Un système de puissance électrique est un réseau complexe de composants électriques utilisés pour fournir, transmettre et utiliser l'énergie électrique. Son objectif final est d'offrir un service fiable, sécurisé et ininterrompu à l'utilisateur final, cela signifie, tension constante et fréquence constante en tout temps. Aujourd'hui, la tendance de la production d'électricité est vers un réseau interconnecté de lignes de transmission reliant la génération et les charges dans des grands systèmes intégrés. En fait, un réseau de système de puissance est considéré comme la machine la plus complexe et plus jamais construite par l'homme car elle peut s'étendre sur tout un continent. Pour cette raison, l'amélioration de la stabilité transitoire des réseaux électriques est d'une grande importance dans la société humaine, car si la stabilité est perdue, le collapse de la puissance peut se produire dans une grande zone peuplée et de graves dommages seront portées à l'économie régionale et les conforts des consommateurs. Par conséquent, compte tenu de tous les problèmes présentés avant, ce travail de recherche aborde la stabilisation transitoire des systèmes de puissance multi-machines soumises à des perturbations du réseau à partir de deux approches: la centralisation, qui considère aucune limitation dans l'échange d'informations d'un réseau donné, et d'autre part, la décentralisation, qui suppose l'échange d'informations n'est pas disponible. À cette fin, d'abord, nous introduisons une nouvelle théorie de commande pour stabiliser globalement systèmes triangulaires non linéarisables globalement en utilisant une commande de rétroaction d'état dynamique non linéaire, qui diffère de backstepping puisque la forme stricte de rétroaction n'est plus nécessaire. Ensuite, sur la base de ces nouvelles idées, le problème de stabilisation transitoire des systèmes de puissance est résolu d'un point de vue centralisé, en assurant la stabilité asymptotique globale du point de fonctionnement, dans certaines conditions sur les paramètres physiques du système. Postérieurement, en utilisant uniquement les mesures locales disponibles avec la technologie existante, le contrôleur central précédent est transformé en un décentralisé, à condition que la dérivée de la puissance active à chaque générateur peut être appropriement estimée. La performance des deux contrôleurs est testée par des simulations numériques envisagent plusieurs scénarios de défaut en utilisant le système de 10 machines de Nouvelle-Angleterre. Contrairement aux solutions non linéaires ci-dessus, nous proposons une méthodologie basée sur observateur pour la stabilisation décentralisée des systèmes linéaires invariants dans le temps. L'originalité de ce travail repose sur le fait que chaque contrôleur local est fourni avec des mesures locales disponibles, il met en œuvre un observateur pour reconstruire l'état des autres sous-systèmes et utilise de manière équivalente ces estimations dans la loi de commande. Les observateurs sont conçus en suivant les principes de l'immersion et l'invariance. De plus, la classe des systèmes est identifiée par une solution d'une inégalité matricielle linéaire, à partir de laquelle on obtient les gains d'observateurs. / An electric power system (EPS) is a complex network of electrical components used to supply, transmit and use electric power. Its final goal is to provide reliable, secure and uninterrupted service to the end-user, this means, constant voltage and frequency at all time. Nowadays, the trend in electric power production is toward an interconnected network of transmission lines linking generators and loads into large integrated systems. Actually, a power system network is considered the most complex and bigger machine ever built by man since it can span an entire continent. For this reason, improving power system transient stability is of great significance in human society, since if the stability is lost, power collapse may occur in a large populated area and serious damages will be brought to a regional economy and the consumer's comforts. Therefore, considering all issues presented before, this research work tackles the transient stabilization of a multi-machine EPS subject to network disturbances from two approaches: centralization which considers no limitation in information exchange at any point of a given network, and on the other hand, decentralization which assumes the information exchange is not available. To this end, first we introduce a novel control theory to globally stabilize non-globally linearizable triangular systems employing a nonlinear dynamic state-feedback controller, which differs from standard backstepping since the strict-feedback form is no longer required. Then, based on these new ideas, the transient stabilization problem of EPS is solved from a centralized point of view ensuring, under some conditions on the physical parameters of the system, global asymptotic stability of the operating point. Subsequently, using only local measurements available with existing technology, the previous central controller is transformed into a truly decentralized one, provided that the derivative of the active power at each generator can be suitable estimated. Performance of both controllers is tested via numerical simulations considering several fault scenarios using the 10-machine New England benchmark. In contrast to the nonlinear solutions above, we offer an observer--based methodology for decentralized stabilization of large--scale linear time--invariant systems. The originality of this work relies on the fact that each local controller is provided with available local measurements, it implements a deterministic observer to reconstruct the state of the other subsystems and uses in a certainty--equivalent way these estimates in the control law. The observers are designed following the principles of immersion and invariance. Furthermore, the class of systems to which the design is applicable is identified via a linear matrix inequality solution, from which the observer gains are obtained.

[pt] No presente trabalho estudou-se numericamente a instabilidade das ondas na interface do escoamento estratificado de água e óleo em um canal plano. Esse padrão de escoamento, associado ao estágio inicial da formação de golfada, é comum em aplicações industriais, de áreas como produção de petróleo, nuclear, química e muitas outras. Através da introdução de perturbações controladas na interface do escoamento estratificado, analisou-se a evolução dessas perturbações à luz das teorias de estabilidade hidrodinâmica. Os experimentos numéricos foram realizados utilizando o método de Volume of Fluid (VOF) do simulador comercial ANSYS Fluent versão 15.0. Analisou-se o comportamento do escoamento em dois regimes distintos com relação a amplitude das ondas interfaciais. No primeiro regime, empregaram-se ondas pequenas o suficiente para que efeitos não lineares fossem desprezíveis. Os resultados obtidos apresentaram boa concordância com as previsões fornecidas por um solver das equações de Orr-Sommerfeld, para escoamento bifásico estratificado em um canal, indicando que a ferramenta numérica foi capaz de reproduzir o comportamento das ondas interfaciais. Mostrou-se que existe uma faixa de amplitudes, em torno de 0,2 porcento da altura do canal, a partir da qual os efeitos não lineares se tornam relevantes. No regime não linear foram avaliados diferentes cenários de interação não linear entre ondas, os quais geralmente são associados a transição do regime do escoamento estratificado para golfadas. Identificou-se o cenário mais relevante, analisando-se,a eficiência de cada uma dessas interações isoladamente. Observou-se que interações não lineares entre ondas de comprimento parecido são as que crescem mais rapidamente. Esse mecanismo parece ser dominante também na presença de um grande número de ondas, como é o caso esperado em um evento natural. Utilizou-se um modelo fracamente não linear, baseado nas equações de Stuart-Landau, para modelar o comportamento das ondas no escoamento, obtendo-se excelente concordância com os resultados das simulações. Isso é interessante do ponto de vista prático, pois sugere que modelos não lineares simples, como é o caso da equação de Stuart-Landau, podem ser implementados para melhorar as ferramentas utilizadas para prever mudanças de regime em escoamentos bifásicos. / [en] In the present work, the instability of waves at the interface of the stratified flow of water and oil in a plane channel was numerically studied. This flow pattern, which is associated with the initial stages of slug formation, is common in industrial application in areas such as oil production, nuclear, chemical and many others. Through the introduction of controlled perturbations at the interface of a stratified flow, the evolution of the perturbations was analyzed based on hydrodynamics stability theories. Numerical experiments were performed using the Volume of Fluid (VOF) method of the ANSYS Fluent release 15.0 commercial simulator. The behavior of the flow in two different regimes with respect to the amplitude of the interfacial waves was analyzed. In the first regime, small enough waves were employed so that non-linear effects were negligible. The results obtained presented good agreement with the predictions provided by a solver of the Orr-Sommerfeld equations for stratified two-phase flow in a channel, indicating that the numerical tool was able to reproduce the behavior of the interfacial waves. It was shown that there is a range of amplitudes, around 0.2 per cent of the channel height, above which the non-linear effects become relevant. In the nonlinear regime, different scenarios of nonlinear interaction between waves, which are usually associated with transition from stratified flow pattern to slug flow, were evaluated. The most relevant scenario was identified, based on the efficiency of each independent interaction. It was observed that non-linear interactions between waves of similar length present the fastest growth. This mechanism seems to be dominant also in the presence of a large number of waves, as present in natural events. A weakly nonlinear model, based on the Stuart- Landau equations, was employed to model the wave behavior in the flow, obtaining an excellent agreement with the results of the simulations. This is interesting from a practical point of view, since it suggests that simple nonlinear models, such as the Stuart-Landau equation, can be implemented to improve the tools used to predict regime changes in two-phase flows.