[pt] O objetivo desta dissertação é desenvolver uma metodologia para oferta estratégia de uma empresa geradora em ambiente de mercado com múltiplas usinas hidrelétricas, levando em consideração múltiplos estágios e a incerteza nas afluências, e ilustrar a aplicação da mesma em sistemas realistas. Mostra-se inicialmente que o problema de oferta estratégica pode ser formulado como uma recursão de programação dinâmica estocástica (PDE), onde as variáveis de estado são os níveis de armazenamento dos reservatórios no início de cada estágio e as afluências observadas nos estágios anteriores. Entretanto, a dificuldade computacional dos algoritmos de PDE restringe sua aplicação a sistemas com poucos reservatórios, limitando bastante a aplicação da técnica a sistemas realistas. Assim, a abordagem proposta nesta dissertação é estender a metodologia de programação dinâmica dual estocástica (PDDE), até então aplicada a problemas de minimização de custos, ao problema de otimização da oferta. Isto é feito através de dois passos principais: (i) Uso de uma estratégia de oferta por quantidade somente (análogo a um modelo de Cournot em problemas de equilíbrio econômico) e (ii) a recursão de PDDE, que por ser baseada numa aproximação por hiperplanos requer que o problema seja convexo, o que não ocorre necessariamente no caso da oferta estratégica. A abordagem proposta consiste em aproximar a cada estágio a função de benefício futuro (FBF) por sua envoltória côncava (concave hull). Com isso, a técnica de PDDE pode ser aplicada para resolver o problema de ofertas multi- estágio e estocástico de uma empresa hidroelétrica com múltiplas usinas. Exemplos e estudos de caso serão ilustrados com os sistemas reais da Romênia e El Salvador, ilustrando a aplicabilidade da metodologia proposta em estudos e análises de poder de mercado. / [en] The objective of this work is to present a methodology for the strategic bidding (or bid-based) problem of a hydropower based company, taking into account multiple hydro plants, time-coupling, multiple inflow scenarios and illustrate its application for real case studies. It is initially show that the bid-based dispatch for a hydro plant can be formulated as a stochastic dynamic programming (SDP) recursion scheme, where the state variables are the storage levels and the past inflows. As widely known, the computational effort of the SDP algorithms restricts its applications for systems with just a few reservoirs, which is not the case of the real world systems. Therefore, the approach proposed in this thesis is to extend the stochastic dual dynamic programming (SDDP) scheme, usually applied to cost minimization problems, to the strategic bidding problem. This is done through two main steps: (i) use of a quantity-only bidding scheme (similar to the Cournot model of economic equilibria); (ii) SDDP recursion, which is based on a linear approximation by piecewise linear segments and thus requires that the underlying problem to be convex. This is not necessarily observed in the strategic bidding problem. Thus, the proposed approach consists in approximating, at each stage, the future benefit function (FBF) by its concave hull, which then assures that the SDDP scheme can be applied to solve the multi-stage and stochastic strategic bidding problem of a company with a portfolio of several hydro plants. The proposed approach is illustrated with examples and case studies from real hydro systems from Rumania and El Savador, where market power analysis will be presented.

No description available.

Par delà les études transversales, étudier l'évolution temporelle de phénomènes connait un intérêt croissant. En effet, pour comprendre un phénomène, il semble plus adapté de comparer l'évolution des marqueurs de celui-ci au cours du temps plutôt que ceux-ci à un stade donné. Le suivi de maladies neuro-dégénératives s'effectue par exemple par le suivi de scores cognitifs au cours du temps. C'est également le cas pour le suivi de chimiothérapie : plus que par l'aspect ou le volume des tumeurs, les oncologues jugent que le traitement engagé est efficace dès lors qu'il induit une diminution du volume tumoral.L'étude de données longitudinales n'est pas cantonnée aux applications médicales et s'avère fructueuse dans des cadres d'applications variés tels que la vision par ordinateur, la détection automatique d'émotions sur un visage, les sciences sociales, etc.Les modèles à effets mixtes ont prouvé leur efficacité dans l'étude des données longitudinales, notamment dans le cadre d'applications médicales. Des travaux récent (Schiratti et al., 2015, 2017) ont permis l'étude de données complexes, telles que des données anatomiques. L'idée sous-jacente est de modéliser la progression temporelle d'un phénomène par des trajectoires continues dans un espace de mesures, que l'on suppose être une variété riemannienne. Sont alors estimées conjointement une trajectoire moyenne représentative de l'évolution globale de la population, à l'échelle macroscopique, et la variabilité inter-individuelle. Cependant, ces travaux supposent une progression unidirectionnelle et échouent à décrire des situations telles que la sclérose en plaques ou le suivi de chimiothérapie. En effet, pour ces pathologies, vont se succéder des phases de progression, de stabilisation et de remision de la maladie, induisant un changement de la dynamique d'évolution globale.Le but de cette thèse est de développer des outils méthodologiques et algorithmiques pour l’analyse de données longitudinales, dans le cas de phénomènes dont la dynamique d'évolution est multiple et d'appliquer ces nouveaux outils pour le suivi de chimiothérapie. Nous proposons un modèle non-linéaire à effets mixtes dans lequel les trajectoires d'évolution individuelles sont vues comme des déformations spatio-temporelles d'une trajectoire géodésique par morceaux et représentative de l'évolution de la population. Nous présentons ce modèle sous des hypothèses très génériques afin d'englober une grande classe de modèles plus spécifiques.L'estimation des paramètres du modèle géométrique est réalisée par un estimateur du maximum a posteriori dont nous démontrons l'existence et la consistance sous des hypothèses standards. Numériquement, du fait de la non-linéarité de notre modèle, l'estimation est réalisée par une approximation stochastique de l'algorithme EM, couplée à une méthode de Monte-Carlo par chaînes de Markov (MCMC-SAEM). La convergence du SAEM vers les maxima locaux de la vraisemblance observée ainsi que son efficacité numérique ont été démontrées. En dépit de cette performance, l'algorithme SAEM est très sensible à ses conditions initiales. Afin de palier ce problème, nous proposons une nouvelle classe d'algorithmes SAEM dont nous démontrons la convergence vers des minima locaux. Cette classe repose sur la simulation par une loi approchée de la vraie loi conditionnelle dans l'étape de simulation. Enfin, en se basant sur des techniques de recuit simulé, nous proposons une version tempérée de l'algorithme SAEM afin de favoriser sa convergence vers des minima globaux. / Beyond transversal studies, temporal evolution of phenomena is a field of growing interest. For the purpose of understanding a phenomenon, it appears more suitable to compare the evolution of its markers over time than to do so at a given stage. The follow-up of neurodegenerative disorders is carried out via the monitoring of cognitive scores over time. The same applies for chemotherapy monitoring: rather than tumors aspect or size, oncologists asses that a given treatment is efficient from the moment it results in a decrease of tumor volume. The study of longitudinal data is not restricted to medical applications and proves successful in various fields of application such as computer vision, automatic detection of facial emotions, social sciences, etc.Mixed effects models have proved their efficiency in the study of longitudinal data sets, especially for medical purposes. Recent works (Schiratti et al., 2015, 2017) allowed the study of complex data, such as anatomical data. The underlying idea is to model the temporal progression of a given phenomenon by continuous trajectories in a space of measurements, which is assumed to be a Riemannian manifold. Then, both a group-representative trajectory and inter-individual variability are estimated. However, these works assume an unidirectional dynamic and fail to encompass situations like multiple sclerosis or chemotherapy monitoring. Indeed, such diseases follow a chronic course, with phases of worsening, stabilization and improvement, inducing changes in the global dynamic.The thesis is devoted to the development of methodological tools and algorithms suited for the analysis of longitudinal data arising from phenomena that undergo multiple dynamics and to apply them to chemotherapy monitoring. We propose a nonlinear mixed effects model which allows to estimate a representative piecewise-geodesic trajectory of the global progression and together with spacial and temporal inter-individual variability. Particular attention is paid to estimation of the correlation between the different phases of the evolution. This model provides a generic and coherent framework for studying longitudinal manifold-valued data.Estimation is formulated as a well-defined maximum a posteriori problem which we prove to be consistent under mild assumptions. Numerically, due to the non-linearity of the proposed model, the estimation of the parameters is performed through a stochastic version of the EM algorithm, namely the Markov chain Monte-Carlo stochastic approximation EM (MCMC-SAEM). The convergence of the SAEM algorithm toward local maxima of the observed likelihood has been proved and its numerical efficiency has been demonstrated. However, despite appealing features, the limit position of this algorithm can strongly depend on its starting position. To cope with this issue, we propose a new version of the SAEM in which we do not sample from the exact distribution in the expectation phase of the procedure. We first prove the convergence of this algorithm toward local maxima of the observed likelihood. Then, with the thought of the simulated annealing, we propose an instantiation of this general procedure to favor convergence toward global maxima: the tempering-SAEM.

This thesis deals with modelling of continuous casting of steel. This process of steel manufacturing has achieved dominant position not only in the Czech Republic but also worldwide. The solved casted bar cross-section shape is circular, because it is rarely studied in academical works nowadays. First part of thesis focuses on creating numerical model of thermal field, using finite difference method with cylindrical coordinates. This model is then employed in optimization part, which represents control problem of abrupt step change of casting speed. The main goal is to find out, whether the computation of numerical model and optimization both can be parallelized using spatial decomposition. To achieve that, Progressive Hedging Algorithm from the field of stochastic optimization has been used.

The main aim of this thesis is to create a sequence of mathematical optimization models with different levels of complexity for the efficient management and waste energy utilization. Stochastic programming approach was utilized to deal with random demand and uncertain heating values. Hence, more applicable model of the waste-to-energy plant has been developed. As the next step, the model is enhanced by heating plant extension. Computations are realized for real-world data and optimal solution is found by using GAMS implementation.

[pt] Este trabalho apresenta um framework de investimento dinâmico para carteiras de energia, baseados em opções reais, que visa maximizar o valor, corrigido pelo risco, do investimento conjunto em projetos de geração de energia com fontes renováveis. Diferente de outros modelos semelhantes, várias classes de incerteza são levadas em consideração simultaneamente e os valores de projeto são calculados por um modelo de otimização híbrido robusto e estocástico. O framework de investimento é adequado para qualquer mercado que permita a negociação bilateral, conforme feita no Ambiente de Contratação Livre, e é construído na visão da empresa de geração, ou comercializadora de energia, que pretende investir em uma carteira de geração. Utilizando este framework é possível definir o quanto investir em cada fonte renovável, quanto vender da carteira de energia e o melhor momento para investir. Além disso, com essa modelagem é calculado o prêmio do investimento simultâneo em fontes renováveis complementares. Ele estende os modelos de decisão estáticos, já abordados na literatura, para um contexto dinâmico, ou seja, considerando a decisão ótima de investimento no tempo. Isso é feito utilizando a abordagem numérica desenvolvida por Bastian-Pinto [9], para descrever cenários de variáveis estocásticas que se comportam como um processo de reversão à média (típico dos preços de energia). Ao final são mostrados estudos de caso realistas que demonstram o valor do framework. Este modelo aprimora as decisões da indústria de energia, contribui para aumentar a competitividade das fontes renováveis e reduz a necessidade de subsídios para o investimento. Com isso, impulsiona a penetração das fontes renováveis no mercado brasileiro de energia elétrica. / [en] This dissertation presents a dynamic framework for renewable energy portfolios, based on real options, that maximize the risk-averse investment value. Differently from similar models, several classes of uncertainty are taken into account simultaneously and the project values are calculated by means of a hybrid robust and stochastic optimization model. The investment framework is suitable for any market that allows bilateral trading (as in the Brazilian free contracting environment) and is designed for a generation company or energy trading company, that intends to invest in a renewablesource portfolio. Using this framework it is possible to define how much to buy or build from each renewable source, how much to sell from the energy portfolio, and the best moment to invest. Additionally, the premium for investing simultaneously in several complementary renewable sources is also determined. The section responsible for supporting the dynamic investment timing decision uses the binomial lattice proposed by Bastian-Pinto et al [9], to describe mean reverting processes. This framework improves industry practices, contributes to increase renewables competitiveness and proposes an arragement that reduces the need for subsidies. As a consequence, this model contributes to foster the penetration of renewable sources in Brazilian electricity market.

Cette thèse s'intéresse à la résolution de très grands problèmes d'optimisation combinatoire stochastique. Les recherches sont appliquées au problème de planification des arrêts pour rechargement des centrales nucléaires. Compte-tenu de la part prépondérante de celles-ci dans le mix-électrique, ce problème structure fortement la chaîne de management d’énergie d'EDF. Une première partie propose une formulation étendue bi-niveau dans laquelle les décisions de premier niveau fixent les plannings d’arrêt et des profils de production des centrales, et celles de second niveau évaluent le coût de satisfaction de la demande associé. Cette formulation permet la résolution à l'optimum d'instances industrielles déterministes par un solveur en PLNE. Dans le cas stochastique, une telle résolution directe du problème n'est plus possible. Nous proposons une formulation permettant d’en résoudre la relaxation linéaire par génération de colonnes et de coupes, correspondant respectivement aux reformulations de Danzig-Wolfe du premier niveau et de Benders du second. Une phase heuristique permet ensuite de déterminer des solutions entières de bonne qualité pour des instances, jusqu'à une cinquantaine de scénarios représentatifs de l’incertitude sur les données. L’apport de l’approche est estimé en utilisant les outils industriels exploités par EDF pour évaluer les plannings. Une seconde partie porte sur l'intégration de méthodes d'optimisation robuste pour la prise en compte d’aléas sur la disponibilité des centrales. Nous nous plaçons dans un cadre où les recours possibles sur les dates d'arrêts ne sont pas exercés. Nous comparons des méthodes bi-objectif et probabiliste permettant de rendre le planning robuste pour les contraintes opérationnelles dont la relaxation est envisageable. Pour les autres, nous proposons une méthode basée sur un budget d’incertitude. Cette méthode permet de renforcer la stabilité du planning en limitant les besoins de réorganisation futurs. La prise en compte d’une loi de probabilité de l’aléa permet d’affiner le contrôle du prix de cette robustesse. / The purpose of this Ph.D. thesis is to study optimization techniques for large-scale stochastic combinatorial problems. We apply those techniques to the problem of scheduling EDF nuclear power plant maintenance outages, which is of significant importance due to the major part of the nuclear energy in the French electricity system. We build on a two-stages extended formulation, the first level of which fixes nuclear outage dates and production profiles for nuclear plants, while the second evaluates the cost to meet the demand. This formulation enables the solving of deterministic industrial instances to optimality, by using a MIP solver. However, the computational time increases significantly with the number of scenarios. Hence, we resort to a procedure combining column generation of a Dantzig-Wolfe decomposition with Benders’ cut generation, to account for the linear relaxation of stochastic instances. We then obtain integer solutions of good quality via a heuristic, up to fifty scenarios. We further assume that outage durations are uncertain and that unexpected shutdowns of plants may occur. We investigate robust optimization methods in this context while ignoring possible recourse on power plants outage dates. We report on several approaches, which use bi-objective or probabilistic methods, to ensure the satisfaction of constraints which might be relaxed in the operating process. For other constraints, we apply a budget uncertainty-based approach to limit future re-organizations of the scheduling. Adding probabilistic information leads to better control of the price of the robustness.

In this thesis we investigate the problem of portfolio re-balancing for fixed periods using a cost/risk adaptation model and stochastic optimization. The cost/risk adaptation model takes theory of optimal liquidity costs and risk preference to build a universe in which we try to find better strategies than conventional ones. The results are focused on the comparison between the conventional execution strategies versus our developed model. We have found that our model outperforms the conventional methods for all assets that has been evaluated, and especially for investors whom value exposure to the markets higher.

We consider a bilevel continuous knapsack problem where the leader controls the capacity of the knapsack, while the follower chooses a feasible packing maximizing his own profit. The leader’s aim is to optimize a linear objective function in the capacity and in the follower’s solution, but with respect to different item values. We address a stochastic version of this problem where the follower’s profits are uncertain from the leader’s perspective, and only a probability distribution is known. Assuming that the leader aims at optimizing the expected value of her objective function, we first observe that the stochastic problem is tractable as long as the possible scenarios are given explicitly as part of the input,which also allows to deal with general distributions using a sample average approximation. For the case of independently and uniformly distributed item values, we show that the problem is #P-hard in general, and the same is true even for evaluating the leader’s objective function. Nevertheless, we present pseudo-polynomial time algorithms for this case, running in time linear in the total size of the items.Based on this,we derive an additive approximation scheme for the general case of independently distributed item values, which runs in pseudo-polynomial time.

Modern power systems are at the forefront of addressing the transition towards a decarbonized energy system, with increased integration of renewable energy sources becoming the key means to achieve the same. In this context, an expansion planning problem provides a decision support framework for investments in new generation and transmission assets over a long time-frame, addressing a wide range of technical and economic criteria aligned with national policies. It is thus an important but complex problem to solve, involving a number of modelling challenges and uncertainties to be considered. These include the treatment of operational variability due to the intermittent nature of renewable sources like wind or solar, while also considering uncertainties such as demand growth, technological developments impacting future investment costs among others, that define the long-term dynamics in the expansion. In this regard, one of the goals of the EU Project Spine is to create open source models to study investment scenarios for the expansion of power systems. This thesis work aims to offer insights for the Spine project, by identifying and exploring various requirements pertinent to a planning problem, along with method development. The primary objective of this thesis is to develop an expansion planning model, that determines the optimal location, size or capacity and time of investment for different candidates in generation and transmission assets, including short-term energy storage. A Mixed-Integer Linear Programming (MILP) optimization problem is formulated for the same, with both investment and operational sub-problems solved together. Operational variability has been modelled in a reduced form using profiles of representative days, while also incorporating contemporary requirements in the planning problem such as penetration targets for renewable generation. The developed model has been evaluated using a case study done on a small test network, in which different expansion scenarios involving varying demand growth and phase-out of conventional generators are investigated. Also, a two-stage stochastic optimization is performed to consider long-term uncertainties in demand growth and the quality of the stochastic solution is analyzed. It is inferred from the results that the expansion solution is indeed different for different scenarios, and stochastic optimization proves to be important in addressing long-term uncertainties, as reflected by a high value of stochastic solution (VSS). / Moderna kraftsystem ligger i framkant när det kommer till omställningen till ett fossilfritt energisystem. Ökad integrering av förnybara energikällor är den främsta lösningen för att uppnå detta. I samband med omställningen kan ett utbyggnadsplanering problem bidra till ramverk för investeringsbeslut för genererings- och transmissionstillgångar över ett långt tidsspann, vilket tar hänsyn till en bredd av tekniska och ekonomiska kriterier i linje med nationella policyer. Detta är ett viktigt, men ett komplext problem som inkluderar ett stort antal modelleringsutmaningar och osäkerheter som måste beaktas. Bland annat inkluderas hur drift varierar på grund av de förnybara energikällornas intermittenta karaktär såsom vind och sol, medan osäkerheter kring hur efterfrågan utvecklas, tekniska framsteg som påverkar framtida investeringar m.m., också behöver vägas in. Med hänsyn till detta är ett av målen för EU Project Spine att skapa en open source för modeller med syftet att studera investeringsscenarier när kraftsystemet expanderar. Syftet med detta examensarbete är att ge insikt i Spine-projektet genom att identifiera och utforska olika relevanta krav för ett planneringsproblem samt att utveckla metoder. Det huvudsakliga målet för detta examensarbete är att utveckla en expanderad planeringsmodell som bestämmer optimal placering, storlek, kapacitet och tidpunkt för investering för olika typer av genererings- och transmissionstillgångar samt kortvarig energilagring. Ett mixed-integer linjär programmering (MILP) optimeringsproblem har formulerats, där både investerings- och drifts-subproblem beräknas tillsammans. Variabel drift av kraftsystemet har modellerats på reducerad form genom att använda profiler för representativa dagar, därtill inkluderas samtida krav i planeringen såsom mål för penetreringsnivå av förnybara energikällor. Den utvecklade modellen har utvärderats i en fallstudie på en liten nätmodell där olika scenarier har utforskats. I scenarierna varieras tillväxten på efterfrågan och när utfasningen av konventionella generatorer sker. Därtill, sker en två-stegs stokastisk optimering för att ta hänsyn till långsiktiga osäkerheter kring tillväxten på efterfrågan och kvalitén av den stokastiska lösningen har analyserats. Resultaten visar att den expanderade lösningen är olika för olika scenarier, och att den stokastiska optimeringen är viktig när långsiktiga osäkerheter måste beaktas som visas genom ett högt värde för den stokastiska lösningen (VSS).