Submitted by DARWIN ALEXIS QUIJANO RODEZNO (alexisqr@yahoo.es) on 2018-05-08T15:12:06Z No. of bitstreams: 1 TeseDarwinQuijano.pdf: 3515679 bytes, checksum: f52d72089eda5a3bb15b367d3afbf33a (MD5) / Approved for entry into archive by Cristina Alexandra de Godoy null (cristina@adm.feis.unesp.br) on 2018-05-08T17:48:35Z (GMT) No. of bitstreams: 1 quijanorodezno_da_dr_ilha.pdf: 3515679 bytes, checksum: f52d72089eda5a3bb15b367d3afbf33a (MD5) / Made available in DSpace on 2018-05-08T17:48:35Z (GMT). No. of bitstreams: 1 quijanorodezno_da_dr_ilha.pdf: 3515679 bytes, checksum: f52d72089eda5a3bb15b367d3afbf33a (MD5) Previous issue date: 2018-04-19 / Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Atualmente, existe uma tendência para aumentar a participação da Geração Distribuída (GD) baseada em Fontes de Energia Renováveis (FER) no suprimento do consumo global de energia elétrica. Esta tendência está sendo impulsionada principalmente por iniciativas governamentais destinadas a aumentar a eficiência energética, aumentar o uso da energia proveniente das FER e reduzir as emissões de gases de efeito estufa. No entanto, à medida que seu nível de penetração aumenta, a GD pode dar origem a um sistema incapaz de fornecer energia de forma confiável e de acordo com os padrões de qualidade. Nesse cenário, a Gestão de Redes Ativas (GRA) surge como uma alternativa para a integração de grandes montantes de GD. A GRA promove a disponibilização de instrumentos comerciais e regulatórios, e o fornecimento das redes de distribuição com tecnologias de automação para procurar serviços ancilares e flexibilidade a partir da GD. A GRA requer o desenvolvimento de ferramentas computacionais para coordenar a implementação de esquemas de controle inteligentes, chamados de esquemas de GRA, a fim de otimizar a utilização e operação das redes. Neste trabalho, são propostos modelos de otimização e técnicas de solução para a GRA considerando a integração de GD solar fotovoltaica e eólica e a eficiência energética. O primeiro modelo é desenvolvido para determinar a capacidade máxima de GD que pode ser alocada em uma rede de distribuição quando se considera o efeito da tensão na eficiência das cargas. No segundo modelo, o procedimento de controlar os níveis de tensão para reduzir a demanda das cargas é implementado para economia de energia e para o balanço de geração e demanda, através de uma estratégia projetada para o planejamento da operação de redes de distribuição ativas. Em ambos os modelos as incertezas são consideradas através de formulações de programação estocásticas de dois estágios. Os esquemas de GRA considerados são o controle coordenado da tensão através de reguladores de tensão e transformadores com comutador de tap sob carga, suporte de potência reativa através da GD, e corte de geração. A técnica de solução envolve a discretização das funções de densidade de probabilidade que definem os parâmetros incertos através de um processo de geração e redução de cenários. Depois, o método de decomposição de Benders é aplicado para reduzir o esforço computacional necessário para resolver os problemas formulados. Os algoritmos desenvolvidos foram testados em dois sistemas teste IEEE e os resultados mostraram benefícios importantes para a integração de GD e a eficiência energética. / Nowadays, there is a trend to increase the participation of distributed generation (DG) based on renewable energy sources in supplying the global electricity consumption. This trend is being driven mainly by government initiatives to increase energy efficiency, convert the energy use to renewable sources and reduce greenhouse gas emissions. However, as its penetration level increases, the DG can give rise to a system unable to deliver energy reliably and according to quality standards. In this scenario, active network management (ANM) emerges as an alternative for the integration of large amounts of DG. ANM promotes the availability of commercial and regulatory instruments and the provision of distribution networks with automation technologies for procuring ancillary services and flexibility from the DG. ANM requires the development of computational tools to coordinate the implementation of intelligent control schemes, called ANM schemes, in order to optimize the utilization and operation of distribution networks. In this work, optimization models and solution techniques are proposed for ANM considering the integration of solar and wind-based DG and energy efficiency. The first model is developed to determine the maximum capacity of DG that can be allocated in a distribution network when considering the effect of voltage on load efficiency. In the second model, the procedure of controlling the voltage levels to reduce the load demand is implemented for energy saving and for balancing the demand and generation, in a strategy designed for the operation planning of active distribution networks. In both models the uncertainties are considered through two-stage stochastic programming formulations. The ANM schemes considered are the coordinated voltage control through voltage regulators and transformers with on-load tap changer, reactive power support from the DG, and DG generation curtailment. The solution technique involves the discretization of the probability density functions that define the uncertain parameters through a scenario generation and reduction process. Then, the Benders decomposition method is applied in order to reduce the computational effort required to solve the formulated problems. The developed algorithms were tested in two IEEE test systems and the results showed important benefits for the integration of DG and energy efficiency. / 2014/14201-0 e 2015/12911-3

Em diferentes aplicações de Aprendizado de Máquina podemos estar interessados na minimização do valor esperado de certa função de perda. Para a resolução desse problema, Otimização estocástica e Sample Size Selection têm um papel importante. No presente trabalho se apresentam as análises teóricas de alguns algoritmos destas duas áreas, incluindo algumas variações que consideram redução da variância. Nos exemplos práticos pode-se observar a vantagem do método Stochastic Gradient Descent em relação ao tempo de processamento e memória, mas, considerando precisão da solução obtida juntamente com o custo de minimização, as metodologias de redução da variância obtêm as melhores soluções. Os algoritmos Dynamic Sample Size Gradient e Line Search with variable sample size selection apesar de obter soluções melhores que as de Stochastic Gradient Descent, a desvantagem se encontra no alto custo computacional deles. / In different Machine Learnings applications we can be interest in the minimization of the expected value of some loss function. For the resolution of this problem, Stochastic optimization and Sample size selection has an important role. In the present work, it is shown the theoretical analysis of some algorithms of these two areas, including some variations that considers variance reduction. In the practical examples we can observe the advantage of Stochastic Gradient Descent in relation to the processing time and memory, but considering accuracy of the solution obtained and the cost of minimization, the methodologies of variance reduction has the best solutions. In the algorithms Dynamic Sample Size Gradient and Line Search with variable sample size selection, despite of obtaining better solutions than Stochastic Gradient Descent, the disadvantage lies in their high computational cost.

A participação da energia eólica na geração de energia elétrica tem apresentado incremento importante nos últimos anos e a tendência é de representar 11,6% da capacidade instalada brasileira em 2024, segundo a Empresa de Pesquisa Energética (EPE). Hoje, nos modelos de otimização para o despacho das usinas no atendimento da carga de energia do sistema, a energia eólica, assim como as pequenas centrais hidrelétricas e a energia a biomassa, são abatidas da carga de forma determinística, não representando a incerteza na produção dessas usinas. Dada a variabilidade na geração de energia eólica, devido às variações nas velocidades dos ventos e considerando o aumento da participação eólica na matriz de eletricidade brasileira, fato que realça a relevância da fonte, este trabalho desenvolve uma representação estocástica da geração eólica a partir de dados históricos reconstruídos de velocidades de vento de 16 coordenadas do Brasil, em especial das regiões Nordeste e Sul. Os valores de velocidade de vento são transformados em energia eólica através de curvas de potência de turbinas e as usinas eólicas são representadas como se fossem usinas hidráulicas a fio d água no modelo de otimização Newave. A representação do histórico de geração eólica é feita através de vazões nos rios, considerando-se também a expansão no horizonte até 2020. O trabalho tem como base os dados do Newave oficial de agosto de 2016. Com a simulação do modelo considerando-se as séries históricas e sintéticas, o trabalho simula o despacho das usinas, o comportamento dos custos marginais, verificando-se as diferenças no comportamento dessas variáveis quando se utiliza uma representação estocástica para a energia eólica, em comparação com a modelagem determinística utilizada hoje. / Wind power has an increasing share of the Brazilian energy market and has the potential to represent 11.6% of the total capacity by 2024, according to Energy Research Company (EPE). The current optimization models, that dispatch power plants to meet demand, only optimize the demand using hydroelectric and thermal power plants. The remaining sources of generation including wind power, small hydroelectric plants and biomass plants, are not part of the optimization model and are included deterministically. There is variability in wind power generation because of wind speeds variations and considering the increase of the wind power share in the Brazilian electricity matrix, which stresses its importance, this work evaluates a stochastic representation for wind power generation through historical wind speed data of 16 coordinates from the Northeast and South of Brazil. It proposes to introduce wind power plants into the optimization model called Newave by proxy of run-of-river hydropower plants and their inflow. This study also considers wind power expansion in Brazil up to 2020 and the database is the official Newave as of August 2016. This work aims to verify the dispatches of the power plants and the marginal costs, considering the differences between the model used today and the stochastic model presented in the study.

This thesis focuses on optimization on a generic objective function based on reflected Brownian motion (RBM). We investigate in several approaches including the partial differential equation approach where we write our objective function in terms of a Hamilton-Jacobi-Bellman equation using the dynamic programming principle and the gradient descent approach where we use two different gradient estimators. We provide extensive numerical results with the gradient descent approach and we discuss the difficulties and future study opportunities for this problem.

[pt] O desenvolvimento da indústria de Gás Natural pelo mundo resultou em um processo de integração entre os setores de gás natural e eletricidade em diversos países. Entretanto, em alguns casos, como o Brasil, apesar de a demanda de gás para uso convencional (industrial, comercial, residencial, GNV) ter crescido a taxas relativamente altas, ela sozinha ainda não justifica novos grandes investimentos na produção e no transporte de gás. Verifica-se que, neste caso, o setor de energia desempenha um papel indispensável por se tratar do maior mercado potencial de gás natural, com a escala suficiente para ser a âncora de demanda que viabiliza os investimentos em produção e transporte do gás. Todavia, devido à predominância hidrológica no sistema elétrico Brasileiro, o despacho das térmicas é bastante volátil e, por conseqüência, o consumo de gás das térmicas é bastante variável. Assim, o produtor de gás está sujeito a um fluxo de caixa muito volátil e incerto e cláusulas de compra compulsória de gás (takeor-pay) e de remuneração do custo da infra-estrutura (ship-or-pay) são observadas. Enquanto estas cláusulas trazem certeza necessária para viabilizar a produção, elas oneram excessivamente os custos das Usinas Térmicas, que se vêem obrigadas a pagar pelo combustível e, portanto, gerar, mesmo quando o preço da energia esteja inferior ao seu custo marginal de produção. Tendo em vista este cenário, foi recentemente discutida no âmbito do Governo Federal a criação de um mercado flexível de gás natural, onde contratos interruptiveis de gás (lastreados no take-or-pay das térmicas) seriam fornecidos a consumidores industriais. Nestes contratos, o fornecimento seria interrompido se a Usina Térmica fosse despachada. O objetivo desta tese é analisar a criação deste mercado sob a ótica dos consumidores. Será verificada a disposição a pagar por um contrato interruptível de gás levando em consideração a incerteza associada ao suprimento (que depende da prioridade de uso do gás pelas térmicas) e o perfil de risco destes consumidores. / [en] With the development of the gas industry worldwide, a process of strengthening the integration between the natural gas and the electricity sectors is underway in several countries. However, although gas demand has been growing at relatively high rates, this demand growth solely is unlikely to justify new large investments in gas production and transportation. This means that the power sector ends up being the largest potential market for natural gas, with the needed scale to provide the necessary anchor demand to spur these production and infrastructure investments. The hydro predominance in the country creates volatility on the dispatch of the gas-fired plants, which ends up creating an undesirable (from the gas-sector point of view) volatility in the natural gas consumption. Since the gas-market is still incipient, gas contracts are typically of long-term with high take or pay and ship or pay clauses to ensure financing of the production- transportation infrastructure. From the power sector point of view, these clauses are undesirable: due to the uncertainty of dispatch gas-based generators want to negotiate a higher flexibility. As such, the aim of this work is to determine the impacts due to the creation of a flexible (secondary) gas market from the costumers´ point of view. It will be also developed the costumers´ willto- contract curve, which will take into account the uncertainty of thermoelectric dispatch (that rules the gas availability over this new proposed market) and the risk-profile of costumers.

This thesis is concerned with identifying the best decision among a set of possible decisions in the presence of uncertainty. We are primarily interested in situations where the objective function value at any feasible solution needs to be estimated, for example via a ``black-box' simulation procedure. We develop adaptive random search methods for solving such simulation optimization problems. The methods are adaptive in the sense that they use information gathered during previous iterations to decide how simulation effort is expended in the current iteration. We consider random search because such methods assume very little about the structure of the underlying problem, and hence can be applied to solve complex simulation optimization problems with little expertise required from an end-user. Consequently, such methods are suitable for inclusion in simulation software. We first identify desirable features that algorithms for discrete simulation optimization need to possess to exhibit attractive empirical performance. Our approach emphasizes maintaining an appropriate balance between exploration, exploitation, and estimation. We also present two new and almost surely convergent random search methods that possess these desirable features and demonstrate their empirical attractiveness. Second, we develop two frameworks for designing adaptive and almost surely convergent random search methods for discrete simulation optimization. Our frameworks involve averaging, in that all decisions that require estimates of the objective function values at various feasible solutions are based on the averages of all observations collected at these solutions so far. We present two new and almost surely convergent variants of simulated annealing and demonstrate the empirical effectiveness of averaging and adaptivity in the context of simulated annealing. Finally, we present three random search methods for solving simulation optimization problems with uncountable feasible regions. One of the approaches is adaptive, while the other two are based on pure random search. We provide conditions under which the three methods are convergent, both in probability and almost surely. Lastly, we include a computational study that demonstrates the effectiveness of the methods when compared to some other approaches available in the literature.

Les Algorithmes Évolutionnaires (AEs) ont été très étudiés en raison de leur capacité à résoudre des problèmes d'optimisation complexes en utilisant des opérateurs de variation adaptés à des problèmes spécifiques. Une recherche dirigée par une population de solutions offre une bonne robustesse par rapport à un bruit modéré et la multi-modalité de la fonction optimisée, contrairement à d'autres méthodes d'optimisation classiques telles que les méthodes de quasi-Newton. La principale limitation de AEs, le grand nombre d'évaluations de la fonction objectif, pénalise toutefois l'usage des AEs pour l'optimisation de fonctions chères en temps calcul. La présente thèse se concentre sur un algorithme évolutionnaire, Covariance Matrix Adaptation Evolution Strategy (CMA-ES), connu comme un algorithme puissant pour l'optimisation continue boîte noire. Nous présentons l'état de l'art des algorithmes, dérivés de CMA-ES, pour résoudre les problèmes d'optimisation mono- et multi-objectifs dans le scénario boîte noire. Une première contribution, visant l'optimisation de fonctions coûteuses, concerne l'approximation scalaire de la fonction objectif. Le meta-modèle appris respecte l'ordre des solutions (induit par la valeur de la fonction objectif pour ces solutions) ; il est ainsi invariant par transformation monotone de la fonction objectif. L'algorithme ainsi défini, saACM-ES, intègre étroitement l'optimisation réalisée par CMA-ES et l'apprentissage statistique de meta-modèles adaptatifs ; en particulier les meta-modèles reposent sur la matrice de covariance adaptée par CMA-ES. saACM-ES préserve ainsi les deux propriété clé d'invariance de CMA-ES~: invariance i) par rapport aux transformations monotones de la fonction objectif; et ii) par rapport aux transformations orthogonales de l'espace de recherche. L'approche est étendue au cadre de l'optimisation multi-objectifs, en proposant deux types de meta-modèles (scalaires). La première repose sur la caractérisation du front de Pareto courant (utilisant une variante mixte de One Class Support Vector Machone (SVM) pour les points dominés et de Regression SVM pour les points non-dominés). La seconde repose sur l'apprentissage d'ordre des solutions (rang de Pareto) des solutions. Ces deux approches sont intégrées à CMA-ES pour l'optimisation multi-objectif (MO-CMA-ES) et nous discutons quelques aspects de l'exploitation de meta-modèles dans le contexte de l'optimisation multi-objectif. Une seconde contribution concerne la conception d'algorithmes nouveaux pour l'optimi\-sation mono-objectif, multi-objectifs et multi-modale, développés pour comprendre, explorer et élargir les frontières du domaine des algorithmes évolutionnaires et CMA-ES en particulier. Spécifiquement, l'adaptation du système de coordonnées proposée par CMA-ES est couplée à une méthode adaptative de descente coordonnée par coordonnée. Une stratégie adaptative de redémarrage de CMA-ES est proposée pour l'optimisation multi-modale. Enfin, des stratégies de sélection adaptées aux cas de l'optimisation multi-objectifs et remédiant aux difficultés rencontrées par MO-CMA-ES sont proposées.

Stochastic and data-distributed optimization algorithms have received lots of attention from the machine learning community due to the tremendous demand from the large-scale learning and the big-data related optimization. A lot of stochastic and deterministic learning algorithms are proposed recently under various application scenarios. Nevertheless, many of these algorithms are based on heuristics and their optimality in terms of the generalization error is not sufficiently justified. In this talk, I will explain the concept of an optimal learning algorithm, and show that given a time budget and proper hypothesis space, only those achieving the lower bounds of the estimation error and the optimization error are optimal. Guided by this concept, we investigated the stochastic minimization of nonsmooth convex loss functions, a central problem in machine learning. We proposed a novel algorithm named Accelerated Nonsmooth Stochastic Gradient Descent, which exploits the structure of common nonsmooth loss functions to achieve optimal convergence rates for a class of problems including SVMs. It is the first stochastic algorithm that can achieve the optimal O(1/t) rate for minimizing nonsmooth loss functions. The fast rates are confirmed by empirical comparisons with state-of-the-art algorithms including the averaged SGD. The Alternating Direction Method of Multipliers (ADMM) is another flexible method to explore function structures. In the second part we proposed stochastic ADMM that can be applied to a general class of convex and nonsmooth functions, beyond the smooth and separable least squares loss used in lasso. We also demonstrate the rates of convergence for our algorithm under various structural assumptions of the stochastic function: O(1/sqrt{t}) for convex functions and O(log t/t) for strongly convex functions. A novel application named Graph-Guided SVM is proposed to demonstrate the usefulness of our algorithm. We also extend the scalability of stochastic algorithms to nonlinear kernel machines, where the problem is formulated as a constrained dual quadratic optimization. The simplex constraint can be handled by the classic Frank-Wolfe method. The proposed stochastic Frank-Wolfe methods achieve comparable or even better accuracies than state-of-the-art batch and online kernel SVM solvers, and are significantly faster. The last part investigates the problem of data-distributed learning. We formulate it as a consensus-constrained optimization problem and solve it with ADMM. It turns out that the underlying communication topology is a key factor in achieving a balance between a fast learning rate and computation resource consumption. We analyze the linear convergence behavior of consensus ADMM so as to characterize the interplay between the communication topology and the penalty parameters used in ADMM. We observe that given optimal parameters, the complete bipartite and the master-slave graphs exhibit the fastest convergence, followed by bi-regular graphs.

This thesis studies two related topics in liner shipping. The first topic is the contract pricing problem for container carriers. The second part studies the interaction of the longer term contracts and the spot markets/exchanges for the same goods/services. Most containerized freight is transported under the provisions of medium term contracts between ocean carriers and shippers. One of the biggest challenges for an ocean carrier is to find optimal ways to structure the prices in those contracts. In particular, an ocean carrier would like to set the prices such that the best match between supply and demand can be obtained to maximize its profit. We propose three optimization models as decision tools that carriers can use to plan the contract price structures, as well as the anticipated freight flows and empty container flows for the period covered by the contracts. Based on the models, we propose algorithms and build decision tools that generate the following output: optimal prices to be charged for the movement of freight, the anticipated freight flows and empty flows, containers to be leased, rented and purchased, and the additional voyage capacities to be procured. The first two models are deterministic and represent the problem at different levels of detail. In addition, a three-stage stochastic model is proposed to handle uncertainties in demand rates, costs, bookings and transit times on feeder arcs. Recent developments in information technology and communication make spot transactions more economical and more convenient. Nevertheless, the incidental spot transactions still count for only a very small portion of freight transported both by the large carriers who are the leaders in implementing e-commerce and in the industry as a whole. The second part of the thesis studies models to provide insight into the effect of spot market participation rates on various economic quantities. This may have implications for freight transportation industries, such as the sea cargo industry, in which longer term contracts are still prevalent. We focus our study on the following situation. Option contracts are signed before the demand is observed. As is common in liner shipping, sellers (carriers) also sell goods/services on the spot. Buyers (shippers) may or may not buy in the spot market as a matter of policy. We investigate the effects of spot market participation on the contract market and on the surpluses of all market players. It is found that the contract market shrinks as more and more buyers participate in the spot market. However, the effects on the surpluses of different market players are much more complicated and depend on the following factors: market structure, demand variation along time, demand variation among buyers and capacity level.

Orientador: Antonio Padilha Feltrin / Resumo: Atualmente, existe uma tendência para aumentar a participação da Geração Distribuída (GD) baseada em Fontes de Energia Renováveis (FER) no suprimento do consumo global de energia elétrica. Esta tendência está sendo impulsionada principalmente por iniciativas governamentais destinadas a aumentar a eficiência energética, aumentar o uso da energia proveniente das FER e reduzir as emissões de gases de efeito estufa. No entanto, à medida que seu nível de penetração aumenta, a GD pode dar origem a um sistema incapaz de fornecer energia de forma confiável e de acordo com os padrões de qualidade. Nesse cenário, a Gestão de Redes Ativas (GRA) surge como uma alternativa para a integração de grandes montantes de GD. A GRA promove a disponibilização de instrumentos comerciais e regulatórios, e o fornecimento das redes de distribuição com tecnologias de automação para procurar serviços ancilares e flexibilidade a partir da GD. A GRA requer o desenvolvimento de ferramentas computacionais para coordenar a implementação de esquemas de controle inteligentes, chamados de esquemas de GRA, a fim de otimizar a utilização e operação das redes. Neste trabalho, são propostos modelos de otimização e técnicas de solução para a GRA considerando a integração de GD solar fotovoltaica e eólica e a eficiência energética. O primeiro modelo é desenvolvido para determinar a capacidade máxima de GD que pode ser alocada em uma rede de distribuição quando se considera o efeito da tensão na eficiência das cargas.... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: Nowadays, there is a trend to increase the participation of distributed generation (DG) based on renewable energy sources in supplying the global electricity consumption. This trend is being driven mainly by government initiatives to increase energy efficiency, convert the energy use to renewable sources and reduce greenhouse gas emissions. However, as its penetration level increases, the DG can give rise to a system unable to deliver energy reliably and according to quality standards. In this scenario, active network management (ANM) emerges as an alternative for the integration of large amounts of DG. ANM promotes the availability of commercial and regulatory instruments and the provision of distribution networks with automation technologies for procuring ancillary services and flexibility from the DG. ANM requires the development of computational tools to coordinate the implementation of intelligent control schemes, called ANM schemes, in order to optimize the utilization and operation of distribution networks. In this work, optimization models and solution techniques are proposed for ANM considering the integration of solar and wind-based DG and energy efficiency. The first model is developed to determine the maximum capacity of DG that can be allocated in a distribution network when considering the effect of voltage on load efficiency. In the second model, the procedure of controlling the voltage levels to reduce the load demand is implemented for energy saving and for b... (Complete abstract click electronic access below) / Doutor