Le problème de dimensionnement et planification d'agents en centre d'appels consiste à déterminer sur une période le nombre d'interlocuteurs requis afin d'atteindre la qualité de service exigée et minimiser les coûts induits. Ce sujet fait l'objet d'un intérêt croissant pour son intérêt théorique mais aussi pour l'impact applicatif qu'il peut avoir. Le but de cette thèse est d'établir des approches en contraintes en probabilités en considérant l'incertitude de la demande.Tout d'abord, la thèse présente un modèle en problème d'optimisation stochastique avec contrainte en probabilité jointe traitant la problématique complète en une étape afin d'obtenir un programme facile à résoudre. Une approche basée sur l'idée de continuité est proposée grâce à des lois de probabilité continues, une nouvelle relation entre les taux d'arrivées et les besoins théoriques et la linéarisation de contraintes. La répartition du risque global est faite pendant le processus d'optimisation, permettant une solution au coût réduit. Ces solutions résultantes respectent le niveau de risque tout en diminuant le coût par rapport à d'autres approches.De plus, le modèle en une étape est étendu pour améliorer sa représentation de la réalité. D'une part, le modèle de file d'attente est amélioré et inclus la patience limitée des clients. D'autre part, une nouvelle expression de l'incertitude est proposée pour prendre la dépendance des périodes en compte.Enfin, une nouvelle représentation de l'incertitude est considérée. L'approche distributionally robust permet de modéliser le problème sous l'hypothèse que la loi de probabilité adéquate est inconnue et fait partie d'un ensemble de lois, défini par une moyenne et une variance données. Le problème est modélisé par une contrainte en probabilité jointe. Le risque à chaque période est définie par une variable à optimiser.Un problème déterministe équivalent est proposé et des approximations linéaires permettent d'obtenir une formulation d'optimisation linéaire. / The staffing and shift-scheduling problems in call centers consist in deciding how many agents handling the calls should be assigned to work during a given period in order to reach the required Quality of Service and minimize the costs. These problems are subject to a growing interest, both for their interesting theoritical formulation and their possible applicative effects. This thesis aims at proposing chance-constrained approaches considering uncertainty on demand forecasts.First, this thesis proposes a model solving the problems in one step through a joint chance-constrained stochastic program, providing a cost-reducing solution. A continuous-based approach leading to an easily-tractable optimization program is formulated with random variables following continuous distributions, a new continuous relation between arrival rates and theoritical real agent numbers and constraint linearizations. The global risk level is dynamically shared among the periods during the optimization process, providing reduced-cost solution. The resulting solutions respect the targeted risk level while reducing the cost compared to other approaches.Moreover, this model is extended so that it provides a better representation of real situations. First, the queuing system model is improved and consider the limited patience of customers. Second, another formulation of uncertainty is proposed so that the period correlation is considered.Finally, another uncertainty representation is proposed. The distributionally robust approach provides a formulation while assuming that the correct probability distribution is unknown and belongs to a set of possible distributions defined by given mean and variance. The problem is formulated with a joint chance constraint. The risk at each period is a decision variable to be optimized. A deterministic equivalent problem is proposed. An easily-tractable mixed-integer linear formulation is obtained through piecewise linearizations.

This thesis deals with application of optimisation methods based on linear and mixed-integer linear programming to various problems in the power sector related to electricity production. The thesis goal is to test the applicability of such methods to formulating and solving various instances from the class of real-world electricity production problems, and to find the advantages and disadvantages associated with using these methods. Introductory chapters describe the main characteristics of power markets, including the historical and regulatory context. Fundamental properties of power markets on both demand and supply side are also described, both from a real-world and a modelling point of view. Benefits of optimisation and modelling are discussed, in particular the solution feasibility and optimality as well as insights gained from sensitivity analysis which is often difficult to replicate with the original system. In the core of the thesis, optimisation techniques are applied to three case studies, each of which deals with a specific problem arising during electricity production. In the first problem, the profit of gas-fired power plant in Slovakia from selling power on the day-ahead market is maximised. The model is set up using both technical and commercial constraints. The second problem deals with the problem of representing a two-dimensional production function which primarily arises for a hydro generator with large variations in the level of its reservoir. Several representations of the original function using piecewise linear subsets are presented, compared, and characterised by their computational intensity both theoretically and practically. In the third problem, the prices on the German day-ahead market in 2011 are modelled. Contrary to the previous two models, the model does not capture an optimisation problem faced by a single producer, but incorporates a large subset of the whole market instead. Consequently the model is formed out of generic constraints relevant to all power plants whose parameters are estimated. By combining information about the aggregate availability of power plants with the estimated efficiencies a full supply curve for each day is created. Different scenarios are analysed to test the impact of uncertain inputs such as unknown or estimated constraints. The choice of the investigated problems stems from the attempt to cover electricity production problems from the point of view of multiple criteria. The three investigated electricity production problems span a broad range from the decisions of a single power plant to the modelling a power market as a whole. Formulations of the production function with different level of detail are presented ranging from a simple linear relationship to several bivariate function formulations. While each problem answers a specific question, they all illustrate the ease with which various electricity production problems can solved using optimisation methods based on linear and mixed-integer linear programming. This is mainly due to the ability of these methods to approximate even non-linear functions and constraints over non-convex domains and find global solutions in reasonable time. Moreover, models formulated with these methods allow sensitivity and scenario analyses to be carried out easily as is illustrated in each of the case studies.

[EN] Nowadays, manufacturing systems meet different new global challenges and the existence of a collaborative manufacturing environment is essential to face with. Distributed manufacturing and assembly systems are two manufacturing systems which allow industries to deal with some of these challenges. This thesis studies a production problem in which both distributed manufacturing and assembly systems are considered. Although distributed manufacturing systems and assembly systems are well-known problems and have been extensively studied in the literature, to the best of our knowledge, considering these two systems together as in this thesis is the first effort in the literature. Due to the importance of scheduling optimization on production performance, some different ways to optimize the scheduling of the considered problem are discussed in this thesis. The studied scheduling setting consists of two stages: A production and an assembly stage. Various production centers make the first stage. Each of these centers consists of several machines which are dedicated to manufacture jobs. A single assembly machine is considered for the second stage. The produced jobs are assembled on the assembly machine to form final products through a defined assembly program. In this thesis, two different problems regarding two different production configurations for the production centers of the first stage are considered. The first configuration is a flowshop that results in what we refer to as the Distributed Assembly Permutation Flowshop Scheduling Problem (DAPFSP). The second problem is referred to as the Distributed Parallel Machine and Assembly Scheduling Problem (DPMASP), where unrelated parallel machines configure the production centers. Makespan minimization of the product on the assembly machine located in the assembly stage is considered as the objective function for all considered problems. In this thesis some extensions are considered for the studied problems so as to bring them as close as possible to the reality of production shops. In the DAPFSP, sequence dependent setup times are added for machines in both production and assembly stages. Similarly, in the DPMASP, due to technological constraints, some defined jobs can be processed only in certain factories. Mathematical models are presented as an exact solution for some of the presented problems and two state-of-art solvers, CPLEX and GUROBI are used to solve them. Since these solvers are not able to solve large sized problems, we design and develop heuristic methods to solve the problems. In addition to heuristics, some metaheuristics are also designed and proposed to improve the solutions obtained by heuristics. Finally, for each proposed problem, the performance of the proposed solution methods is compared through extensive computational and comprehensive ANOVA statistical analysis. / [ES] Los sistemas de producción se enfrentan a retos globales en los que el concepto de fabricación colaborativa es crucial para poder tener éxito en el entorno cambiante y complejo en el que nos encontramos. Una característica de los sistemas productivos que puede ayudar a lograr este objetivo consiste en disponer de una red de fabricación distribuida en la que los productos se fabriquen en localizaciones diferentes y se vayan ensamblando para obtener el producto final. En estos casos, disponer de modelos y herramientas para mejorar el rendimiento de sistemas de producción distribuidos con ensamblajes es una manera de asegurar la eficiencia de los mismos. En esta tesis doctoral se estudian los sistemas de fabricación distribuidos con operaciones de ensamblaje. Los sistemas distribuidos y los sistemas con operaciones de ensamblaje han sido estudiados por separado en la literatura. De hecho, no se han encontrado estudios de sistemas con ambas características consideradas de forma conjunta. Dada la complejidad de considerar conjuntamente ambos tipos de sistemas a la hora de realizar la programación de la producción en los mismos, se ha abordado su estudio considerando un modelo bietápico en la que en la primera etapa se consideran las operaciones de producción y en la segunda se plantean las operaciones de ensamblaje. Dependiendo de la configuración de la primera etapa se han estudiado dos variantes. En la primera variante se asume que la etapa de producción está compuesta por sendos sistemas tipo flowshop en los que se fabrican los componentes que se ensamblan en la segunda etapa (Distributed Assembly Permutation Flowshop Scheduling Problem o DAPFSP). En la segunda variante se considera un sistema de máquinas en paralelo no relacionadas (Distributed Parallel Machine and Assembly Scheduling Problem o DPMASP). En ambas variantes se optimiza la fecha de finalización del último trabajo secuenciado (Cmax) y se contempla la posibilidad que existan tiempos de cambio (setup) dependientes de la secuencia de trabajos fabricada. También, en el caso DPMASP se estudia la posibilidad de prohibir o no el uso de determinadas máquinas de la etapa de producción. Se han desarrollado modelos matemáticos para resolver algunas de las variantes anteriores. Estos modelos se han resuelto mediante los programas CPLEX y GUROBI en aquellos casos que ha sido posible. Para las instancias en los que el modelo matemático no ofrecía una solución al problema se han desarrollado heurísticas y metaheurísticas para ello. Todos los procedimientos anteriores han sido estudiados para determinar el rendimiento de los diferentes algoritmos planteados. Para ello se ha realizado un exhaustivo estudio computacional en el que se han aplicado técnicas ANOVA. Los resultados obtenidos en la tesis permiten avanzar en la comprensión del comportamiento de los sistemas productivos distribuidos con ensamblajes, definiendo algoritmos que permiten obtener buenas soluciones a este tipo de problemas tan complejos que aparecen tantas veces en la realidad industrial. / [CAT] Els sistemes de producció s'enfronten a reptes globals en què el concepte de fabricació col.laborativa és crucial per a poder tindre èxit en l'entorn canviant i complex en què ens trobem. Una característica dels sistemes productius que pot ajudar a aconseguir este objectiu consistix a disposar d'una xarxa de fabricació distribuïda en la que els productes es fabriquen en localitzacions diferents i es vagen acoblant per a obtindre el producte final. En estos casos, disposar de models i ferramentes per a millorar el rendiment de sistemes de producció distribuïts amb acoblaments és una manera d'assegurar l'eficiència dels mateixos. En esta tesi doctoral s'estudien els sistemes de fabricació distribuïts amb operacions d'acoblament. Els sistemes distribuïts i els sistemes amb operacions d'acoblament han sigut estudiats per separat en la literatura però, en allò que es coneix, no s'han trobat estudis de sistemes amb ambdós característiques conjuntament. Donada la complexitat de considerar conjuntament ambdós tipus de sistemes a l'hora de realitzar la programació de la producció en els mateixos, s'ha abordat el seu estudi considerant un model bietàpic en la que en la primera etapa es consideren les operacions de producció i en la segona es plantegen les operacions d'acoblament. Depenent de la configuració de la primera etapa s'han estudiat dos variants. En la primera variant s'assumix que l'etapa de producció està composta per sengles sistemes tipus flowshop en els que es fabriquen els components que s'acoblen en la segona etapa (Distributed Assembly Permutation Flowshop Scheduling Problem o DAPFSP). En la segona variant es considera un sistema de màquines en paral.lel no relacionades (Distributed Parallel Machine and Assembly Scheduling Problem o DPMASP). En ambdós variants s'optimitza la data de finalització de l'últim treball seqüenciat (Cmax) i es contempla la possibilitat que existisquen temps de canvi (setup) dependents de la seqüència de treballs fabricada. També, en el cas DPMASP s'estudia la possibilitat de prohibir o no l'ús de determinades màquines de l'etapa de producció. S'han desenvolupat models matemàtics per a resoldre algunes de les variants anteriors. Estos models s'han resolt per mitjà dels programes CPLEX i GUROBI en aquells casos que ha sigut possible. Per a les instàncies en què el model matemàtic no oferia una solució al problema s'han desenrotllat heurístiques i metaheurísticas per a això. Tots els procediments anteriors han sigut estudiats per a determinar el rendiment dels diferents algoritmes plantejats. Per a això s'ha realitzat un exhaustiu estudi computacional en què s'han aplicat tècniques ANOVA. Els resultats obtinguts en la tesi permeten avançar en la comprensió del comportament dels sistemes productius distribuïts amb acoblaments, definint algoritmes que permeten obtindre bones solucions a este tipus de problemes tan complexos que apareixen tantes vegades en la realitat industrial. / Hatami, S. (2016). The Distributed and Assembly Scheduling Problem [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/64072 / TESIS

Dans les réseaux traditionnels, la transmission de flux de données s'effectuaient par routage des paquets de la source vers le ou les destinataires. Le codage réseau (NC) permet aux nœuds intermédiaires du réseau d'effectuer des combinaisons linéaires des paquets de données qui arrivent à leurs liens entrants. Les opérations de codage ont lieu dans un corps de Galois de taille finie q. Aux destinataires, le décodage se fait par une élimination de Gauss des paquets codés-réseau reçus. Cependant, dans les réseaux sans fils, le codage réseau doit souvent faire face à des erreurs de transmission causées par le bruit, les effacements, et les interférences. Ceci est particulièrement problématique pour les applications temps réel, telle la transmission de contenus multimédia, où les contraintes en termes de délais d'acheminement peuvent aboutir à la réception d'un nombre insuffisant de paquets, et par conséquent à des difficultés à décoder les paquets transmis. Dans le meilleurs des cas, certains paquets arrivent à être décodés. Dans le pire des cas, aucun paquet ne peut être décodé.Dans cette thèse, nous proposons des schémas de codage conjoint source-réseau dont l'objectif est de fournir une reconstruction approximative de la source, dans des situations où un décodage parfait est impossible. L'idée consiste à exploiter la redondance de la source au niveau du décodeur afin d'estimer les paquets émis, même quand certains de ces paquets sont perdus après avoir subi un codage réseau. La redondance peut être soit naturelle, c'est-à-dire déjà existante, ou introduite de manière artificielle.Concernant la redondance artificielle, le codage à descriptions multiples (MDC) est choisi comme moyen d'introduire de la redondance structurée entre les paquets non corrélés. En combinant le codage à descriptions multiples et le codage réseau, nous cherchons à obtenir une qualité de reconstruction qui s'améliore progressivement avec le nombre de paquets codés-réseau reçus.Nous considérons deux approches différentes pour générer les descriptions. La première approche consiste à générer les descriptions par une expansion sur trame appliquée à la source avant la quantification. La reconstruction de données se fait par la résolution d'un problème d' optimisation quadratique mixte. La seconde technique utilise une matrice de transformée dans un corps de Galois donné, afin de générer les descriptions, et le décodage se fait par une simple éliminationde Gauss. Ces schémas sont particulièrement intéressants dans un contexte de transmission de contenus multimédia, comme le streaming vidéo, où la qualité s'améliore avec le nombre de descriptions reçues.Une seconde application de tels schémas consiste en la diffusion de données vers des terminaux mobiles à travers des canaux de transmission dont les conditions sont variables. Dans ce contexte, nous étudions la qualité de décodage obtenue pour chacun des deux schémas de codage proposés, et nous comparons les résultats obtenus avec ceux fournis par un schéma de codage réseau classique.En ce qui concerne la redondance naturelle, un scénario typique est celui d'un réseau de capteurs, où des sources géographiquement distribuées prélèvent des mesures spatialement corrélées. Nous proposons un schéma dont l'objectif est d'exploiter cette redondance spatiale afin de fournir une estimation des échantillons de mesures transmises par la résolution d'un problème d'optimisation quadratique à variables entières. La qualité de reconstruction est comparée à celle obtenue à travers un décodage réseau classique. / While network data transmission was traditionally accomplished via routing, network coding (NC) broke this rule by allowing network nodes to perform linear combinations of the upcoming data packets. Network operations are performed in a specific Galois field of fixed size q. Decoding only involves a Gaussian elimination with the received network-coded packets. However, in practical wireless environments, NC might be susceptible to transmission errors caused by noise, fading, or interference. This drawback is quite problematic for real-time applications, such as multimediacontent delivery, where timing constraints may lead to the reception of an insufficient number of packets and consequently to difficulties in decoding the transmitted sources. At best, some packets can be recovered, while in the worst case, the receiver is unable to recover any of the transmitted packets.In this thesis, we propose joint source-network coding and decoding schemes in the purpose of providing an approximate reconstruction of the source in situations where perfect decoding is not possible. The main motivation comes from the fact that source redundancy can be exploited at the decoder in order to estimate the transmitted packets, even when some of them are missing. The redundancy can be either natural, i.e, already existing, or artificial, i.e, externally introduced.Regarding artificial redundancy, we choose multiple description coding (MDC) as a way of introducing structured correlation among uncorrelated packets. By combining MDC and NC, we aim to ensure a reconstruction quality that improves gradually with the number of received network-coded packets. We consider two different approaches for generating descriptions. The first technique consists in generating multiple descriptions via a real-valued frame expansion applied at the source before quantization. Data recovery is then achieved via the solution of a mixed integerlinear problem. The second technique uses a correlating transform in some Galois field in order to generate descriptions, and decoding involves a simple Gaussian elimination. Such schemes are particularly interesting for multimedia contents delivery, such as video streaming, where quality increases with the number of received descriptions.Another application of such schemes would be multicasting or broadcasting data towards mobile terminals experiencing different channel conditions. The channel is modeled as a binary symmetric channel (BSC) and we study the effect on the decoding quality for both proposed schemes. Performance comparison with a traditional NC scheme is also provided.Concerning natural redundancy, a typical scenario would be a wireless sensor network, where geographically distributed sources capture spatially correlated measures. We propose a scheme that aims at exploiting this spatial redundancy, and provide an estimation of the transmitted measurement samples via the solution of an integer quadratic problem. The obtained reconstruction quality is compared with the one provided by a classical NC scheme.

[pt] Em um cenário mundial de intensa competição, com indústrias cada vez mais complexas e dinâmicas, a tomada de decisão se tornou um fator chave que muitas vezes pode determinar a sobrevivência e o sucesso de uma empresa. A indústria do petróleo possui investimentos anuais na ordem de bilhões de dólares e por isso toda decisão é considerada vital para esta indústria. Em função da grande quantidade de projetos sendo estudados e propostos, tanto de ampliação quanto de manutenção da estrutura das empresas, do alto valor de cada um e da restrição orçamentária de cada empresa, a tomada de decisão é um passo importante e complexo, devido às inúmeras variáveis e restrições envolvidas. Esta dissertação elabora uma proposta de método de apoio para o processo de tomada de decisão usado na seleção de projetos de investimento em logística de uma grande empresa do setor de óleo e gás. Na maioria das vezes, os critérios utilizados para a tomada de decisão na seleção de projetos consideram apenas os aspectos financeiros, sem levar em consideração outros critérios, que são abordados nesse trabalho. Em função da importância dos projetos e da quantidade de variáveis envolvidas durante a seleção e priorização deles, esse trabalho utiliza o método AHP (Analytic Hierarchy Process) Simplificado e os conceitos de Programação Linear Inteira Mista (PLIM) para determinar os projetos prioritários e otimizar a satisfação do grupo tomador de decisão. Foram estudados seis dos principais projetos da empresa, que juntos somam, aproximadamente, 2,7 bilhões de doláres e representam quase 80 por cento da carteira da área de logística da empresa estudada. Este estudo possibilitou uma seleção mais eficaz, resultando em uma maior integração entre os tomadores de decisão e um maior conhecimento sobre cada um dos projetos avaliados. / [en] In a global scenario of intense competition with increasingly complex and dynamic industries, decision-making has become a key factor that can often determine the survival and success of a company. The oil and gas industry has annual investments of billions of dollars and every decision is considered vital for this industry. Due to the large number of projects being proposed and studied, involving both expansion and maintenance of business structure, the high value of each and the budget constraint of each company, the decision-making is important and complex, due to numerous variables and constraints involved. This dissertation develops a proposed method of support to the process of decision making used in the selection of logistics investment projects, in a large company in the oil and gas industry. In most cases, the criteria used for decision making in project selection only considers the financial aspects, without taking into account other ones, which are addressed in this work. Due to the importance of the projects and the amount of variables involved when selecting and prioritizing them, this work uses the Simplified AHP (Analytic Hierarchy Process) and concepts of Mixed Integer Linear Programming (MILP) to determine priority projects and optimize the satisfaction of the decision-making group. Six of the main projects of the enterprise were studied, which together sum up to approximately 2.7 billion dollars and represent almost 80 per cent of the logistics portfolio of the studied company. This study allowed a more effective selection, resulting in a greater integration between decision makers and a greater knowledge of each of the evaluated projects.

Orientador: Antônio Fernando Branco Costa / Resumo: O presente trabalho desenvolveu um modelo matemático que contribui para o projeto de redes cicloviárias conexas e integradas ao sistema de transporte público por ônibus em áreas urbanas, fazendo uso da Otimização Multiobjetivo (OM) por meio da Programação Linear Inteira Mista (MILP ou MIP). A investigação dos critérios de infraestrutura que influenciam a demanda cicloviária e que deveriam compor a modelagem matemática foi realizada por meio da Revisão da Literatura e cinco critérios identificados, referentes ao: Conforto, Segurança, Objetividade, Intermodalidade e Continuidade. Cada critério é composto por uma série de subcritérios, dezesseis no total, classificados como parâmetros de projeto de entrada e/ou de saída para o modelo matemático, segundo funções objetivos e restrições. Uma aplicação do modelo foi realizada em um caso real para cidade de grande porte, em São José dos Campos – SP, e obtidas soluções exatas com redes parcialmente conexas (abordagem 1) e redes totalmente conexas (abordagem 2), integradas aos pontos de parada de ônibus e as linhas de ônibus da região delimitada, por meio da linguagem de modelagem General Algebric Modeling System (GAMS) e o solver CPLEX. Os resultados obtidos foram aderentes a realidade da macrozona de tráfego na região estudada e contribuem para o planejamento urbano como validado pela Secretaria de Mobilidade Urbana de São José dos Campos. A análise de diversos cenários para o modelo matemático, ponderando os objetivos do modelo, pos... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: The present work created a mathematical model that contributes to design bikeway networks integrated to the public transport service by buses in urban areas, using the Network Multiobjective Optimization through Mixed by means of Mixed Integer Linear Programming (MILP or MIP). The investigation of the infrastructure criteria that influence cycling demand and that should set the mathematical modeling was carried out through the Literature Review and five criteria were identified, referring to: Comfort, Safety, Objectivity, Intermodality and Connectivity. Each criteria is composed of a subcriteria set, sixteen in total, which performed as input and/or output design parameters for the mathematical model, according to fitness functions and constraints. An application of the model was carried out in a real case in São Jose dos Campos city - Sao Paulo state, and exact solutions were obtained with partially connected networks (approach 1) and fully connected networks (approach 2), integrated to the stop points by bus and bus lines in the defined region, using the General Algebric Modeling System (GAMS) modeling language and the CPLEX solver. The results obtained were adherent to the reality of the traffic macrozone in the studied region and contribute to urban planning, as validated by the São José dos Campos Urban Mobility Secretariat. The analysis of several scenarios for the mathematical model, considering the model's objectives, made it possible to verify the behavior of the res... (Complete abstract click electronic access below) / Doutor

La plupart des systèmes ferroviaires subissent une demande croissante de capacité. Pour y faire face, il faut construire de nouvelles infrastructures ou exploiter plus efficacement celles existantes, notamment en définissant des grilles horaires optimisées. Dans la littérature, la plupart des approches de construction des grilles sont basées sur des représentations macroscopiques de l'infrastructure, ce qui peut conduireà des solutions infaisables ou inefficaces. En revanche, les approches microscopiques reposent sur une modélisation réaliste du système ferroviaire, ce qui garantit la faisabilité et l'efficacité des résultats. Néanmoins, en raison de leur complexité, l'utilisation de ces approches est généralement limitée à une seule gare. Malgré l'optimisation de la grille horaire, les travaux de maintenance peuvent avoir un fort impact sur les circulations des trains. En présence de maintenances, il peut donc être nécessaire de redéfinir la grille horaire pour assurer une exploitation efficace de la capacité. Nous présentons deux contributions principales sous forme de deux approches microscopiques : une pour la conception de grilles horaires et l'autre pour leur redéfinition en cas de maintenance. La deuxième est la première approche microscopique qui apparaît dans la littérature pour aborder ce problème tout en considérant des aspects comme les limitations temporaires de vitesse. Nous démontrons la validité de nos approches et leur applicabilité dans des scénarios réels. De plus, nous montrons que les approches microscopiques peuvent être utilisées pour traiter des zones de l'infrastructure contenant plusieurs gares. / Most railway systems experience a growing demand of railway capacity. To face this demand, either new infrastructure must be built or a more efficient exploitation of the existing one must be attained. Timetables play a determinant role in the efficient capacity exploitation. Most timetabling approaches in the literature are based on macroscopic representations of the infrastructure. This may lead to inefficient and in some cases, impractical solutions. Instead, microscopic approaches are based on more realistic modelling of the elements of the railway system. This guarantees the feasibility of the timetables while promoting an efficient capacity exploitation. However, due to their complexity, the scope of microscopic approaches is typically restricted to main stations. Despite the optimization of timetables, the performance of infrastructure maintenance may severely impact the trains' circulations in the network. Therefore, the timetable may have to be rearranged to ensure an efficient capacity exploitation. We present two main contributions in this thesis: first, a microscopic approach for timetable design. Second, a microscopic approach for timetable rearrangement to cope with maintenance. This is the first microscopic approach in the literature to tackle this problem while also considering specific aspects as temporary speed limitations. After a thorough experimental analysis, we demonstrate the validity of our approaches and their practical applicability in real life scenarios. In particular, we show that microscopic approaches can be used to tackle large areas of the infrastructure, including several stations.

MSc (Statistics) / Department of Statistics / Generally, short term load forecasting is essential for any power generating utility. In this dissertation the main objective was to develop short term load forecasting models for the peak demand periods (i.e. from 18:00 to 20:00 hours) in South Africa using. Quantile semi-parametric additive models were proposed and used to forecast electricity demand during peak hours. In addition to this, forecasts obtained were then used to nd an optimal number of generating units to commit (switch on or o ) daily in order to produce the required electricity demand at minimal costs. A mixed integer linear programming technique was used to nd an optimal number of units to commit. Driving factors such as calendar e ects, temperature, etc. were used as predictors in building these models. Variable selection was done using the least absolute shrinkage and selection operator (Lasso). A feasible solution to the unit commitment problem will help utilities meet the demand at minimal costs. This information will be helpful to South Africa's national power utility, Eskom. / NRF

In modern days electric power systems, the penetration of renewable resources and the introduction of free market principles have led to new challenges facing the power producers and regulators. Renewable production is intermittent which leads to fluctuations in the grid and requires more control for regulators, and the free market principle raises the challenge for power plant producers to operate their plants in the most profitable way given the fluctuating prices. Those problems are addressed in the literature as the Economic Dispatch, and they have been discussed from both regulator and producer view points. Combined Cycle Power plants have the privileges of being dispatchable very fast and with low cost which put them as a primary solution to power disturbance in grid, this fast dispatch-ability also allows them to exploit price changes very efficiently to maximize their profit, and this sheds the light on the importance of prices forecasting as an input for the profit optimization of power plants. In this project, an integrated solution is introduced to optimize the dispatch of combined cycle power plants that are bidding for electricity markets, the solution is composed of two models, the forecasting model and the optimization model. The forecasting model is flexible enough to forecast electricity and fuel prices for different markets and with different forecasting horizons. Machine learning algorithms were used to build and validate the model, and data from different countries were used to test the model. The optimization model incorporates the forecasting model outputs as inputs parameters, and uses other parameters and constraints from the operating conditions of the power plant as well as the market in which the plant is selling. The power plant in this mode is assumed to satisfy different demands, each of these demands have corresponding electricity price and cost of energy not served. The model decides which units to be dispatched at each time stamp to give out the maximum profit given all these constraints, it also decides whether to satisfy all the demands or not producing part of each of them.

No description available.