Significant stray magnetic field exists around the coils when charging the electric vehicles (EVs) with inductive power transfer (IPT), owning to the large air gap between the transmitter and receiver. The methods for field attenuation usually introduce extra losses and reduce the efficiency. This study focuses on the synergetic attenuation of stray magnetic field which is optimized simultaneously with the efficiency. The optimization is realized with Pareto front. In this dissertation, three methods are discussed for the field attenuation. The first method is to tune the physical parameters of the winding, such as the inner radii, outer radii, distribution of the turns, and types of the litz wires. The second method is to add metal shields around the IPT coils, in which litz wires are used as shields to reduce the shielding losses. The third method is to control the phases of winding currents, which avoids increasing the size and weight of the IPT coils. To attenuate the stray magnetic field by tuning the physical parameters, the conventional method is to sweep all the physical parameters in finite-element simulation. This takes thousands of simulations to derive the Pareto front, and it's especially time-consuming for three-dimensional simulations. This dissertation demonstrates a faster method to derive the Pareto front. The windings are replaced by the lumped loops. As long as the number of turns for each loop is known, the efficiency and magnetic field are calculated directly from the permeance matrices and current-to-field matrices. The sweep of physical parameters in finite-element simulation is replaced by the sweep of the turns numbers for the lumped loops in calculation. Only tens of simulations are required in the entire procedure, which are used to derive the matrices. An exemplary set of coils was built and tested. The efficiency from the matrix calculation is the same as the experimental measurement. The difference for stray magnetic field is less than 12.5%. Metal shields attenuate the stray magnetic field effectively, but generates significant losses owning to the uneven distribution of shield currents. This dissertation uses litz wires to replace the conventional plate shield or ring shield. Skin effect is eliminated so the shield currents are uniformly distributed and the losses are reduced. The litz shields are categorized to two types: shorted litz shield and driven litz shield. Circuit models are derived to analyze their behaviors. The concept of lumped-loop model is applied to derive the Pareto front of efficiency versus stray magnetic field for the coils with litz shield. In an exemplary IPT system, coils without metal shield and with metal shields are optimized for the same efficiency. Both the simulation and experimental measurement verify that the shorted litz shield has the best performance. The stray magnetic field is attenuated by 65% compared to the coils without shield. This dissertation also introduces the method to attenuate the stray magnetic field by controlling the phases of winding currents. The magnetic field around the coils is decomposed to the component in the axial direction and the component in the radial direction. The axial component decreases with smaller phase difference between windings' currents, while the radial component exhibits the opposite property. Because the axial component is dominant around the IPT coils, decreasing the phase difference is preferred. The dual-side-controlled converter is applied for the circuit realization. Bridges with active switches are used for both the inverter on the transmitter side and the rectifier on the receiver side. The effectiveness of this method was verified both in simulation and experiment. Compared to the conventional series-series IPT with 90° phase difference between winding currents, stray magnetic field was attenuated by up to 30% and 40% when the phase differences of winding currents are 50° and 40°, respectively. Furthermore, an analytical method is investigated to calculate the proximity-effect resistance of the planar coils with ferrite plate. The objective of this method is to work together with the fast optimization which uses the lumped-loop model. The existence of the ferrite plate complicates the calculation of the magnetic field across each turn which is critical to derive the proximity-effect resistance. In this dissertation, the ferrite plate is replaced by the mirrored turns according to the method of image. The magnetic fields are then obtained from Ampere's Law and Biot-Savart Law. Up to 200 kHz, the difference of the proximity-effect resistance is less than 15% between calculation and measurement. / Ph. D.

[ES] El aumento en el uso de combustibles fósiles para la generación de energía ha contribuido significativamente a la crisis del calentamiento global. Diferentes lugares alejados de la infraestructura eléctrica emplean generadores a base de gasolina que aumentan la contaminación ambiental. En este contexto, la introducción masiva de microrredes en la sociedad ha traído oportunidades para la generación de energía de forma distribuida, beneficiando a personas en todo el mundo. Por ejemplo, las microrredes pueden brindar electricidad a poblaciones vulnerables que viven en áreas remotas con acceso limitado a infraestructuras de transmisión y distribución. Además, las microrredes promueven el uso de recursos renovables, reduciendo el impacto ambiental en comparación con los métodos tradicionales de generación de electricidad, como las plantas de energía térmica o las instalaciones nucleares. Además, las microrredes permiten la generación de electricidad a pequeña escala, lo que permite que las familias logren la independencia energética y vendan el exceso de energía a la compañía eléctrica local. Cualquier inversor en una microrred necesita un algoritmo de control para realizar una regulación en bucle cerrado. En este contexto, el control por modos deslizantes de segundo orden es una estrategia de control robusta que ha ganado atención en las aplicaciones de inversores de microrredes. Mediante el uso de este enfoque, el inversor puede lograr un control preciso y rápido, incluso en presencia de incertidumbres y perturbaciones. El uso de estrategias de control robustas mejora la estabilidad y el rendimiento general del sistema de microrredes, asegurando una gestión de energía óptima. El proceso de ajuste es esencial para los algoritmos de control en bucle cerrado, ya que modifica la respuesta del controlador para alcanzar los objetivos de control. La optimización por enjambre de partículas (PSO por sus siglas en inglés) es un eficiente algoritmo de optimización empleado en controladores en lazo cerrado que puede resolver de manera efectiva problemas multi-objetivo formulados en una sola función de costo. Los parámetros de control del inversor de la microrred pueden ser optimizados mediante la utilización de PSO para lograr los objetivos deseados, ajustando de manera eficiente una estrategia de control. Para controladores por modos deslizantes, algunas estrategias de ajuste se basan en técnicas heurísticas. La función de costo única resuelve varios problemas en una microrred, pero existen dificultades cuando diferentes objetivos en un proceso no pueden ser mejorados simultáneamente debido a su relación conflictiva. Estrategias como Algoritmos Genéticos Multi-Objetivo (MOGA por sus siglas en inglés), Evolución Diferencial Multi-Objetivo (MODE por sus siglas en inglés) y Algoritmo Artificial de Ovejas Multi-Objetivo (MOASA por sus siglas en inglés), han demostrado su capacidad para mejorar el rendimiento del inversor mediante la optimización de objetivos conflictivos. Estos algoritmos pueden equilibrar de manera efectiva objetivos como la reducción del tiempo de respuesta y la minimización del sobreimpulso en la señal de salida del inversor. En consecuencia, el rendimiento general y la eficiencia de los inversores de la microrred pueden mejorar. La integración de algoritmos de control multi-objetivo en los inversores de la microrred tiene un gran potencial para abordar los desafíos de gestión de energía y optimizar el rendimiento. Los inversores de la microrred pueden lograr una mayor estabilidad, eficiencia y confiabilidad utilizando técnicas como el control por modos deslizantes de segundo orden y algoritmos de optimización como PSO, MOGA, MODE y MOASA. Al adoptar estos enfoques, se presenta una nueva metodología para un futuro energético más sostenible y resiliente, al tiempo que se mitigan los efectos adversos del calentamiento global causado por el consumo de combustibles fósiles en la generación convencional de energía. / [CA] L'augment en l'ús de combustibles fòssils per a la generació d'energia ha contribuït significativament a la crisi de l'escalfament global. Diferents llocs allunyats de la infraestructura elèctrica empleen generadors a base de gasolina que augmenten la contaminació ambiental. En aquest context, la introducció massiva de microxarxes a la societat ha comportat oportunitats per a la generació d'energia de forma distribuïda, beneficiant persones arreu del món. Per exemple, les microxarxes poden proporcionar electricitat a poblacions vulnerables que viuen en àrees remotes amb accés limitat a infraestructures de transmissió i distribució. A més, les microxarxes promouen l'ús de recursos renovables, reduint l'impacte ambiental en comparació amb els mètodes tradicionals de generació d'electricitat, com les plantes d'energia tèrmica o les instal·lacions nuclears. A més a més, les microxarxes permeten la generació d'electricitat a petita escala, la qual cosa permet que les famílies aconsegueixin la independència energètica i venguen l'excedent d'energia a la companyia elèctrica local. Qualsevol inversor en una microxarxa necessita un algoritme de control per a realitzar una regulació en bucle tancat. En aquest context, el control per modes lliscants de segon ordre és una estratègia de control robusta que ha guanyat atenció en les aplicacions d'inversors de microxarxes. Mitjançant l'ús d'aquest enfocament, l'inversor pot aconseguir un control precís i ràpid, fins i tot en presència d'incerteses i pertorbacions. L'ús d'estratègies de control robustes millora l'estabilitat i el rendiment general del sistema de microxarxes, assegurant una gestió d'energia òptima. El procés d'ajust és essencial pels algoritmes de control en bucle tancat, ja que modifica la resposta del controlador per a aconseguir els objectius de control. L'optimització per enjambre de partícules (PSO per les seues sigles en anglés) és un eficient algoritme d'optimització emprat en controladors en bucle tancat que pot resoldre de manera efectiva problemes multi-objectiu formulats en una sola funció de cost. Els paràmetres de control de l'inversor de la microxarxa poden ser optimitzats mitjançant l'utilització de PSO per a aconseguir els objectius desitjats, ajustant de manera eficient una estratègia de control. Per a controladors per modes lliscants, algunes estratègies d'ajust es basen en tècniques heurístiques. La funció de cost única resol diversos problemes en una microxarxa, però existeixen dificultats quan diferents objectius en un procés no poden ser millorats simultàniament a causa de la seua relació conflictiva. Estratègies com Algorismes Genètics Multi-Objectiu (MOGA per les seues sigles en anglés), Evolució Diferencial Multi-Objectiu (MODE per les seues sigles en anglés) i Algorisme Artificial de Xais Multi-Objectiu (MOASA per les seues sigles en anglés), han demostrat la seua capacitat per a millorar el rendiment de l'inversor mitjançant l'optimització d'objectius conflictius. Aquests algorismes poden equilibrar de manera efectiva objectius com la reducció del temps de resposta i la minimització del sobreguiny a la senyal de sortida de l'inversor. En conseqüència, el rendiment general i l'eficiència dels inversors de la microxarxa poden millorar. La integració d'algorismes de control multi-objectiu en els inversors de la microxarxa té un gran potencial per a abordar els desafiaments de gestió d'energia i optimitzar el rendiment. Els inversors de la microxarxa poden aconseguir una major estabilitat, eficiència i fiabilitat utilitzant tècniques com el control per modes lliscants de segon ordre i algorismes d'optimització com PSO, MOGA, MODE i MOASA. En adoptar aquests enfocaments, es presenta una nova metodologia per a un futur energètic més sostenible i resilient, al mateix temps que es mitiguen els efectes adversos de l'escalfament global causat pel consum de combustibles fòssils en la generació convencional d'energia. / [EN] The increase in fossil fuel usage for power generation has significantly contributed to the global warming crisis. Various remote areas, detached from electrical infrastructure, rely on gasoline-based generators that escalate environmental pollution. In this context, the widespread implementation of microgrids in society has brought forth opportunities for distributed energy generation, benefiting people worldwide. For instance, microgrids can provide electricity to vulnerable populations in remote areas with limited access to transmission and distribution infrastructures. Furthermore, these microgrids advocate for using renewable resources, diminishing environmental impact compared to traditional methods such as thermal power plants or nuclear facilities. Additionally, microgrids enable small-scale electricity generation, empowering families to achieve energy independence and sell surplus energy to local power companies. Any investor in a microgrid requires a closed-loop control algorithm. In this realm, the second-order sliding mode control is a robust strategy garnering attention in microgrid inverter applications. Through this approach, the inverter can achieve precise and rapid control despite uncertainties and disturbances. Using robust control strategies enhances microgrid systems' stability and overall performance, ensuring optimal energy management. Adjustment processes are pivotal for closed-loop control algorithms, modifying the controller's response to meet control objectives. Particle Swarm Optimization (PSO) is an efficient optimization algorithm employed in closed-loop controllers that can effectively solve multi-objective problems formulated in a single cost function. Control parameters of the microgrid inverter can be optimized using PSO to attain desired objectives, efficiently fine-tuning a control strategy. For sliding mode controllers, some adjustment strategies rely on heuristic techniques. While a single cost function resolves various issues within a microgrid, difficulties arise when different objectives in a process cannot be simultaneously improved due to conflicting relationships. Strategies like Multi-Objective Genetic Algorithms (MOGA), Multi-Objective Differential Evolution (MODE), and Multi-Objective Artificial Sheep Algorithm (MOASA) have proven their ability to enhance inverter performance by optimizing conflicting objectives. These algorithms effectively balance objectives like reducing response time and minimizing overshoot in the inverter's output signal. Consequently, the overall performance and efficiency of microgrid inverters can be enhanced. Integrating multi-objective control algorithms into microgrid inverters holds significant potential in addressing energy management challenges and optimizing performance. Microgrid inverters can achieve greater stability, efficiency, and reliability by utilizing second-order sliding mode control and optimization algorithms like PSO, MOGA, MODE, and MOASA. By embracing these approaches, a new methodology emerges for a more sustainable and resilient energy future while mitigating the adverse effects of global warming caused by conventional fossil fuel consumption in power generation. / Gonzales Zurita, ÓO. (2024). Multi-objective Control on Inverter-Based Microgrids [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/203120

Doctor of Philosophy / Department of Electrical and Computer Engineering / Sanjoy Das / Stochastic algorithms are widely used in various modeling and optimization problems. Evolutionary algorithms are one class of population-based stochastic approaches that are inspired from Darwinian evolutionary theory. A population of candidate solutions is initialized at the first generation of the algorithm. Two variation operators, crossover and mutation, that mimic the real world evolutionary process, are applied on the population to produce new solutions from old ones. Selection based on the concept of survival of the fittest is used to preserve parent solutions for next generation. Examples of such algorithms include genetic algorithm (GA) and genetic programming (GP). Nevertheless, other stochastic algorithms may be inspired from animals’ behavior such as particle swarm optimization (PSO), which imitates the cooperation of a flock of birds. In addition, stochastic algorithms are able to address multi-objective optimization problems by using the concept of dominance. Accordingly, a set of solutions that do not dominate each other will be obtained, instead of just one best solution. This thesis proposes a multi-objective GP-PSO hybrid algorithm to recover gene regulatory network models that take environmental data as stimulus input. The algorithm infers a model based on both phenotypic and gene expression data. The proposed approach is able to simultaneously infer network structures and estimate their associated parameters, instead of doing one or the other iteratively as other algorithms need to. In addition, a non-dominated sorting approach and an adaptive histogram method based on the hypergrid strategy are adopted to address ‘convergence’ and ‘diversity’ issues in multi-objective optimization. Gene network models obtained from the proposed algorithm are compared to a synthetic network, which mimics key features of Arabidopsis flowering control system, visually and numerically. Data predicted by the model are compared to synthetic data, to verify that they are able to closely approximate the available phenotypic and gene expression data. At the end of this thesis, a novel breeding strategy, termed network assisted selection, is proposed as an extension of our hybrid approach and application of obtained models for plant breeding. Breeding simulations based on network assisted selection are compared to one common breeding strategy, marker assisted selection. The results show that NAS is better both in terms of breeding speed and final phenotypic level.

Many concrete problems may be cast in a multi-objective optimisation framework. The redundancy of existing methods for solving multi-objective programming problems susceptible to inconsistencies, coupled with the necessity for making in- herent assumptions before using a given method, make it hard for a nonspecialist to choose a method that ¯ts the situation at hand well. Moreover, using a method blindly, as suggested by the hammer principle (when you only have a hammer, you want everything in your hand to be a nail) is an awkward approach at best and a caricatural one at worst. This brings challenges to the design, development, implementation and deployment of a Decision Support System able to choose a method that is appropriate for a given problem and to apply the chosen method to solve the problem under consideration. The choice of method should be made according to the structure of the problem and the decision maker's opinion. The aim here is to embed a sample of methods representing the main multi-objective programming techniques and to help the decision maker find the most appropriate method for his problem. / Decisions Sciences / M. Sc. (Operations Research )

Les techniques de groupement technologique sont aujourd’hui utilisées dans de nombreux ateliers de fabrication; elles consistent à décomposer les systèmes industriels en sous-systèmes ou cellules constitués de pièces et de machines. Trouver le groupement technologique le plus efficace est formulé en recherche opérationnelle comme un problème de formation de cellules. La résolution de ce problème permet de tirer plusieurs avantages tels que la réduction des stocks et la simplification de la programmation. Plusieurs critères peuvent être définis au niveau des contraintes du problème tel que le flot intercellulaire,l’équilibrage de charges intracellulaires, les coûts de sous-traitance, les coûts de duplication des machines, etc. Le problème de formation de cellules est un problème d'optimisation NP-difficile. Par conséquent les méthodes exactes ne peuvent être utilisées pour résoudre des problèmes de grande dimension dans un délai raisonnable. Par contre des méthodes heuristiques peuvent générer des solutions de qualité inférieure, mais dans un temps d’exécution raisonnable. Dans ce mémoire, nous considérons ce problème dans un contexte bi-objectif spécifié en termes d’un facteur d’autonomie et de l’équilibre de charge entre les cellules. Nous présentons trois types de méthodes métaheuristiques pour sa résolution et nous comparons numériquement ces métaheuristiques. De plus, pour des problèmes de petite dimension qui peuvent être résolus de façon exacte avec CPLEX, nous vérifions que ces métaheuristiques génèrent des solutions optimales. / Group technology techniques are now widely used in many manufacturing systems. Those techniques aim to decompose industrial systems into subsystems or cells of parts and machines. The problem of finding the most effectivegroup technology is formulated in operations research as the Cell Formation Problem. Several criteria can be used to specify the optimal solution such as flood intercellular, intracellular load balancing, etc. Solving this problem leads to several advantages such as reducing inventory and simplifying programming. The Cell Formation Problem is an NP-hard problem; therefore, exact methods cannot be used to solve large problems within a reasonabletime, whereas heuristics can generate solutions of lower quality, but in a reasonable execution time. We suggest in this work, three different metaheuristics to solve the cell formation problem having two objectives functions: cell autonomy and load balancing between the cells.We compare numerically these metaheuristics. Furthermore, for problems of smaller dimension that can be solved exactly with CPLEX, we verify that the metaheuristics can reach the optimal value.

Cette thèse se compose de deux parties. Dans la partie I, nous avons comparé les différences entre les polymérisations en macro- et mini- émulsion. Nous avons établi un modèle de polymérisation en macro-émulsion. Ce modèle peut prédire la distribution de la taille de la chaîne radicalaire pour une classe donnée de tailles des particules. Par comparaison, nous avons établi un modèle de polymérisation en mini-émulsion. Ce modèle peut fournir la distribution des tailles des particules et l'évolution des tailles des gouttelettes. Dans la partie II, afin d'améliorer les propriétés du PP, deux types de monomères ont été utilisés pour être greffés sur PP par extrusion réactive. Ces monomères sont l'anhydride maléique MAH et le silane. Pour améliorer le degré de greffage de MAH sur PP, la montmorillonite organiquement modifiée o-MMT a été utilisée comme un nanoréacteur. Les influences des quantités de MAH et d'initiateur ont été étudiées. Afin d'améliorer le degré de greffage du silane sur PP, les conditions opératoires ont été étudiées. Toutefois, une relation simple entre les propriétés du produit et les conditions de fonctionnement n'a pas pu être trouvée. Pour obtenir le montant le plus élevé de silane greffé sur PP, et en même temps, restreindre les deux réactions secondaires, c'est à dire la polymérisation du silane et la dégradation de la PP chaîne, modèles polynomiaux ont été tentées pour décrire ce processus / This thesis consists of two parts: In Part I, the differences between macro-and mini-emulsion polymerization were compared. A model on macro-emulsion polymerization has been established. This model can predict the radical chain size distribution for a given class of particle sizes. By comparison, a model on mini-emulsion polymerization has been established. This model can provide the particle size distribution and the change of droplets size. In Part II, in order to improve the properties of PP, two kinds of monomers were used to be grafted on PP by reactive extrusion. These monomers were maleic anhydride (MAH) and silane. To improve the grafting degree of MAH on PP, organically modified montmorillonite (MMT) was used as a nanoreactor. The effects of the amounts of MAH and that of initiator were studied. In order to improve the grafting degree of silane on PP, the operating conditions were studied. However, a simple relation between the product properties and the operating conditions could not be found. To obtain the highest amount of silane grafted on PP, and meanwhile, restraining the two side reactions, i.e. the polymerization of silane and the PP chain degradation, polynomial models were attempted to describe this process

Le but de cette thèse est de proposer des méthodes générales afin de contourner l’intractabilité de problèmes d’optimisation multi-objectifs.Dans un premier temps, nous essayons d’apprécier la portée de cette intractabilité en déterminant une borne supérieure, facilement calculable, sur le nombre de points non dominés, connaissant le nombre de valeurs prises par chaque critère.Nous nous attachons ensuite à produire des représentations discrètes et tractables de l’ensemble des points non dominés de toute instance de problèmes d’optimisation multi-objectifs. Ces représentations doivent satisfaire des conditions de couverture, i.e. fournir une bonne approximation, de cardinalité, i.e. ne pas contenir trop de points, et si possible de stabilité, i.e. ne pas contenir de redondances. En s’inspirant de travaux visant à produire des ensembles ε-Pareto de petite taille, nous proposons tout d’abord une extension directe de ces travaux, puis nous axons notre recherche sur des ensembles ε-Pareto satisfaisant une condition supplémentaire de stabilité. Formellement, nous considérons des ensembles ε-Pareto particuliers, appelés (ε, ε′)-noyaux, qui satisfont une propriété de stabilité liée à ε′. Nous établissons des résultats généraux sur les (ε, ε′)-noyaux puis nous proposons des algorithmes polynomiaux qui produisent des (ε, ε′)-noyaux de petite taille pour le cas bi-objectif et nous donnons des résultats négatifs pour plus de deux objectifs. / The goal of this thesis is to propose new general methods to get around the intractability of multi-objective optimization problems.First, we try to give some insight on this intractability by determining an, easily computable, upper bound on the number of nondominated points, knowing the number of values taken on each criterion. Then, we are interested in producingsome discrete and tractable representations of the set of nondominated points for each instance of multi-objective optimization problems. These representations must satisfy some conditions of coverage, i.e. providing a good approximation, cardinality, i.e. it does not contain too many points, and if possible spacing, i.e. it does not include any redundancies. Starting from works aiming to produce ε-Pareto sets of small size, we first propose a direct extension of these works then we focus our research on ε-Pareto sets satisfying an additional condition of stability. Formally, we consider special ε-Pareto sets, called (ε, ε′)-kernels, which satisfy a property of stability related to ε′. We give some general results on (ε, ε′)-kernels and propose some polynomial time algorithms that produce small (ε, ε′)-kernels for the bicriteria case and we give some negative results for the tricriteria case and beyond.

Dans cette thèse, nous nous intéressons à l'étude des capacités de la méta heuristique d'optimisation par colonie de fourmis (Ant Colony Optimization - ACO) pour résoudre des problèmes d’optimisation combinatoire multi-objectif. Dans ce cadre, nous avons proposé une taxonomie des algorithmes ACO proposés dans la littérature pour résoudre des problèmes de ce type. Nous avons mené, par la suite, une étude expérimentale de différentes stratégies phéromonales pour le cas du problème du sac à dos multidimensionnel mono-objectif. Enfin,nous avons proposé un algorithme ACO générique pour résoudre des problèmes d'optimisation multi-objectif. Cet algorithme est paramétré par le nombre de colonies de fourmis et le nombre de structures de phéromone considérées. Il permet de tester et de comparer, dans un même cadre,plusieurs approches. Nous avons proposé six variantes de cet algorithme dont trois présentent de nouvelles approches et trois autres reprennent des approches existantes. Nous avons appliqué et comparé ces variantes au problème du sac à dos multidimensionnel multi-objectif / In this thesis, we investigate the capabilities of Ant Colony Optimization (ACO) metaheuristic to solve combinatorial and multi-objective optimization problems. First, we propose a taxonomy of ACO algorithms proposed in the literature to solve multi-objective problems. Then, we studydifferent pheromonal strategies for the case of mono-objective multidimensional knapsackproblem. We propose, finally, a generic ACO algorithm to solve multi-objective problems. Thisalgorithm is parameterised by the number of ant colonies and the number of pheromonestructures. This algorithm allows us to evaluate and compare new and existing approaches in thesame framework. We compare six variants of this generic algorithm on the multi-objectivemultidimensional knapsack problem

La présente thèse porte sur l'étude de prise de décisions séquentielles multi-Objectif (MOSDM). La motivation de ce travail est double. D'un côté, la prise de décision, par exemple, dans les domaines de robotique et de planification, concerne l'optimisation séquentielle. De l'autre côté, nombreuses applications dans le monde réel sont plus naturellement formulés en termes d'optimisation multi-Objectif (MOO). La méthode proposée dans la thèse adapte le cadre bien connue de recherche Monte-Carlo arborescente (MCTS) à l'optimisation multi-Objectif, dans lequel multiple séquences de décision optimales sont développées dans un seul arbre de recherche. Le principal défi est de proposer une nouvelle récompense, capable de guider l'exploration de l'arbre bien que le problème de MOO n'applique pas un ordre total entre les solutions. La contribution principale de cette thèse est de proposer et d'étudier expérimentalement ces deux récompenses : l'indicateur de hypervolume et la récompense de dominance Pareto, qui sont inspirées de la littérature de MOO et basés sur une archive de solutions antérieures (archives Pareto). L'étude montre la complémentarité de ces deux récompenses. L'indicateur de hypervolume souffre de sa complexité algorithmique. Cependant, cet indicateur fournit des informations à grains fins de la qualité des solutions à l'égard de l'archive actuelle. Bien au contraire, la complexité de la récompense de dominance Pareto est linéaire, mais cette récompense fournit des informations de plus en plus rare au long de la recherche. Les preuves de principe de l'approche sont donnés sur les problèmes articiaux et les défis internationaux, et confirment la valeur de l'approche. En particulier, MOMCTS est capable de découvrir les politiques se trouvant dans les régions non-Convexes du front Pareto, qui contraste avec l'état de l'art: les algorithmes d'apprentissage par renforcement multi-Objectif existants sont basés sur scalarization linéaire et donc ne sont pas capables de explorer ces régions non-Convexes. Enfin, MOMCTS a fait honorablement la concurrence avec l'état de l'art sur la compétition internationale de MOPTSP 2013. / This thesis is concerned with multi-Objective sequential decision making (MOSDM). The motivation is twofold. On the one hand, many decision problems in the domains of e.g., robotics, scheduling or games, involve the optimization of sequences of decisions. On the other hand, many real-World applications are most naturally formulated in terms of multi-Objective optimization (MOO). The proposed approach extends the well-Known Monte-Carlo tree search (MCTS) framework to the MOO setting, with the goal of discovering several optimal sequences of decisions through growing a single search tree. The main challenge is to propose a new reward, able to guide the exploration of the tree although the MOO setting does not enforce a total order among solutions. The main contribution of the thesis is to propose and experimentally study two such rewards, inspired from the MOO literature and assessing a solution with respect to the archive of previous solutions (Pareto archive): the hypervolume indicator and the Pareto dominance reward. The study shows the complementarity of these two criteria. The hypervolume indicator suffers from its known computational complexity; however the proposed extension thereof provides fine-Grained information about the quality of solutions with respect to the current archive. Quite the contrary, the Pareto-Dominance reward is linear but it provides increasingly rare information. Proofs of principle of the approach are given on artificial problems and challenges, and confirm the merits of the approach. In particular, MOMCTS is able to discover policies lying in non-Convex regions of the Pareto front, contrasting with the state of the art: existing Multi-Objective Reinforcement Learning algorithms are based on linear scalarization and thus fail to sample such non-Convex regions. Finally MOMCTS honorably competes with the state of the art on the 2013 MOPTSP competition.

There are manufacturing systems all over the world and all of them present dif- ferent characteristics. To get close to those manufacturing systems and aid them to analyze data and improve their efficiency, it arises FACTS Analyzer®. The present project concerns the development of a bottleneck analysis using REVERSE- SCORE (Simulation based COnstraint REmoval), feature included in FACTS Ana- lyzer. It is used Simulation-based Multi-Objective Optimization (SMO) to ana- lyze the different variables of a production line and investigate how to best extend previous application of SMO for bottleneck detection to not only consider im- provements of system parameters but also degradations of them. Degrading some system parameters can have many hidden advantages such as reduce power con- sumption, increase material efficiency or lengthen the useful life of the machines or tools, advantages that can draw near sustainability.