertainties and the norm-bounded model uncertainties) have been considered. Finally, the experimental test-bench and the real system implementation have been designed, which involved the identification and linearizing control of the subsystems (the master/slave robots). The experimental results have illustrated the effectiveness of the approaches proposed in this thesis

Orientador: Rodrigo Cardim / Resumo: Neste trabalho são apresentados uma série de resultados relacionados com as técnicas de controle para sistemas lineares chaveados incertos que asseguram índices de desempenho e custos garantidos no projeto. Inicialmente a técnica abordada para este estudo consiste na utilização das desigualdades de Lyapunov-Metzler e as propriedades dos sistemas Estritamente Reais Positivos (ERP). São abordados os sistemas Lyapunov-Metzler-ERP (LMERP), que permitem o desenvolvimento de um método de projeto de estabilização para sistemas que apresentam comutação e incertezas no modelo, usando para isto a realimentação do vetor de estado. A análise de estabilidade é descrita por meio de Desigualdades Matriciais Lineares (em inglês: Linear Matrix Inequalities), LMIs, que, quando factíveis, são facilmente resolvidas por meio de ferramentas disponíveis de programação convexa. Neste trabalho trata-se também da síntese via realimentação de estado com chaveamento no ganho que assegura o critério de desempenho Hoo. Para a validação das estratégias de controle mencionadas foram realizadas simulações e experimentos práticos em um sistema de suspensão ativa de bancada e em um sistema ball balancer, equipamentos fabricados pela Quanser. Os resultados comprovam a eficácia dos método propostos tanto nas simulações quanto nos testes realizados em bancada. / Mestre

Le contrôle par mode glissant est une technique d'automatique qui possède une longue histoire, la littérature remontant jusqu'au année 50. Son essence est la suivante : le contrôle est définit comme étant l'image d'une fonction discontinue de la variable de glissement, contraignant le système à évolué sur une variété, le système glisse alors dessus, d'où le nom. Cette variable de glissement est elle définie à partir de l'état du système. Les développements ont mené à la constitution d'une théorie bien établie à propos de cette technique, avec de nombreuses propriétés théoriques fort intéressante. Toutefois ceci ne porte que sur la version continue, c'est à dire quand le contrôle peut changer de valeur à chaque instant. En comparaison la version discrète du ce contrôleur est définie par le fait que la valeur du contrôle ne peut changer qu'à des instants isolés discrets. On a alors une fonction en escalier, constante sur la période d'échantillonnage. Cette situation est rencontrée par exemple lorsque le contrôleur est implémenté à l'aide d'un micro-contrôleur, ce qui est le cas dans nombre d'applications industrielles. Le principal problème avec le mode glissant est l'apparition d'un phénomène largement indésirable, le chattering (ou broutement) avec la version discrète du contrôleur, où même déjà en simulation. Dans ce dernier cas, nous appelons ceci du chattering numérique que nous attribuons à une mauvaise discrétisation du contrôle. L'approche développée ici se focalise sur ce point et est largement inspirée par les travaux effectués en mécanique non régulière, où ce type de comportement a aussi été observé lors de la simulation de système avec frottements et/où impacts. L'idée principale est de discretisé le contrôle de manière implicite et non explicite. Ceci permet d'éliminer le chattering numérique dans les cas simples (systèmes linéaires par exemple) où bien de le réduire grandement. Pour mener à bien l'analyse, des outils provenant de l'analyse convexe ainsi que des inégalités variationnelles en dimension finie sont utilisés. Le contrôleur proposé possède des propriétés intéressantes et proches de celles du temps continu. Ainsi on peut montrer que la variable de glissement est régie par une dynamique stable en temps finie, avec une fonction de Lyapunov. Le contrôle discret convergence vers celui du cas continu quand la période d'échantillonnage tends vers 0. Une atténuation d'éventuelles perturbations de type "matching" peut être établie. Ces travaux ont essentiellement portés sur le contrôle par mode glissant classique. L'algorithme dit twisting a pu être discrétisé avec la même technique et sa stabilité en temps finie grâce à une fonction de Lyapunov a pu être montrée. Ces propriétés ont été vérifiée en simulation, mais aussi de manière expérimentale. Ainsi des essais ont pu être menés sur deux banc d'essai: le premier est basé sur un système electropneumatique où à la fois le contrôle par mode glissant classique ainsi que le twisting ont pu être implémentés. L'objectif étant de suivre une trajectoire de référence. Le second système est un pendule inverse où le système doit être stabilisé à la position d'équilibre instable. Ici seul le contrôleur classique a été testé. L'analyse des données expérimentales a permis de mettre en lumière les performances supérieures des contrôleurs proposés par rapport à ceux classiquement usités. Les objectifs de contrôle sont mieux atteint et le chattering est grandement diminué. / Sliding Mode Control is a control technique with a long history, with research efforts dating back to the 50's. The basic idea is to define the control input as a discontinuous function of the sliding variable, which solely depends on the state, and to constraint the system to evolve on a manifold, hence the term sliding. Over the years a strong theory was build around this technique, but only in continuous time. In our context, this means that control input value can change value at any time. The discrete-time case is when the control input can only change at isolated time instants and the dynamical system on which the control is still a continuous-time process. The control input is therefore a step function. This case appears when the controller is digitally implemented, for instance with the help of a microcontroller. This kind of setup is nowadays ubiquitous in benchmarks and industrial applications. One of the main limitation of the applicability of sliding mode control is the chattering phenomenon that is witnessed when this control technique is applied in practice, but already in simulations. In contrast to previous approaches, we single out the chattering that is already witnessed in simulation, even with no disturbance and with perfect knowledge of the dynamics. This is called the numerical chattering and one of its distinct feature is the constant chattering, or high-frequency bang-bang behavior, of the control input. This naturally induces a chattering of the sliding variable. The claim that this type of chattering is usually predominant and that it is due to a bad discretization of the signum multifunction. The approach developed in this work was inspired by the research effort in the nonsmooth mechanical to properly simulate some systems like those with dry friction and/or unilateral constraints. The main point is to discretize the signum in an implicit fashion, that is its argument is the value of the sliding variable at the end of the next sampling period. With this change, the numerical chattering can be removed in the simplest cases, largely attenuated. The research effort was focused on classical sliding mode controller, rather than the higher order ones. The frameworks used to perform the analysis are convex analysis and variational inequalities. This discrete-time controller enjoys several interesting theoretical properties. First it is finite-time Lyapunov stable: the sliding variable goes to 0 in finite-time. The discrete-time control input converges to the continuous-time one as the sampling period goes to 0. The control action also attenuates the effect of matched perturbations. Also the increase of the gain of the controller does not affect the performances when the system is sliding. The twisting controller can be discretized in the same way and is also finite-time Lyapunov stable. This good theoretical properties have been verified in simulations, but also on experimental setups. Two tests were conducted: the first one on an electropneumatic system, where both the classical first-order sliding mode controller and the twisting algorithm were tested. The objective was to track a reference trajectory. The second one was an inverted pendulum on a cart with only the classical SMC. The goal was to stabilize the system at the unstable equilibrium. The analysis from the data collected during those experiments shows that the proposed controllers perform better than the their explicitly discretized versions. The performances are better and the chattering is effectively reduced.

La stabilité en roulis des véhicules est un problème de sécurité très critique, en particulier pour les poids lourds. Actuellement, la plupart des poids lourds sont équipés de systèmes de barres anti-roulis passifs. Malheureusement ceux-ci ne sont pas capables, en général, de surmonter les situations critiques. Cette thèse se concentre sur les systèmes de barres anti-roulis actifs, qui constituent l'approche la plus communément utilisée pour améliorer la stabilité en roulis des poids lourds. Le travail de recherche de cette thèse est divisé en trois parties principales. Dans la première partie, un modèle intégré est développé, comprenant quatre actionneurs hydrauliques commandés par des servo-valves, associés à un modèle linéaire lacet-roulis de poids lourd. Dans la deuxième partie, le système anti-roulis actif est développé suivant deux méthodologies de contrôle dans le cadre LTI: LQR et Hinfty. Dans la troisième partie, une approche LPV, basée sur le maillage, est utilisée pour synthétiser le contrôleur Hinfty/LPV de barre anti-roulis actif avec des fonctions de pondération dépendant de paramètres variants, à l'aide du progiciel LPVTools. Les résultats de simulation dans les domaines fréquentiel et temporel, ainsi que la validation avec le logiciel de simulation TruckSim, montrent que les systèmes de barres anti-roulis actifs sont une solution réaliste et efficace qui améliore considérablement la stabilité en roulis des poids lourds par rapport aux systèmes de barres anti-roulis passifs. / Vehicle rollover is a very serious problem for the safety of heavy vehicles. Most modern heavy vehicles are equipped with passive anti-roll bars, however they may be not sufficient to overcome critical situations. This thesis focuses on the active anti-roll bar system, which is the most common method used to improve roll stability of heavy vehicles.The thesis research work is divided into three main parts. In the first part, an integrated model is proposed with four electronic servo-valve hydraulic actuators mounted in a linear yaw-roll model of a single unit heavy vehicle. In the second part, the active anti-roll bar system uses two control approaches in the LTI framework: LQR, Hinfty. In the third part, the grid-based LPV approach is used to synthesize the Hinfty/LPV active anti-roll bar controller with parameter dependant weighting functions, by using LPVTools.The simulation results, in the frequency and time domains, as well as the validation by using the TruckSim simulation software, show that the active anti-roll bar control is a realistic and efficient solution which drastically improves roll stability of a single unit heavy vehicle, compared to the passive anti-roll bar.

La thèse étudie les implications des anticipations des agents privés formées avec l’apprentissage adaptatif pour la politique monétaire optimale dans des modèles Nouveau Keynésien. Les résultats obtenus sont comparés avec la littérature adoptant l’hypothèse d’anticipations rationnelles. La banque centrale fait l’arbitrage intertemporel introduit par les anticipations à apprentissage adaptatif entre stabiliser l’inflation maintenant et dans le futur. La distorsion générée par ces anticipations rend la politique monétaire davantage agressive, même si la présence du marché financier diminue légèrement l’agressivité de la politique monétaire. Il est optimal pour le gouvernement de choisir un banquier central libéral, ce résultat peut être mitigé par un contrat d’inflation linéaire. Un autre résultat est qu’il est optimal pour la banque centrale d’être moins indépendante en termes d’instruments par rapport aux anticipations rationnelles. La possibilité de contrôle robuste de la politique monétaire est limitée par l’apprentissage adaptatif en économie fermée, et encore plus limitée en économie ouverte. Mes travaux de recherche donnent des recommandations nouvelles. / The dissertation studies the implications of private agents expectations formed with adaptive learning for the optimal monetary policy using New Keynesian models. The obtained results are compared with the literature adopting the hypothesis of rational expectations. The central bank makes the intertemporal trade-off between stabilizing current inflation or the future one that is introduced by adaptive learning expectations. The distortion which is introduced by this latter makes the monetary policy more aggressive, even if the presence of financial market slightly reduces the aggressiveness of the monetary policy. It is optimal for the government to choose a liberal central banker, but this result may be mitigated by adopting a linear inflation contract. Another result is that it is optimal for the central bank to be less instrument-independent compared to rational expectations. The possibility of robust control for monetary policy is limited by adaptive learning in a closed economy, and even more limited in an open economy. My research works give new recommendations for policy making.

Les robots flexibles sont de plus en plus utilisés dans les applications pratiques. Ces robots sont caractérisés par une conception mécanique légère, réduisant ainsi leur encombrement, leur consommation d’énergie et améliorant leur sécurité. Cependant, la présence de vibrations transitoires rend difficile un contrôle précis de la trajectoire de ces systèmes. Cette thèse est précisément consacrée à l’asservissement en position des manipulateurs flexibles dans les espaces articulaire et opérationnel. Des méthodes de commande avancées, basées sur des outils de la commande robuste et de l’optimisation convexe, ont été proposées. Ces méthodes font en particulier appel à la théorie des systèmes linéaires à paramètres variants (LPV) et aux inégalités matricielles linéaires (LMI). En comparaison avec des lois de commande non-linéaires disponibles dans la littérature, les lois de commande LPV proposées permettent de considérerdes contraintes de performance et de robustesse de manière simple et systématique. L’accent est porté dans notre travail sur la gestion appropriée de la dépendance paramétrique du modèle LPV, en particulier les dépendances polynomiale et rationnelle. Des simulations numériques effectuées dans des conditions réalistes, ont permis d’observer une meilleure robustesse de la commande LPV par rapport à la commande non-linéaire par inversion de modèle face aux bruits de mesure, aux excitations de haute fréquence et aux incertitudes de modèle. / Flexible robots are becoming more and more common in practical applications. This type of robots is characterized by the use of lightweight materials, which allows reducing their size, their power consumption and improves their safety. However, an accurate trajectory tracking of these systems is difficult to achieve because of the transient vibrations they undergo. This PhD thesis work is particularly devoted to the position control of flexible robotic manipulators at the joint and end-effector levels. Advanced control methods, based on some tools of the robust control theory and convex optimization, have been proposed. These methods are based on the theory of Linear Parameter Varying (LPV) systems and Linear Matrix Inequalities (LMI). Compared to some nonlinear control laws available in the literature that involve model inversion, theproposed LPV control laws make it possible to consider performance and robustness constraints in a simple and systematic manner. Our work particularly emphasizes on the appropriate management of the parametric dependence of the LPV model, especially the polynomial and rational dependences. Numerical simulations carried out in realistic operating conditions have shown a better robustness of the LPV control compared to the inversion-based nonlinear control withrespect to measurement noise, high frequency inputs and model uncertainties.

Ces travaux de recherche portent sur la synthèse systématique et l’optimisation de correcteurs numériques à structure polynomiale RST par approches métaheuristiques. Les problèmes classiques de placement de pôles et de calibrage des fonctions de sensibilité de la boucle fermée RST sont formulés sous forme de problèmes d’optimisation multi-objectif sous contraintes pour lequel des algorithmes métaheuristiques de type NSGA-II, MODE, MOPSO et epsilon-MOPSO sont proposés et adaptés. Deux formulations du problème de synthèse RST ont été proposées. La première approche, formulée dans le domaine temporel, consiste à minimiser des indices de performance, de type ISE et MO, issus de la théorie de la commande optimale et liés essentiellement à la réponse indicielle du système corrigé. Ces critères sont optimisés sous des contraintes non analytiques définis par des gabarits temporels sur la dynamique de la boucle fermée. Dans la deuxième approche de synthèse RST, une formulation dans le domaine fréquentiel est retenue. La stratégie proposée consiste à définir et calculer une fonction de sensibilité de sortie désirée en satisfaisant des contraintes de robustesse de H∞. L’utilisation de parties fixes dans la fonction de sensibilité de sortie désirée assurera un placement partiel des pôles de la boucle fermée RST. L’inverse d’une telle fonction de sensibilité désirée définira le filtre de pondération H∞ associé. Un intérêt particulier est porté à l’approche d’optimisation par essaim particulière PSO pour la résolution des problèmes multi-objectif de commande reformulés. Un algorithme MOPSO à grille adaptative est proposé et puis perfectionné à base des concepts de l’epsilon-dominance. L’algorithme epsilon-MOPSO obtenu a montré, par comparaison avec les algorithmes MOPSO, NSGA-II et MODE, des performances supérieures en termes de diversité des solutions de Pareto et de rapidité en temps de convergence. Des métriques de type distance générationnelle, taux d’erreurs et espacement sont toutefois considérées pour l’analyse statistique des résultats de mise en œuvre obtenus. Une application à la commande en vitesse variable d’un moteur électrique DC est effectuée, également pour la commande en position d’un système de transmission flexible à charges variables. La mise en œuvre par simulations numériques sur les procédés considérés est également présentée dans le but de montrer la validité et l’efficacité de l’approche de commande optimale RST proposée / This research focuses on the systematic synthesis and optimization of digital RST structure based controllers thanks to global metaheuristics approaches. The classic and hard problems of closed-loop poles placement and sensitivity functions shaping of RST control are well formulated as constrained multi-objective problems to be solved with proposed metaheuristics algorithms NSGA-II, MODE, MOPSO and especially epsilon-MOPSO. Two formulations of the metaheuristics-tuned RST problem have been proposed. The first one, which is given in the time domain, deals with the minimization of several performance criteria like the Integral Square Error (ISE) and the Maximum Overshoot (MO) indices. These optimal criteria, related primarily to the step response of the controlled plant, are optimized under non-analytical constraints defined by temporal templates on the closed-loop dynamics. In the second approach, a formulation in the frequency domain is retained. The proposed strategy aims to optimize a desired output sensitivity function satisfying H∞ robustness constraints. The use of a suitable fixed part of the optimized output sensitivity function will provide partial pole placement of the closed-loop dynamics of the digital RST controller. The opposite of such desired sensitivity function will define the associated H∞ weighting filter. The Multi-Objective Particle Swarm Optimization (MOPSO) technique is particularly retained for the resolution of all formulated multi-objective RST control problems. An adaptive grid based MOPSO algorithm is firstly proposed and then improved based on the epsilon-dominance concepts. Such proposed epsilon-MOPSO algorithm, with a good diversity of the provided Pareto solutions and fast convergence time, showed a remarkable superiority compared to the standard MOPSO, NSGA-II and MODE algorithms. Performance metrics, such as generational distance, error rate and spacing, are presented for the statistical analysis of the achieved multi-optimization results. An application to the variable speed RST control of an electrical DC drive is performed, also for the RST position control of a flexible transmission plant with varying loads. Demonstrative simulations and comparisons are carried out in order to show the validity and the effectiveness of the proposed metaheuristics-based tuned RST control approach, which is formulated in the multi-objective optimization framework

Les applications du calcul fractionnaire en automatique se sont considérablement développées ces dernières années, surtout en commande robuste. Ce mémoire est une contribution à la commande robuste des systèmes d'ordre entier à l'aide d'un correcteur PID d'ordre fractionnaire.Le conventionnel régulateur PID, unanimement apprécié pour le contrôle des processus industriels, a été adapté au cas fractionnaire sous la forme PInDf grâce à l'introduction d'un modèle de référence d'ordre non entier, réputé pour sa robustesse vis-à-vis des variations du gain statique.Cette nouvelle structure a été étendue aux systèmes à retard sous la forme d'un Prédicteur de SMITH fractionnaire. Dans leur forme standard, ces correcteurs sont adaptés à la commande des systèmes du premier et du second ordre, avec ou sans retard pur.Pour des systèmes plus complexes, deux méthodologies de synthèse du correcteur ont été proposées, grâce à la méthode des moments et à l'approche retour de sortie.Pour les systèmes dont le modèle est obtenu à partir d'une identification, la boucle fermée doit en outre être robuste aux erreurs d'estimation. Un modèle pire-cas, déduit de la matrice de covariance de l'estimateur et des domaines d'incertitudes fréquentielles, a été proposé pour la synthèse du correcteur.Les différentes simulations numériques montrent l'efficacité de cette méthodologie pour l'obtention d'une boucle fermée robuste aux variations du gain statique et aux incertitudes d'identification. / The application of fractional calculus in automatic control have received much attention these last years, mainly in robust control. This PhD dissertation is a contribution to the control of integer order systems using a fractional order PID controller.The classical PID, well known for its applications to industrial plants, has been adapted to the fractional case as a PInDf controller, thanks to a fractional order reference model, characterized by its robustness to static gain variations.This new controller has been generalized to time delay systems as a fractional SMITH Predictor. In standard case, these controllers are adapted to first and second order systems, with or without a time delay. For more complex systems, two design methodologies have been proposed, based on the method of moments and on output feedback approach.For systems whose model is obtained by an identification procedure, the closed loop has to be robust to estimation errors. So, a worst-case model, derived from the covariance matrix of the estimator and the frequency uncertainty domains, has been proposed for the design of the controller.The different numerical simulations demonstrate that this methodology is able to provide robustness to static gain variations and to identification uncertainties.

Submitted by JOSÉ CARLOS UGAZ PEÑA null (josecarlos84@gmail.com) on 2016-06-20T13:29:30Z No. of bitstreams: 1 Tese_JoseCarlosPena_PPGEE.pdf: 9650212 bytes, checksum: 77b105d009c0b473cbc424e681ebe9a5 (MD5) / Approved for entry into archive by Ana Paula Grisoto (grisotoana@reitoria.unesp.br) on 2016-06-22T13:10:00Z (GMT) No. of bitstreams: 1 pena_jcu_dr_ilha.pdf: 9650212 bytes, checksum: 77b105d009c0b473cbc424e681ebe9a5 (MD5) / Made available in DSpace on 2016-06-22T13:10:00Z (GMT). No. of bitstreams: 1 pena_jcu_dr_ilha.pdf: 9650212 bytes, checksum: 77b105d009c0b473cbc424e681ebe9a5 (MD5) Previous issue date: 2016-06-02 / Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) / Os inversores fonte de tensão com filtro de saída LCL (VSI+LCL) são amplamente utilizados em sistemas de geração distribuída. Nestas aplicações o sistema é controlado como uma fonte de corrente, no entanto, há a possibilidade de controlar o conjunto como uma fonte de tensão. Desta forma, a mencionada configuração pode ser utilizada em geração distribuída no ambiente de microrredes onde os sistemas, monofásicos ou trifásicos, devem operar conectados à rede de distribuição elétrica com controle das potências injetadas (ativa e reativa) e serem capazes de, em ausência da rede, passar a operar no modo autônomo. Ainda, após o restabelecimento da rede, o controle deve levar o sistema a operar novamente no modo conectado. Sendo as transições realizadas sem transientes que possam danificar qualquer componente do sistema. O filtro LCL, de terceira ordem, caracteriza um comportamento ressonante que pode comprometer a estabilidade do sistema. Para resolver esta situação, diversas técnicas ativas e passivas são aplicadas. Para aplicações de baixa potência, preferem-se as técnicas passivas de amortecimento devido a sua simplicidade e baixo custo, porém estas implicam em perdas adicionais. Já as técnicas ativas de amortecimento, consideram procedimentos de controle para atenuar a ressonância, e, portanto, não adicionam perdas, porém, sua realização requer da realimentação de variáveis adicionais elevando assim o custo do sistema. Todavia, mesmo que aplicáveis a ambos os modos de operação, as técnicas de amortecimento disponíveis na literatura consideram apenas um modo de operação. O presente trabalho de doutorado explora a possibilidade de controlar sistemas VSI+LCL, monofásicos e trifásicos, em ambos os modos de operação, com atenção a objetivos específicos em cada modo e transições suaves entre estes. Assim, são apresentadas duas estratégias de controle. A primeira estratégia considera o amortecimento da ressonância por técnicas passivas, mediante a utilização de um ramo de amortecimento de segunda ordem, projetado para garantir o efeito desejado em ambos os modos de operação e simplificar a dinâmica do sistema a fim de facilitar o projeto dos controladores, abordagem não utilizada nos métodos conhecidos na literatura. Logo, o sistema amortecido é controlado em uma configuração de duas malhas, controlando a corrente injetada mediante a tensão no capacitor. A segunda estratégia proposta considera a utilização de controladores por realimentação de estados em tempo discreto, sintetizados mediante desigualdades matriciais lineares, para simultaneamente, realizar ativamente o amortecimento da ressonância e atender os objetivos de controle em ambos os modos de operação. Ambas as estratégias propostas consideram controladores ressonantes com o objetivo de rastrear sinais senoidais de referência com erro nulo e suprimir componentes harmônicos de baixa ordem na corrente de saída. Ainda, os controladores são projetados considerando a necessidade de garantir a estabilidade robusta do sistema, isto é, frente a perturbações externas (tais como variações na carga local, oscilações na tensão do barramento CC ou distúrbios na rede) e às variações em parâmetros do sistema, como a indutância de rede. As propostas são apresentadas em detalhe, incluindo os procedimentos de projeto assim como critérios para a geração e coordenação dos sinais de controle e referência. As estratégias propostas são avaliadas experimentalmente sendo os resultados obtidos discutidos e analisados considerando-se as respectivas normas para os casos de operação conectada e ilhada. / The voltage source inverter utilization with LCL filters (VSI+LCL) is extended in Distributed Electrical Energy Systems. In these applications, the system is controlled as a current source, however, it can also be controlled as a voltage source. Hence, this configuration is suitable for microgrids environment. In this scenario, the system should operate connected to utility grid with control of the supplied power (active or reactive) and also be capable, in case of grid absence, to operate in islanded mode. Then, if the grid is reestablished, system should be reconnected to grid. Moreover, these transitions should be smooth, with no hazardous transients. The third order filter leads to a resonant behavior that can compromise the system stability. In order to overcome this limitation, passive and active damping methods are used. In low power applications, passive damping methods are preferred due to their simplicity and low. Nevertheless, these methods lead to additional losses. On the other hand, active damping methods consider the feedback of additional variables in order to damp the resonance in closed loop, with no additional losses. This implies additional sensors, thus increasing the overall cost. Despite their effectiveness to damp the resonance in both autonomous and grid connected applications, the most of the damping methods are usually designed only for a specific operation mode. This work explores the possibility to control VSI+LCL systems, single and three-phase, in both operation modes, attending to specific goals in each one, and with smooth transitions between them. For that purpose, two control strategies are proposed. The first one considers passive damping methods, by using a second order damping branch which is designed in order to guarantee the desired effect in both operation modes, thus simplifying the system dynamics in order to ease the control. This approach is not known in the literature. Then, the damped system is controlled in a two loop strategy, where the output current is controlled by means of the capacitor voltage. The second strategy considers the utilization of discrete time state-feedback controllers, synthesized by Linear Matrix Inequalities, in order to simultaneously achieve the active damping and the control goals for both operation modes. The proposed strategies use resonant controllers in order to achieve the tracking of sinusoidal references and to suppress low order harmonics in the output current. Moreover, controllers are designed to achieve robust stability of the system, thus, even in front of external disturbances (such as local load variations, DC bus oscillations or grid disturbances) and variation on system parameters, such as the grid inductance. The two introduced strategies are detailed including the design procedure and the criterion to generate and coordinate the reference and control signals. The two proposed strategies were experimentally verified. The results were analyzed and compared to the requirements imposed by the related standards for both modes of operation. / CNPq: 141757/2012-4