The main topic of this research is the efficient operation of a modernized distribution grid from both the customer side and utility side. For the customer side, this dissertation discusses the planning and operation of a customer with multiple demand response programs, energy storage systems and distributed generators; for the utility side, this dissertation addresses the implementation and assessment of voltage/VAR control and conservation voltage reduction in a distribution grid with distributed generators. The objectives of this research are as follows: (1) to develop methods to assist customers to select appropriate demand response programs considering the integration of energy storage systems and DGs, and perform corresponding energy management including dispatches of loads, energy storage systems, and DGs; (2) to develop stochastic voltage/VAR control techniques for distribution grids with renewable DGs; (3) to develop optimization and validation methods for the planning of integration of renewable DGs to assist the implementation of voltage/VAR control; and (4) to develop techniques to assess load-reduction effects of voltage/VAR control and conservation voltage reduction. In this dissertation, a two-stage co-optimization method for the planning and energy management of a customer with demand response programs is proposed. The first level is to optimally select suitable demand response programs to join and integrate batteries, and the second level is to schedule the dispatches of loads, batteries and fossil-fired backup generators. The proposed method considers various demand response programs, demand scenarios and customer types. It can provide guidance to a customer to make the most beneficial decisions in an electricity market with multiple demand response programs. For the implementation of voltage/VAR control, this dissertation proposes a stochastic rolling horizon optimization-based method to conduct optimal dispatches of voltage/VAR control devices such as on-load tap changers and capacitor banks. The uncertainties of renewable DG output are taken into account by the stochastic formulation and the generated scenarios. The exponential load models are applied to capture the load behaviors of various types of customers. A new method to simultaneously consider the integration of DGs and the implementation of voltage/VAR control is also developed. The proposed method includes both solution and validation stages. The planning problem is formulated as a bi-level stochastic program. The solution stage is based on sample average approximation (SAA), and the validation stage is based on multiple replication procedure (MRP) to test the robustness of the sample average approximation solutions of the stochastic program. This research applies big data-driven analytics and load modeling techniques to propose two novel methodologies to assess the load-reduction effects of conservation voltage reduction. The proposed methods can be used to assist utilities to select preferable feeders to implement conservation voltage reduction.

This thesis investigates how estimates of reservoir compartmentalization impact facility sizing decisions and the degree to which inaccurate estimates destroy project value. An uncertainty analysis workflow is proposed and an asset development optimization model is specified to simulate the decision making process during concept selection. The model endogenizes drilling decisions and includes a real option to expand facility capacity after the uncertain variables are realized. The value of information analysis suggests that cost of erroneous estimates of reservoir compartmentalization is significant and can reduce asset value by more than 30%. We also find that the negative impacts are larger when the degree of compartmentalization is underestimated (too optimistic) than when it is overestimated (too pessimistic). / text

Reservoir systems are subject to several uncertainties that are the result of imperfect knowledge about system behavior and inputs. A major source of uncertainty arises from the inability to predict future inflows. Fortunately, it is often possible to generate probabilistic forecasts of inflow volumes in the form of probability density functions or ensembles. These inflow forecasts can be coupled with stochastic management models to determine reservoir release policies and provide stakeholders with meaningful information of upcoming system responses such as reservoir levels, releases, flood damage risks, hydropower production, water supply withdrawals, water quality conditions, navigation opportunities, and environmental flows, among others. This information on anticipated system responses is also expressed in the form of forecasts that must reliably represent the actual system behavior when it eventually occurs. The first part of this study presents an assessment methodology that can be used to determine the consistency of ensemble forecasts through the use of relative frequency histograms and minimum spanning trees (MST). This methodology is then used to assess a management model's ability to produce reliable ensemble forecasts. It was found that neglecting to account for hydrologic state variables and improperly modeling the finite management horizon decrease ensemble consistency. Several extensions to the existing management model are also developed and evaluated. The second portion of this study involves the management of the uncertainties in reservoir systems. Traditional management models only find management policies that optimize the expected values of system benefits or costs, thereby not allowing operators and stakeholders to explicitly explore issues related to uncertainty and risk management. A technique that can be used to derive management policies that produce desired probabilistic distributions of reservoir system outputs reflecting stakeholder preferences is developed. This technique can be embedded in a user-interactive framework that can be employed to evaluate the trade-offs and build consensus in multi-objective and multi-stakeholder systems. The methods developed in this dissertation are illustrated in case studies of real reservoir systems, including a seven-reservoir, multi-objective system in California's Central Valley.

No description available.

In this thesis, we provide computationally efficient algorithms with provable statistical guarantees, for the problem of active learning, by using ideas from sequential analysis. We provide a generic algorithmic framework for active learning in the pool setting, and instantiate this framework by using ideas from learning with experts, stochastic optimization, and multi-armed bandits. For the problem of learning convex combination of a given set of hypothesis, we provide a stochastic mirror descent based active learning algorithm in the stream setting.

The subject of this thesis is to study approximation algorithms for some network design problems in face of uncertainty. We consider two widely studied models of handling uncertainties - Robust Optimization and Stochastic Optimization. We study a robust version of the well studied Uncapacitated Facility Location Problem (UFLP). In this version, once the set of facilities to be opened is decided, an adversary may close at most β facilities. The clients must then be assigned to the remaining open facilities. The performance of a solution is measured by the worst possible set of facilities that the adversary may close. We introduce a novel LP for the problem, and provide an LP rounding algorithm when all facilities have same opening costs. We also study the 2-stage Stochastic version of the Steiner Tree Problem. In this version, the set of terminals to be covered is not known in advance. Instead, a probability distribution over the possible sets of terminals is known. One is allowed to build a partial solution in the first stage a low cost, and when the exact scenario to be covered becomes known in the second stage, one is allowed to extend the solution by building a recourse network, albeit at higher cost. The aim is to construct a solution of low cost in expectation. We provide an LP rounding algorithm for this problem that beats the current best known LP rounding based approximation algorithm.

[pt] A busca por uma matriz limpa de geração de energia vem incentivando a expansão de fontes alternativas de geração de energia ao redor do mundo. No Brasil, Pequenas Centrais Hidroelétricas (PCHs) e Usinas a Biomassa de Cana de Açúcar (Biomassa) vêm se mostrando alternativas atraentes nos últimos anos. No entanto, ambas as tecnologias são caracterizadas por perfis de geração sazonais (mas complementares). Este fato gera riscos que por muitas vezes inviabilizam a comercialização de maneira individual da energia produzida por essas usinas. As PCHs, em particular, têm uma opção de mitigação de parte desse risco participando do Mecanismo de Realocação de Energia (MRE). O MRE traz às PCHs a flexibilidade de sazonalizar sua Garantia Física ao longo do ano, o que se mostra uma ferramenta adicional para mitigar o risco da sazonalidade da geração hidráulica no Brasil. Neste trabalho, será estudado como a combinação de PCHs e Biomassas em um mesmo portfólio pode trazer ganhos sinérgicos para os Geradores. Em particular, será estudado como essa combinação altera a estratégia de sazonalização da Garantia Física da PCH participante do MRE e como essa sazonalização diferenciada resulta em benefícios para os geradores. Para isto, será proposto um modelo de otimização estocástica utilizado para simular o processo decisório de como sazonalizar a Garantia Física de PCHs combinadas com Biomassas em uma proporção fixa ou no contexto de otimização de portfólios compostos por estes dois tipos de usinas. Serão apresentados estudos de caso mostrando diferentes estratégias de comercialização de energia por parte destes Geradores e como a decisão de sazonalização da Garantia Física da PCH se comporta em cada um desses casos. / [en] The search for clean energy development has motivated the expansion of renewable sources of generation around the world. In Brazil, Small Hydro Plants (SHP) and Cogenaration from Sugarcane waste (Biomass) have proven themselves to be attractive alternatives during the last years. Nevertheless, both tecnologies have seazonal (yet complementary) availability. This fact results in financial risks that can make the commercialization of these plants energy individually too risky. SHPs have the option of mitigating their risk by joining the Energy Realocation Mecanism (ERM). The ERM, additionally, gives the SHPs the flexibility of allocate its firm energy in different manners along the year, which can be a valuable tool in mitigating the risks due to the seasonal availability of these plants. In this work, the combination of SHPs and Biomass in a single portfolio will be studied as a tool to mitigate the risks each plant faces individually. In particular, we will study the impact that this combination has over the decision process of SHPs on how to allocate their firm energy and how this different allocation can prove to be beneficial to both generators. In order to do so, a stochastic optimization model will be proposed to simulate the decision process of the SHPs on how to allocate its firm energy when combined in a portfolio with a Biomass in a fixed proportion or in the context of portfolio optimization. Case studies will be presented showing different strategies of commercialization by these generators and how the firm energy allocation decision by the SHP changes in each case.

[pt] A natureza incerta e os altos incentivos econômicos do negócio de refino são forças motrizes para melhorias nos processos de planejamento das refinarias. Decisões tomadas na cadeia do petróleo diferem principalmente na gama de atividades (integração espacial) e no horizonte de planejamento (integração temporal). O objetivo desta tese é abordar o problema da integração da cadeia do petróleo sob incerteza em diferentes níveis de decisão. Modelos de programação matemática tático e operacional são propostos. O modelo tático maximiza o lucro esperado da cadeia de suprimentos e aloca metas de produção para as refinarias considerando restrições logísticas. O modelo operacional maximiza o lucro esperado de cada refinaria determinando a quantidade de material processada por unidade de processo em um dado período. Ambos os modelos são lineares estocásticos de dois estágios, onde a incerteza é incorporada nos parâmetros dominantes de cada nível (preço e demanda no nível tático e suprimento de petróleo e capacidade das unidades no nível operacional). A integração espacial é discutida no nível tático (considerando a cadeia de suprimentos), enquanto a integração temporal é discutida na interação entre os dois níveis. Duas abordagens de integração temporal são consideradas: hierárquica, onde o fluxo de informações é somente do modelo tático para o operacional, e iterativa, onde há retorno do nível operacional para o tático. Um estudo de escala industrial foi conduzido para demonstrar os benefícios da integração em ambiente estocástico. Resultados são oferecidos no contexto de um estudo usando dados da indústria brasileira do petróleo para demonstrar a eficácia das abordagens propostas. / [en] The uncertain nature and high economic incentives of the refining business are driving forces for improvements in the refinery planning process. Decisions made at the oil chain differ mainly in the range of activities (spatial integration) and planning horizon (temporal integration). This thesis purpose is to address the problem of the oil chain integration under uncertainty at different decision levels. Tactical and operational mathematical programming models are proposed. The tactical model maximizes the expected profit of the supply chain and allocates the production targets to refineries taking logistics constraints into account. The operational model maximizes the expected profit of each refinery determining the amount of material that is processed at each process unit in a given period. Both models are two-stage stochastic linear programs where uncertainty is incorporated in the dominant random parameters at each level (price and demand at the tactical level and oil supply and process capacity unit at the operational level).Spatial integration is discussed at the tactical level (considering supply chain), whereas the temporal integration is discussed in the interaction between the two levels. Two temporal integration approaches are considered: hierarchical, where the flow of information is only from the tactical to the operational model, and iterative, where there is feedback from the tactical to the operational model. An industrial scale study was conducted to discuss the benefits of integration in a stochastic environment. Results are offered in the context of a study using data from the Brazilian oil industry to demonstrate the effectiveness of the proposed approaches.

[pt] Nos últimos anos, nota-se uma forte tendência no Brasil de oferta de petróleos cada vez mais pesados e ácidos em contraposição a uma crescente demanda de derivados mais leves dentro de especificações mais rígidas. Dessa forma, o Brasil se depara com a necessidade em adaptar suas refinarias e rede logística a esse novo perfil. Nesse contexto é importante a avaliação da cadeia integrada de petróleo e derivados no longo prazo, visando auxiliar a tomada de decisão em relação aos projetos que devem ser considerados na carteira de investimentos. Por se tratar de uma decisão de longo prazo, é importante levar em consideração as incertezas relacionadas aos parâmetros considerados, como: oferta e preço de petróleos, demanda e preço de derivados e outros. Assim, tornase possível a avaliação de uma carteira de projetos de investimentos considerando os riscos existentes. Este trabalho propõe apresentar uma metodologia de otimização sob incerteza, que utilize programação estocástica em conjunto com técnicas de otimização de portfólio, aplicada ao estudo de uma carteira de investimentos na área de abastecimento de petróleo. O estudo é focado em um modelo de programação linear que maximiza o resultado presente líquido esperado ao longo de um horizonte de tempo estipulado, dado um nível de risco aceitável. Foram propostas duas abordagens de medida de risco: Conditional Value-at-Risk (CVaR) e Minimax. A partir dos resultados numéricos, ficou comprovado que a decisão otimizada de investimento na área de petróleo e derivados apresenta variação com o nível de risco que se pretende assumir. / [en] Over the last years, a strong trade-off between crude oil offer and oil product demand has been posed in Brazil: while the oil produced in Brazil is getting heavier, its` products must be light, constrained by rigid specifications. Hence, the country needs to adapt its refineries and logistic network to this new profile. In this context, a long term analysis of the integrated oil chain is a relevant task. This analysis helps the decision maker to choose projects that should be considered in portfolio investment. During the decision process, it is important to take into account uncertainties related to some parameters: crude oil prices, crude oil offer, product prices, expected demand and others. By doing that, it is possible for the analyst to evaluate a project portfolio considering risks. The present work proposes a methodology for optimization under uncertainty, applied to the study of a portfolio investment for the downstream oil industry, employing both stochastic programming and portfolio optimization techniques. The study is focused on a linear programming model that maximizes the expected net present value (NPV) along the specified time horizon and risk level. Two approaches have been proposed to measure risk: Conditional Value-at-Risk (CVaR) and Minimax. The results show that the investment choice in the oil chain varies with the imposed risk level.

abstract: In this dissertation, I propose potential techniques to improve the quality-of-service (QoS) of real-time applications in cognitive radio (CR) systems. Unlike best-effort applications, real-time applications, such as audio and video, have a QoS that need to be met. There are two different frameworks that are used to study the QoS in the literature, namely, the average-delay and the hard-deadline frameworks. In the former, the scheduling algorithm has to guarantee that the packet's average delay is below a prespecified threshold while the latter imposes a hard deadline on each packet in the system. In this dissertation, I present joint power allocation and scheduling algorithms for each framework and show their applications in CR systems which are known to have strict power limitations so as to protect the licensed users from interference. A common aspect of the two frameworks is the packet service time. Thus, the effect of multiple channels on the service time is studied first. The problem is formulated as an optimal stopping rule problem where it is required to decide at which channel the SU should stop sensing and begin transmission. I provide a closed-form expression for this optimal stopping rule and the optimal transmission power of secondary user (SU). The average-delay framework is then presented in a single CR channel system with a base station (BS) that schedules the SUs to minimize the average delay while protecting the primary users (PUs) from harmful interference. One of the contributions of the proposed algorithm is its suitability for heterogeneous-channels systems where users with statistically low channel quality suffer worse delay performances. The proposed algorithm guarantees the prespecified delay performance to each SU without violating the PU's interference constraint. Finally, in the hard-deadline framework, I propose three algorithms that maximize the system's throughput while guaranteeing the required percentage of packets to be transmitted by their deadlines. The proposed algorithms work in heterogeneous systems where the BS is serving different types of users having real-time (RT) data and non-real-time (NRT) data. I show that two of the proposed algorithms have the low complexity where the power policies of both the RT and NRT users are in closed-form expressions and a low-complexity scheduler. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2016