Στην παρούσα διπλωματική εργασία θα ασχοληθούμε με την αξιολόγηση της προστασίας σε κατανεμημένη παραγωγή. Γενικά η κατανεμημένη παραγωγή, καλύπτει ένα μεγάλο εύρος νέων και παραδοσιακών τεχνολογιών με μικρές μονάδες εγκατεστημένες κοντά στην κατανάλωση και πιστεύεται ότι θα συμβάλει στην κάλυψη ενός μεγάλου μέρους της ενεργειακής ζήτησης στα επόμενα χρόνια με τη δυναμική της είσοδο στην απελευθερωμένη αγορά ενέργειας. Αρχικά γίνεται μια ανασκόπηση της υφιστάμενης παγκόσμιας ενεργειακής κατάστασης. Αναλύονται οι κυριότερες τεχνολογίες παραγωγής ηλεκτρικής ενέργειας και τα καύσιμα που χρησιμοποιούνται, τα οποία κατά συντριπτική πλειοψηφία βασίζονται στους ορυκτούς πόρους. Αποτυπώνεται αφενός, η συνολική συνεισφορά αυτών στην ενεργειακή ζήτηση, αφετέρου τα παγκόσμια αποθέματα αυτών. Στη συνέχεια παρουσιάζεται η υφιστάμενη κατάσταση στον ηλεκτρικό τομέα, οι αλλαγές που συνεπάγονται με τη απελευθέρωση της αγοράς ενέργειας, καθώς και το γενικό πλαίσιο και οι ρυθμίσεις που διέπουν τον τομέα της ενέργειας με βάση τα νέα δεδομένα. Γίνεται εισαγωγή και ορισμός της κατανεμημένης παραγωγής και αναλύονται τα βασικά χαρακτηριστικά που συνθέτουν την έννοια αυτής καθώς επίσης γίνεται μια παρουσίαση των κυριότερων τεχνολογιών που ολοκληρώνουν τη κατανεμημένη παραγωγή. Στη συνέχεια ασχολούμαστε με το φαινόμενο της νησιδοποίησης. Η αύξηση της παραγωγής ισχύος από διανεμημένους παραγωγούς καθιστά την εμφάνιση του φαινομένου όλο και συχνότερη. Η ανίχνευση και διακοπή του φαινομένου είναι καθοριστικής σημασίας για την ορθή λειτουργία του Δικτύου και την ασφάλεια του προσωπικού που εργάζεται σε αυτό. Εδώ περιγράφουμε μερικές από τις πιο βασικές εφαρμοζόμενες μεθόδους ανίχνευσης και διακοπής του. Ύστερα παρουσιάζουμε τα βασικά χαρακτηριστικά των μέσων προστασίας έναντι υπερεντάσεων. Προχωράμε σε μελέτη των διατάξεων προστασίας ενός παραδείγματος μιας εγκατάστασης φωτοβολταϊκών με σκοπό τη διασφάλιση της σε περίπτωση εμφάνισης σφαλμάτων. Τέλος αναφερόμαστε σε κάποιες νέες προσεγγίσεις και μελλοντικές τάσεις και προοπτικές για τον καθορισμό των τρόπων ενσωμάτωσης ομαλών τεχνολογιών ΚΠ και τη ρύθμιση θεμάτων σχετικά με τη διασύνδεση και τη λειτουργία τους. / In this work we will do a research about the appraisal of the distributed generation .All in all the distributed generation encloses a great range of new and traditional technologies. Some of them are installed close to the consumption and it is believed that the distributed generation will contribute to cover a great part of the energy demand in the next years. In the beginning we refer to the global energy situation. We analyze the most important technologies of distributed generation and the fuels that are used which are based on the mineral resources. These fuels contribute in the energy demand and we also refer to the global store of fuel. We present the situation of the electric section today, the changes and the regulations that happen in this section day by day. What is more, it is given the meaning of the “distributed generation”, we analyze the basic characteristics of it and we present the most important technologies that complete the distributed generation. Furthermore we refer to the problem of islanding. Because of the fact that the power production is increasing lately, the problem is appeared more and more often. We must look for this phenomenon and stop it immediately in order the network work correctly and the persons who work in this part of network be safe. We describe some of the most important methods of “tracking” the islanding and stopping it. Afterwards we present the basic characteristics of the protection arrangement that we must use in case of errors and short circuits. We continue presenting an instance for a photovoltaic system installation and what sort of protection must be used in order this system to work with safe when errors happen. In the end we refer to what must be done in the future so that the technologies of distributed generation be used more and more often and much more efficiently.

Made available in DSpace on 2014-06-11T19:30:50Z (GMT). No. of bitstreams: 0 Previous issue date: 2009-05-25Bitstream added on 2014-06-13T21:01:23Z : No. of bitstreams: 1 falavinhajunior_jn_dr_ilha.pdf: 3487083 bytes, checksum: 5eeeb56b23091b46b46acaafba4babe4 (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Em sistemas distribuídos de larga escala; onde os recursos compartilhados são de propriedade de entidades distintas; existe a necessidade de refletir o fator propriedade dos recursos no processo de escalonamento de tarefas e alocação de recursos. Um sistema de gerenciamento de recursos apropriado deve garantir que os proprietários de recursos tenham acesso aos seus recursos ou ao menos a uma parcela de recursos que seja equivalente a eles. Diferentes políticas podem ser estabelecidas para que o sistema garanta esse direito aos proprietários de recursos; e nessa tese defende-se uma política de escalonamento e alocação de reucrsos chamada Owner-Share Enforcement Policy (OSEP) ou Política de Garantia da Porção do Proprietário; que tem por objetivo garantir o direito de acesso aos recursos através de um sistema de escalonamento baseado em preempção de tarefas e realocação de recursos. Avalia-se a política através da análise de testes e resultados envolvendo métricas de desempenho que descrevem fatores como violação da política; perdada capacidade de processamento; custo da política e satisfação do usuário. Os testes ainda envolveram a análise de desempenho da política em ambientes com a possibilidade de chekcpointing de tarefas; minimizando assim o desperdício de processamento. Fez-se ainda comparações com a política de compartilhamento justo Fair-Share; que permitiram estabelecer as vantagens e desvantagens de cada política e ainda identificar futuros problemas. Por fim; conclui-se a tese identificando as contribuições oferecidas por este trabalho e os trabalhos futuros que podem ser desenvolvidos. / In large distributed systems, where shared resources are owned by distinct entities, there is a need to reflect resource ownership in resource allocation. An appropriate resource management system should guarantee that owners of resources have access to their resources or at least to a share of resources proportional to the share of resources they provide. Different policies can be established for guaranteeing the access to resources, and in this thesis we introduce a policy for scheduling and resource allocation named Owner Share Enforcement Policy (OSEP). This policy is based on the concept of distributed ownership and itguarantees the owner's right of accessing their share of resources in a distributed system with a preemptive share space. We evaluate this policy through tests and results analysis involving performance metrics that describe policy violation, loss of capacity, policy cost and user satisfaction. The tests were also conducted in environments withand without job checkpointing, and comparisons with the Fair-Share scheduling policy were made in order to capture the trade-offs of each policy. Finally, we conclude the thesis describing the contributions achieved with this work and pointing directions for future work.

Many real-life problems such as distributed meeting scheduling, mobile frequency allocation and resource allocation can be solved using multi-agent paradigms. Distributed constraint satisfaction problems (DisCSPs) is a framework for describing such problems in terms of related subproblems, called a complex local problem (CLP), which are dispersed over a number of locations, each with its own constraints on the values their variables can take. An agent knows the variables in its CLP plus the variables (and their current value) which are directly related to one of its own variables and the constraints relating them. It knows little about the rest of the problem. Thus, each CLP is solved by an agent which cooperates with other agents to solve the overall problem. Algorithms for solving DisCSPs can be classified as either systematic or local search with the former being complete and the latter incomplete. The algorithms generally assume that each agent has only one variable as they can solve DisCSP with CLPs using “virtual” agents. However, in large DisCSPs where it is appropriate to trade completeness off against timeliness, systematic search algorithms can be expensive when compared to local search algorithms which generally converge quicker to a solution (if a solution is found) when compared to systematic algorithms. A major drawback of local search algorithms is getting stuck at local optima. Significant researches have focused on heuristics which can be used in an attempt to either escape or avoid local optima. This thesis makes significant contributions to local search algorithms for DisCSPs. Firstly, we present a novel combination of heuristics in DynAPP (Dynamic Agent Prioritisation with Penalties), which is a distributed synchronous local search algorithm for solving DisCSPs having one variable per agent. DynAPP combines penalties on values and dynamic agent prioritisation heuristics to escape local optima. Secondly, we develop a divide and conquer approach that handles DisCSP with CLPs by exploiting the structure of the problem. The divide and conquer approach prioritises the finding of variable instantiations which satisfy the constraints between agents which are often more expensive to satisfy when compared to constraints within an agent. The approach also exploits concurrency and combines the following search strategies: (i) both systematic and local searches; (ii) both centralised and distributed searches; and (iii) a modified compilation strategy. We also present an algorithm that implements the divide and conquer approach in Multi-DCA (Divide and Conquer Algorithm for Agents with CLPs). DynAPP and Multi-DCA were evaluated on several benchmark problems and compared to the leading algorithms for DisCSPs and DisCSPs with CLPs respectively. The results show that at the region of difficult problems, combining search heuristics and exploiting problem structure in distributed constraint satisfaction achieve significant benefits (i.e. generally used less computational time and communication costs) over existing competing methods.

In recent years, many machine learning algorithms have been developed and widely applied in various applications. However, most of them have considered the data distributions of the training and test datasets to be similar. This thesis concerns on the decrease of generalization ability in a test dataset when the data distribution is different from that of the training dataset. As labels may be unavailable in the test dataset in practical applications, we follow the effective approach of unsupervised domain adaptation and propose distribution alignment methods to improve the generalization ability of models learned from the training dataset in the test dataset. To solve the problem of joint distribution alignment without target labels, we propose a new criterion of domain-shared group sparsity that is an equivalent condition for equal conditional distribution. A domain-shared group-sparse dictionary learning model is built with the proposed criterion, and a cross-domain label propagation method is developed to learn a target-domain classifier using the domain-shared group-sparse representations and the target-specific information from the target data. Experimental results show that the proposed method achieves good performance on cross-domain face and object recognition. Moreover, most distribution alignment methods have not considered the difference in distribution structures, which results in insufficient alignment across domains. Therefore, a novel graph alignment method is proposed, which aligns both data representations and distribution structural information across the source and target domains. An adversarial network is developed for graph alignment by mapping both source and target data to a feature space where the data are distributed with unified structure criteria. Promising results have been obtained in the experiments on cross-dataset digit and object recognition. Problem of dataset bias also exists in human pose estimation across datasets with different image qualities. Thus, this thesis proposes to synthesize target body parts for cross-domain distribution alignment, to address the problem of cross-quality pose estimation. A translative dictionary is learned to associate the source and target domains, and a cross-quality adaptation model is developed to refine the source pose estimator using the synthesized target body parts. We perform cross-quality experiments on three datasets with different image quality using two state-of-the-art pose estimators, and compare the proposed method with five unsupervised domain adaptation methods. Our experimental results show that the proposed method outperforms not only the source pose estimators, but also other unsupervised domain adaptation methods.

Este trabalho propõe um modelo hierárquico para coleta de dados de monitoramento em sistemas distribuídos. Seu objetivo é proporcionar a análise on-line do comportamento de sistemas e programas distribuídos. O meio escolhido para realizar essa análise foi a visualização. Inicialmente é apresentada uma contextualização sobre monitoramento de sistemas distribuídos. Também são abordados aspectos específicos ao monitoramento de Grid. Após, é analisado um conjunto de ferramentas de monitoramento. Então tem-se a apresentação do modelo proposto. Esse é composto por coletores locais, por uma hierarquia de agregadores e por clientes. É utilizado o modelo push de transmissão de dados e há um mecanismo de subscrição aos coletores. Foi implementado um protótipo do modelo de coleta proposto, que foi utilizado na implementação de um protótipo de ferramenta de monitoramento on-line. Nessa, os dados coletados são fornecidos ao DIMVisual, que é um modelo de integração de dados para visualização. Para visualização, o protótipo utiliza a ferramenta TRIVA, que recebe os dados integrados como entrada. Essa ferramenta foi modificada para gerar uma visualização que é atualizada de maneira on-line. Também foram realizados experimentos para avaliar o tempo necessário para enviar mensagens com diferentes hierarquias e configurações dos coletores. Além disso, foi avaliada a capacidade de o cliente implementado processar os dados recebidos, gerando sua visualização. / This work proposes a hierarchical model for collecting monitoring data from distributed systems. Its goal is to allow the on-line analysis of the behavior of distributed systems and applications. The means we chose to perform this analysis is to generate a visualization of the collected information. In the beginning of this dissertation we present an overview of the monitoring of distributed systems. Aspects that are specific to the monitoring of Grid systems are also reviewed. Next, we have an analysis of a set of monitoring tools. Then we present the proposed model, which is composed by local collectors, an hierarchical structure of aggregators and clients. A push data transmission model is used in the model and it also has a subscription mechanism. A prototype monitoring tool was implemented, integrating the data collection model with DIMVisual and TRIVA. The former is a data integration model whose output is formatted to be used as input for a visualization tool. The later is a visualization tool which, in the prototype, receives the integrated data from DIMVisual. TRIVA generates a visualization of the received information, which is updated in an on-line fashion. In order to evaluate the model, we performed a set of experiments using the prototype. One of the experiments measured the time spent to send data though different hierarchies. In these tests we have also varied the quantity and the configuration of the collectors. In another experiment we evaluated the capacity of the client to process the received data.

Linguagens de programação orientadas a objetos possuem diversas características que facilitam sua utilização frente a outras linguagens imperativas. No projeto e desenvolvimento de software, o mecanismo de herança permite a construção de sistemas na forma incremental e evolutiva, possibilitando a reutilização de códigos já escritos. Também a possível atingir aplicações com bons níveis de segurança e confiabilidade, através do encapsulamento de dados e funções sob forma de objetos, que também representam a unidade básica de execução em uma linguagem orientada a objetos. O mesmo recurso que possibilita níveis elevados de segurança permite que linguagens orientadas a objetos sejam inerentemente distribuídas. Objetos possuem tanto área de dados e código de execução independentes dos demais. Acessos aos dados internos de um objeto somente são possíveis através de mensagens explicitas entre objetos. Neste caso um objeto solicita uma ação específica a outro objeto, podendo ser enviados parâmetros e existir retorno de resultados. Este trabalho apresenta um modelo para construção de uma linguagem orientada a objetos distribuída. O ambiente para suportar a execução ao modelo é compostos por vários nodos de processamento com memórias locais individuais e contando com uma rede de comunicação para troca de mensagens entre os nodos. O modelo e discutido em dois níveis distintos: a nível de linguagem e a nível operacional. A nível de linguagem são analisados os recursos de programação normalmente utilizados em linguagens orientadas a objetos quando implementados em ambientes distribuídos. O ambiente de suporte A execução necessário ao suporte do modelo da linguagem a analisado pelo nível operacional. A apresentação do modelo a nível de linguagem discute as características de uma linguagem orientada a objetos distribuída frente as implementações seqüenciais convencionais. E ressaltada a implementado de herança em um ambiente de execução distribuído, que, não podendo ser através de compartilhamento, é efetuada através de copia de código. Também são apresentadas novas diretivas de compilação necessárias exclusivamente a ambientes distribuídos. Tais diretivas visam explorar níveis de concorrência de uma aplicação durante sua execução, diferenciando classes que definem objetos locais ou distribuídos e diferentes tipos de mensagens entre objetos. As formas de extrair o melhor desempenho nas aplicações e o gerenciamento do ambiente de execução são os pontos abordados pelo nível operacional do modelo. Em operação neste nível, um elemento de gerencia de execução permite o controle tanto dos objetos da aplicação quanto dos nodos de processamento disponíveis para execução. A tarefa de controle de objetos viabiliza a criação e remoção de objetos durante a execução da aplicação, bem como a identificação de localização destes. O controle dos nodos de processamento possibilita analisar continuamente a carga computacional dos nodos de processamento. Assim, cada objeto a ser criado pode ser alocado em um nodo onde a carga computacional esteja baixa, propiciando um melhor desempenho no momento de execução da aplicação distribuído a carga entre os nodos. A unido do modelo de execução distribuído proposto a uma linguagem orientada a objetos resulta em uma linguagem eficiente tanto na produção de software como no desempenho de aplicações. A eficiência na produção de sistemas 6 obtida através de dois itens, a utilização do paradigma de orientação a objetos e a transparência do nível operacional para o programador, que não necessita conhecer os mecanismos utilizados para ativação de objetos e envio de mensagens. A eficiência de execução é obtida através da utilização de múltiplos nodos processadores servindo como base a execução. Neste trabalho é também apresentado um protótipo para uma linguagem suportando o modelo distribuído proposto. A linguagem, denominada DPC++ (Processamento Distribuído em C++), é voltada para execução em redes de estações de trabalho, sobre o sistema operacional Unix, utilizando sockets como mecanismo de comunicação. O estilo de programação em DPC++ é baseado em C++. / The objects-oriented programming languages have many features who make simple their use in front of others imperatives languages. In the software project and development, the inheritance mechanism allows an increasing and evaluative way of codes that have been written. It also possible gain applications with goods levels of security and confiability with the encapsulation of both data and functions in the form of object, which represent the basic execution unit in an object-oriented language. The same resource that provides high levels of security also permits that object-oriented languages may be inherently distributed. Objects have their own area of data, their execution codes are independent from the other. Accesses to the internal data of an object are possible only through a specific protocol among objets. When this occurs, an object requests a specific action to other object with or without parameters or results return. This work presents a model for build a distributed object-oriented language, devoted to environments compounded by several processing nodes with local memory and linked by a communication network. The model is discussed in two different levels: language level and operational level. In the language level are analyzed the programming resources usually used in object-oriented languages when implemented in distributed environments. The executing environments support are analyzed in the operational level. In the language level presentation are made a discussion about distributed object-oriented language features in front of conventional sequential implementations. It is emphasized inheritance in a distributed executing environment, who is done by code copy, due to can not be by memory sharing. Also are presented news compilation directives necessaries to the distributed environment. Those directives aim to explorer concurrence levels in an application during its execution, differing class who defining local or distributed objects and the different messages types among objects. The operational level boards the ways that mean to extract the best performance for the applications and the execution environment management. An execution manager element allows the control as the application objects as the available to execution processor nodes. The task of object control makes possible the objects creation and removal during the application executing as well their network identification. The processor nodes control allows the continuous analyzes of the computational load in the nodes available to processing. In this way, every object to be created can be allocated in a node with low occupation rates, propitiating a better performance in the application executing. The union of the proposed distributed execution model to an objectoriented language results in an efficient language as in the software production as in execution performance. The systems production efficiency is obtained from two items: the utilization of the object-oriented paradigm an the transparency of the operational level to the programmer, that no need know the used mechanisms to object activation and message exchange. The execution efficiency is gained by the utilization of multiples processor nodes supporting the application executing. In this work is presented a prototype that implements the proposed model. The language, called DPC++, Distributed Processing in C++, is turned to execute in workstation network with Unix operational system, using sockets as communication mechanism. The style of DPC++ programming are based in C++.

Linguagens de programação orientadas a objetos possuem diversas características que facilitam sua utilização frente a outras linguagens imperativas. No projeto e desenvolvimento de software, o mecanismo de herança permite a construção de sistemas na forma incremental e evolutiva, possibilitando a reutilização de códigos já escritos. Também a possível atingir aplicações com bons níveis de segurança e confiabilidade, através do encapsulamento de dados e funções sob forma de objetos, que também representam a unidade básica de execução em uma linguagem orientada a objetos. O mesmo recurso que possibilita níveis elevados de segurança permite que linguagens orientadas a objetos sejam inerentemente distribuídas. Objetos possuem tanto área de dados e código de execução independentes dos demais. Acessos aos dados internos de um objeto somente são possíveis através de mensagens explicitas entre objetos. Neste caso um objeto solicita uma ação específica a outro objeto, podendo ser enviados parâmetros e existir retorno de resultados. Este trabalho apresenta um modelo para construção de uma linguagem orientada a objetos distribuída. O ambiente para suportar a execução ao modelo é compostos por vários nodos de processamento com memórias locais individuais e contando com uma rede de comunicação para troca de mensagens entre os nodos. O modelo e discutido em dois níveis distintos: a nível de linguagem e a nível operacional. A nível de linguagem são analisados os recursos de programação normalmente utilizados em linguagens orientadas a objetos quando implementados em ambientes distribuídos. O ambiente de suporte A execução necessário ao suporte do modelo da linguagem a analisado pelo nível operacional. A apresentação do modelo a nível de linguagem discute as características de uma linguagem orientada a objetos distribuída frente as implementações seqüenciais convencionais. E ressaltada a implementado de herança em um ambiente de execução distribuído, que, não podendo ser através de compartilhamento, é efetuada através de copia de código. Também são apresentadas novas diretivas de compilação necessárias exclusivamente a ambientes distribuídos. Tais diretivas visam explorar níveis de concorrência de uma aplicação durante sua execução, diferenciando classes que definem objetos locais ou distribuídos e diferentes tipos de mensagens entre objetos. As formas de extrair o melhor desempenho nas aplicações e o gerenciamento do ambiente de execução são os pontos abordados pelo nível operacional do modelo. Em operação neste nível, um elemento de gerencia de execução permite o controle tanto dos objetos da aplicação quanto dos nodos de processamento disponíveis para execução. A tarefa de controle de objetos viabiliza a criação e remoção de objetos durante a execução da aplicação, bem como a identificação de localização destes. O controle dos nodos de processamento possibilita analisar continuamente a carga computacional dos nodos de processamento. Assim, cada objeto a ser criado pode ser alocado em um nodo onde a carga computacional esteja baixa, propiciando um melhor desempenho no momento de execução da aplicação distribuído a carga entre os nodos. A unido do modelo de execução distribuído proposto a uma linguagem orientada a objetos resulta em uma linguagem eficiente tanto na produção de software como no desempenho de aplicações. A eficiência na produção de sistemas 6 obtida através de dois itens, a utilização do paradigma de orientação a objetos e a transparência do nível operacional para o programador, que não necessita conhecer os mecanismos utilizados para ativação de objetos e envio de mensagens. A eficiência de execução é obtida através da utilização de múltiplos nodos processadores servindo como base a execução. Neste trabalho é também apresentado um protótipo para uma linguagem suportando o modelo distribuído proposto. A linguagem, denominada DPC++ (Processamento Distribuído em C++), é voltada para execução em redes de estações de trabalho, sobre o sistema operacional Unix, utilizando sockets como mecanismo de comunicação. O estilo de programação em DPC++ é baseado em C++. / The objects-oriented programming languages have many features who make simple their use in front of others imperatives languages. In the software project and development, the inheritance mechanism allows an increasing and evaluative way of codes that have been written. It also possible gain applications with goods levels of security and confiability with the encapsulation of both data and functions in the form of object, which represent the basic execution unit in an object-oriented language. The same resource that provides high levels of security also permits that object-oriented languages may be inherently distributed. Objects have their own area of data, their execution codes are independent from the other. Accesses to the internal data of an object are possible only through a specific protocol among objets. When this occurs, an object requests a specific action to other object with or without parameters or results return. This work presents a model for build a distributed object-oriented language, devoted to environments compounded by several processing nodes with local memory and linked by a communication network. The model is discussed in two different levels: language level and operational level. In the language level are analyzed the programming resources usually used in object-oriented languages when implemented in distributed environments. The executing environments support are analyzed in the operational level. In the language level presentation are made a discussion about distributed object-oriented language features in front of conventional sequential implementations. It is emphasized inheritance in a distributed executing environment, who is done by code copy, due to can not be by memory sharing. Also are presented news compilation directives necessaries to the distributed environment. Those directives aim to explorer concurrence levels in an application during its execution, differing class who defining local or distributed objects and the different messages types among objects. The operational level boards the ways that mean to extract the best performance for the applications and the execution environment management. An execution manager element allows the control as the application objects as the available to execution processor nodes. The task of object control makes possible the objects creation and removal during the application executing as well their network identification. The processor nodes control allows the continuous analyzes of the computational load in the nodes available to processing. In this way, every object to be created can be allocated in a node with low occupation rates, propitiating a better performance in the application executing. The union of the proposed distributed execution model to an objectoriented language results in an efficient language as in the software production as in execution performance. The systems production efficiency is obtained from two items: the utilization of the object-oriented paradigm an the transparency of the operational level to the programmer, that no need know the used mechanisms to object activation and message exchange. The execution efficiency is gained by the utilization of multiples processor nodes supporting the application executing. In this work is presented a prototype that implements the proposed model. The language, called DPC++, Distributed Processing in C++, is turned to execute in workstation network with Unix operational system, using sockets as communication mechanism. The style of DPC++ programming are based in C++.

Doctor of Philosophy / Electrical and Computer Engineering / Ruth D. Miller / Optimizing renewable distributed generation in distribution systems has gained popularity with changes in federal energy policies. Various studies have been reported in this regard and most of the studies are based on optimum wind and/or solar generation planning in distribution system using various optimization techniques such as analytical, numerical, and heuristic. However, characteristics such as high energy density, relatively lower footprint of land, availability, and local reactive power compensation ability, have gained increased popularity for optimizing distributed wind generation (DWG) in distribution systems. This research investigated optimum distributed generation planning (ODGP) using two primary optimization techniques: analytical and heuristic. In first part of the research, an analytical optimization method called “Combined Electrical Topology (CET)” was proposed in order to minimize the impact of intentional structural changes in distribution system topology, in distributed generation/ DWG placement. Even though it is still rare, DWG could be maximized to supply base power demand of three-phase unbalanced radial distribution system, combined with distributed battery energy storage systems (BESS). In second part of this research the usage of DWG/BESS as base power generation, and to extend the ability to sustain the system in a power grid failure for a maximum of 1.5 hours was studied. IEEE 37-node, three-phase unbalanced radial distribution system was used as the test system to optimize wind turbines and sodium sulfide (NaS) battery units with respect to network real power losses, system voltage profile, DWG/BESS availability and present value of cost savings. In addition, DWG’s ability to supply local reactive power in distribution system was also investigated. Model results suggested that DWG/NaS could supply base power demand of a threephase unbalanced radial distribution system. In addition, DWG/NaS were able to sustain power demand of a three-phase unbalanced distribution system for 1.5 hours in the event of a power grid failure.

A fundamental assumption behind most machine learning (ML) algorithms and analyses is the sequential execution. That is, any update to the ML model can be immediately applied and the new model is always available for the next algorithmic step. This basic assumption, however, can be costly to realize, when the computation is carried out across multiple machines, linked by commodity networks that are usually 104 times slower than the memory speed due to fundamental hardware limitations. As a result, concurrent ML computation in the distributed settings often needs to handle delayed updates and perform learning in the presence of staleness. This thesis characterizes learning with staleness from three directions: (1) We extend the theoretical analyses of a number of classical ML algorithms, including stochastic gradient descent, proximal gradient descent on non-convex problems, and Frank-Wolfe algorithms, to explicitly incorporate staleness into their convergence characterizations. (2)We conduct simulation and large-scale distributed experiments to study the empirical effects of staleness on ML algorithms under indeterministic executions. Our results reveal that staleness is a key parameter governing the convergence speed for all considered ML algorithms, with varied ramifications. (3) We design staleness-minimizing parameter server systems by optimizing synchronization methods to effectively reduce the runtime staleness. The proposed optimization of a bounded consistency model utilizes the additional network bandwidths to communicate updates eagerly, relieving users of the burden to tune the staleness level. By minimizing staleness at the framework level, our system stabilizes diverging optimization paths and substantially accelerates convergence across ML algorithms without any modification to the ML programs.

Este trabalho propõe um modelo hierárquico para coleta de dados de monitoramento em sistemas distribuídos. Seu objetivo é proporcionar a análise on-line do comportamento de sistemas e programas distribuídos. O meio escolhido para realizar essa análise foi a visualização. Inicialmente é apresentada uma contextualização sobre monitoramento de sistemas distribuídos. Também são abordados aspectos específicos ao monitoramento de Grid. Após, é analisado um conjunto de ferramentas de monitoramento. Então tem-se a apresentação do modelo proposto. Esse é composto por coletores locais, por uma hierarquia de agregadores e por clientes. É utilizado o modelo push de transmissão de dados e há um mecanismo de subscrição aos coletores. Foi implementado um protótipo do modelo de coleta proposto, que foi utilizado na implementação de um protótipo de ferramenta de monitoramento on-line. Nessa, os dados coletados são fornecidos ao DIMVisual, que é um modelo de integração de dados para visualização. Para visualização, o protótipo utiliza a ferramenta TRIVA, que recebe os dados integrados como entrada. Essa ferramenta foi modificada para gerar uma visualização que é atualizada de maneira on-line. Também foram realizados experimentos para avaliar o tempo necessário para enviar mensagens com diferentes hierarquias e configurações dos coletores. Além disso, foi avaliada a capacidade de o cliente implementado processar os dados recebidos, gerando sua visualização. / This work proposes a hierarchical model for collecting monitoring data from distributed systems. Its goal is to allow the on-line analysis of the behavior of distributed systems and applications. The means we chose to perform this analysis is to generate a visualization of the collected information. In the beginning of this dissertation we present an overview of the monitoring of distributed systems. Aspects that are specific to the monitoring of Grid systems are also reviewed. Next, we have an analysis of a set of monitoring tools. Then we present the proposed model, which is composed by local collectors, an hierarchical structure of aggregators and clients. A push data transmission model is used in the model and it also has a subscription mechanism. A prototype monitoring tool was implemented, integrating the data collection model with DIMVisual and TRIVA. The former is a data integration model whose output is formatted to be used as input for a visualization tool. The later is a visualization tool which, in the prototype, receives the integrated data from DIMVisual. TRIVA generates a visualization of the received information, which is updated in an on-line fashion. In order to evaluate the model, we performed a set of experiments using the prototype. One of the experiments measured the time spent to send data though different hierarchies. In these tests we have also varied the quantity and the configuration of the collectors. In another experiment we evaluated the capacity of the client to process the received data.