Doutorado em Engenharia Elétrica e da Computação (EMC)

URI Permanente para esta coleção

http://200.137.215.59/tede/handle/tede/3828

Navegar

Agora exibindo 1 - 20 de 38

Diagnóstico otimizado de transformadores de potência mediante a integração de técnicas preditivas
(Universidade Federal de Goiás, 2018-02-22) Marques, André Pereira; Brito, Leonardo da Cunha; http://lattes.cnpq.br/6660680440182900; Ribeiro, Cacilda de Jesus; http://lattes.cnpq.br/8998911644222586; Ribeiro, Cacilda de Jesus; Brito, Leonardo da Cunha; Domingues, Elder Geraldo; Rocha, Adson Silva; Arruda, Colemar
The effective application of predictive maintenance techniques for power transformers is essential for the early detection of defects and failures, and in order to trigger scheduled preventive or corrective actions. This precludes the need for urgent and/or emergency interventions, aiming to reduce unscheduled power outages, which are usually more costly and cause great inconvenience to consumers and the electric power system. In this context, this work focuses on the development of an optimized diagnostic methodology for power transformers, by integrating eleven predictive techniques (with 27 quantities). These techniques are analyzed with respect to voltage level, type of electrical connection and age of the transformers – when applicable –, based on the experience of specialists/analysts, on standards and technical guidelines, and on statistical analyses of a database of actual field and laboratory tests. The analyses of the validation of the results are significant and presented through a specialized system, demonstrating their practical application. The novelty of this thesis consists in the development of an original classification method, called Normalized Doubly Weighted Sum (NDWS), and in the detailed description of the predictive techniques that are integrated, their functionalities, the creation of equations, the definition of criteria and parameters, with their weights and scores for the classification of “A” (excellent) to “E” (very poor), and recommended actions aimed at underpinning decision- making, thereby contributing to the body of studies in this field. Hence, it can be concluded that this work offers a comprehensive and efficient tool to aid in the optimized diagnosis of power transformers, insulated with kraft paper and immersed in mineral insulating oil, providing effective diagnostics and maintenance of these devices, and thus increasing the reliability of electric power systems.
Efeitos da operação do gerador de indução no comportamento do gerador síncrono operando em um sistema isolado alimentando cargas não lineares
(Universidade Federal de Goiás, 2018-07-13) Oliveira, José Mário Menescal de; http://lattes.cnpq.br/2672882686688933; Alves, Antônio César Baleeiro; http://lattes.cnpq.br/1188328474646154; Alves, Antônio César Baleeiro; Rodrigues, Kleiber David; Salerno, Carlos Henrique; Santos, Euler Bueno dos; Nerys, José Wilson Lima
This thesis demonstrates the effects of harmonic pollution in a salient pole synchronous generator and an induction generator operating in parallel on an isolated system, supplying a non-linear load. The main contributions of this research-study consist of identifying and quantifying the oscillations that non-linear load cause on the electric variables of synchronous and induction generators, such as, the electromagnetic conjugate that presents oscillations of sixth harmonic due to the distorted currents.
Escalonamento de recursos em redes LTE utilizando processo envelope de tráfego multifractal e curva de serviço mínima
(Universidade Federal de Goiás, 2018-12-14) Abrahão, Diego Cruz; Vieira, Flávio Henrique Teles; http://lattes.cnpq.br/0920629723928382; Vieira, Flávio Henrique Teles; Sousa, Marcos Antônio de; Lemos, Rodrigo Pinto; Lima, Marcos Antônio Cardoso de; Cardoso, Kleber Vieira
In this work, a variation of the MWM (Multifractal Wavelet Model) model is proposed for network traffic flows, in such a way that its parameters are estimated adaptively. Next, it proposes an envelope process of network traffic based on the parameters' adaptive estimation of this model, whose final objective is to provide quality of service (QoS) in real time. The proposed envelope is compared to the main envelopes processes known in the literature, that are based on traffic models, such as: Brownian Motion (Bm), Fractional Brownian Motion (fBm) and Multifractal Brownian Motion (mBm). This work investigates the use of F-OFDM multi-carrier modulation, which is one of the candidates for 5G networks. It is known that the LTE / LTE-A (Long-Term Evolution) network uses Adaptive Modulation and Coding (AMC) technique that have the function of adjusting the modulation order and code rate based on the user's channel state information, in order to achieve a Block Error Rate (BLER) lesser than 10%. In this work, simulations of the LTE downlink are carried out using OFDM and F-OFDM multicarrier modulation, with the objective of mapping the BLER as a function of SNR (Signal-to-Noise Ratio). This mapping is necessary for the network to adjust the modulation and code scheme appropriate to each user. It's also proposed an adaptive algorithm for resource allocation in the LTE/LTE-A downlink, with admission control of the users. This algorithm aims to improve the performance of some network parameters and to guarantee maximum delay, through of following information: backlog, channel condition and user’s traffic behavior. In order to control the admission of users and to estimate the maximum delay of the network, a minimum adaptive service curve of the LTE / LTE-A network is proposed. The proposed algorithm for resource allocation is compared with several scheduling schemes known in the literature through computational simulations of different LTE network scenarios using OFDM and F-OFDM multi-carrier modulation.
Controle adaptativo de fluxos de tráfego de redes baseado em modelagem multifractal e sistemas fuzzy
(Universidade Federal de Goiás, 2019-08-13) Cardoso, Alisson Assis; Vieira, Flávio Henrique Teles; http://lattes.cnpq.br/0920629723928382; Vieira, Flávio Henrique Teles; Sousa, Marcos Antônio de; Lemos, Rodrigo Pinto; Vieira, Robson Domingos; Dantas, Maria José Pereira
The network traffic flows that arrive at the base station to be transmitted to the mobile users, in a 5G network system, enter the queuing process until transmission rates are provided. In order to minimize the delay, this work proposes the use of flow control algorithms based on the prediction of user queue behavior. Thus, the more accurate the data prediction, the greater the accuracy and control of flow control algorithms. To improve accuracy, models describing the behavior of network traffic are employed. In this work, two adaptive modeling algorithms based on the Lognormal Beta and BetaMWM models are proposed to model the network traffic and allow its use in real-time applications, such as the 5G network. Simulations are performed in comparisons to multifractal models found in the literature to validate the proposed algorithms, where results in terms of expected value, variance, moments of 2º to 4º order, mean squared errors of autocorrelation and distribution function prove the adaptively use of the algorithms. To perform the flow control, an equation is also proposed to obtain the optimal prediction-based control rate, where generalized ortonormal functions and fuzzy modeling are employed. Simulations of the Downlink 5G link are also performed to validate the proposed flow control algorithms. For this, results in terms of Flow, Utilization, Loss Rate, Delay and Average Waiting Queue are presented, proving the efficiency in the use of multifractal models, orthonormal basis functions, and fuzzy modeling in flow control algorithms for Downlink 5G systems. Taking advantage of the proposed multifractal modeling, an equation is also proposed to estimate the delay limitation for the first recommendations of the 5G network using the network calculation theory. For this, it is proposed a stochastic envelope process for network traffic based on the Adaptive Beta Lognormal model where comparisons with envelope processes known in the literature are performed.
Modelos de simulação e controle preditivo generalizado de sistemas fotovoltaicos conectados à rede
(Universidade Federal de Goiás, 2019-11-28) Franco, Ricardo Augusto Pereira; Vieira, Flávio Henrique Teles; http://lattes.cnpq.br/0920629723928382; Vieira, Flávio Henrique Teles; Negrete, Lina Paola Garces; Castro, Marcelo Stehling de; Soares, Telma Worle Lima; Belchior, Fernando Nunes
After the Second Industrial Revolution, the electric energy became an essential item for the world population. Beyond the environmental impacts caused by polluting energy sources, the awareness of clean energy generation has been growing worldwide. In this way, renewable and clean energy sources gain prominence in relation to traditional dirty energy sources. Photovoltaic systems can help to reduce the energy consumption at the consumer units in which they are deployed and it can contribute to reducing the energy demand provided by the utility power grid. Thus, the objective of this work is to provide modeling, simulation and control methods for photovoltaic systems in order to assist the photovoltaic systems projects, to analyze the quality of the energy produced, the impacts that these systems cause on the distribution network and to perform the control of power generation of photovoltaic systems. It is proposed to develop optimization and analytical methods for the estimation of the parameters of the single diode model. The estimation of the parameters is performed to improve the modeling of a photovoltaic module and, consequently, its performance in the generation of energy. Then, simulations of grid-connected photovoltaic systems are proposed with the purpose of analyzing the impacts that the photovoltaic energy generation implies on the consumer loads and on the distribution network. Finally, it is proposed a method of energy flow control in a system consisting of photovoltaic system and energy storage system, in order to reduce the energy consumed from the utility grid. The energy flows in this electric system with energy storage system (hybrid system) can be combined to provide power for a load or inject power into the distribution network. The results presented show the benefits of the estimation of parameters of a photovoltaic module in terms of power generation. The results of the simulations show the impacts of photovoltaic systems for the consumer who installed the system and for the utility to design its energy distribution networks in medium and long-term. The results of the controller present the optimization achieved through the control of hybrid energy systems, reducing the costs of the energy consumption of a load and making the most efficient use of the energy storage system.
Development and analysis of mathematical methods for estimating statistical parameters in sensor array-based systems
(Universidade Federal de Goiás, 2020-02-07) Kunzler, Jonas Augusto; Sander, Oliver; Lemos, Rodrigo Pinto; lattes.cnpq.br/3333000136853156; Lemos, Rodrigo Pinto; Sander, Oliver; Fleury, Claudio Afonso; Vieira, Robson Domingos; Castro, Marcelo Stehling de
A estimação da direção de chegada e a estimação de pulsos de energia na espectroscopia de raios-x são baseadas no mesmo parâmetro, o deslocamento de fase de sinais em relação a uma referência. Neste trabalho é demonstrada a possibilidade de aplicar técnicas já estabelecidas para a estimação da direção de chegada ao problema de espectroscopia. Os dois temas são correlacionados em uma parte introdutória e em seguida eles são discutidos separadamente dando ênfase nas características intrínsecas de cada um. No problema de estimação da direção de chegada, inicialmente apresenta-se o modelo de sinal para o arranjo de sensores e alguns métodos baseados no estimador de máxima verossimilhança. Considerando um arranjo de sensores linearmente distribuídos e com espaçamento uniforme entre os elementos, os sinais induzidos em cada circuito serão cópias defasadas de um sinal de referência. O defasamento está intimamente relacionado com a direção de chegada. No domínio espacial, as direções de chegada representam frequências de funções exponenciais complexas, as quais deverão ser estimadas. Os principais métodos de estimação são o MODE, MODEX, o MODEX modificado, e o SEAD. Atenção especial é dispensada ao método SEAD que é baseado na decomposição em autovalores da matriz de correlação espacial modificada. A diferença entre os dois maiores autovalores gera uma curva com picos proeminentes indicando as direções de chegada de ondas planas. A esta curva dá-se o nome de espectro diferencial. Uma análise matemática do espectro diferencial, denominada de espectro diferencial total, é desenvolvida e demonstra-se que a norma matricial induzida pela norma-2 vetorial é a principal componente do cálculo. Portanto, propõe-se uma abordagem baseada em normas matriciais para a estimação das direções de chegada. Uma descrição matemática geral foi desenvolvida, a qual explicita a relação entre os verdadeiros ângulos de chegada e um ângulo sintético usado para varrer todo o espectro, a correlação entre as fontes, o número de fontes e o número de sensores. Demonstra-se que a diferença entre os ângulos determina a amplitude do pico gerado. A formulação matemática do espectro angular constitui em uma das principais contribuições deste trabalho, porém, outros aprimoramentos foram alcançados através da proposta de uso de normas matriciais. A utilização da abordagem baseada em normas evita a necessidade de realização da decomposição da matriz em autovalores e, consequentemente, o tempo de execução total do método é reduzido e o erro quadrático médio é reduzido para situação de fontes afastadas. Comparando as propostas com os métodos estabelecidos na literatura, a abordagem de normas supera o método MODE e seus derivados com relação ao erro quadrático médio. Porém, com relação ao tempo de execução os métodos baseados no SEAD são mais custosos computacionalmente para um número de fontes menor do que 4. A estimação de pulsos de energia na espectroscopia de raios-x não é um tema novo, porém, o multiplexador SQUID de micro-ondas é uma abordagem relativamente nova e, atualmente, ele consiste em um tema de pesquisa pujante. O objetivo do sistema é estimar a energia de partículas energéticas que incidem em detectores, os quais podem ser do tipo transition edge sensor ou metallic magnetic calorimeters. Os calorímetros metálicos são sensores paramagnéticos que estão situados em um campo magnético fraco, eles traduzem variação de temperatura em variação de fluxo magnético. Para realizar a leitura das variação de fluxo magnético decorrente da incidência de partículas emprega-se um componente supercondutor denominado de superconducting quantum interference device (SQUID). O SQUID é um interferômetro de extrema sensibilidade e se comporta como um indutor variável. O SQUID é acoplado a uma terminação de uma linha de transmissão supercondutora e produz alterações na frequência de ressonância do circuito. Portanto, a energia de partículas pode ser lida eletronicamente através da demodulação de uma onda de rádio que percorre o ressonador. A modulação ocorre na amplitude e fase de uma portadora complexa. Após a exclusão de frequências intermediárias, resultado do processo de mixagem da portadora, obtem-se uma portadora complexa de frequência baixa sobre a qual métodos de estimação de fase são aplicados. Pela semelhança com o problema de estimação da direção de chegada, propõe-se uma abordagem baseada em arranjo de sensores e decomposição da matriz de correlação espacial em autovalores. Desta forma, estabelece-se uma analogia entre os dois problemas. Para a obtenção da matriz de correlação, considera-se a existência de um vetor de referência cuja fase não depende da incidência de partículas, i.e., quando o sistema está em repouso. Este vetor de referência é comparado por meio da função de correlação com os dados que são recebidos a cada instante. Define-se um comprimento do vetor de dados que seja conveniente e, desta forma, obtem-se uma matriz de snapshots com dois sensores e uma matriz de correlação de dimensões 2 × 2. A decomposição em autovalores pode ser resolvida explicitamente em função das entradas da matriz, de tal maneira, que uma formulação geral para o método é apresentada. Além disso, simplificações podem ser impostas com o objetivo de implementação em hardware dedicado. O erro de estimação é analisado para o método proposto e o método de máxima verossimilhança. O método de autovalores apresentou maior robustez ao ruído e em algumas circunstâncias consegue resolver o problema de estimação sem artifícios externos, algo que não ocorre para o método de máxima verossimilhança.
Técnicas de acionamento e controle ótimo aplicados ao motor a relutância chaveado para maximizar o rendimento
(Universidade Federal de Goiás, 2020-03-06) Reis, Márcio Rodrigues da Cunha; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Calixto, Wesley Pacheco; Nerys, José Wilson Lima; Sousa, Marcos Antônio de; Lemos, Rodrigo Pinto; Alves, Aylton José
This work presents modeling, driving and speed control techniques for the switched reluctance motor. The objective is to improve the computational model, the control response and the machine efficiency. A parametric regression model was used to find the inductance profile of the switched reluctance motor and from the new inductance profile model. The drive and control techniques are shown: i) with classical speed control acting on the excitation voltage and fixed switching angles, ii) with classical speed control acting on the switching angles and fixed excitation voltage and iii) with classical speed control acting on the excitation voltage, in this case, with dynamic switching angles and controller parameters. The inductance profile is represented analytically and inserted into the machine computer model, allowing greater precision and low computational cost. The speed controls acting on the excitation voltage with dynamic controller parameters and dynamic switching angles allowed shorter response time for a wide range of control, higher efficiency, low computational cost and simplified implementation and maintenance. The techniques proposed in this work obtained precision of the computational model with respect to the system (in workbench) and optimized parameters in a wide range of the speed control, allowing an improvement of switched reluctance motor efficiency.
Repotencialização na operação paralela de gerador síncrono com gerador de indução
(Universidade Federal de Goiás, 2020-03-27) Magalhães, Alana da Silva; Alves, Ayton José; http://lattes.cnpq.br/2762752291082988; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Calixto, Wesley Pacheco; Alves, Aylton José; Wainer, Gabriel Andrés; Rodrigues, Clóves Gonçalves; Lemos, Rodrigo Pinto
This work presents the comparison between simulation and experimental tests of the repowering system to validate the electrical interactions between an induction generator and a synchronous generator. Parametric regression and optimization models are used to find the constructive parameters of the machines. Two generators are connected to a common bus in steady state, subject to non-linear loads. The results comparing modeling and experimental tests show that the induction generator besides the active power increasing, has a better way for harmonic currents flowing in common bus. It is concluded that the parametric regression has the advantages of not needing to know the parameters provided by the machine's manufacturers and does not need to perform destructive tests and that the induction generator repowering and attenuates current harmonic components present at the connection point, improving the network voltage profile.
Detecção de imagens falsificadas baseada em descritores locais de textura e rede neural convolucional
(Universidade Federal de Goiás, 2020-06-30) Ferreira, William Divino; Soares, Fabrizzio Alphonsus Alves de Melo Nunes; http://lattes.cnpq.br/7206645857721831; Cruz Júnior, Gélson da; http://lattes.cnpq.br/4370555454162131; Cruz Júnior, Gélson da; Pedrini, Hélio; Salvini, Rogério Lopes; Costa, Ronaldo Martins da; Lemos, Rodrigo Pinto
Nowadays, digital image transformation has become a widespread activity. Hence, image copying, cloning, and resizing are easily performed, making it challenging to check image integrity and authenticity. Moreover, a criminal investigation from digital images becomes extremely hard, because using those images as proof demands to ensure its legitimately,under a risk to implicate the whole legal process.In this sense, this work develops a model for forged images based on local texture descriptors with convolutional neural networks. Henceforth, in this work, firstly, we evaluated fourteen local texture descriptors in five public image texture datasets, and then we selected descriptors with the best efficacy. Second, the selected descriptors are applied to four public datasets to extract texture features from forged and legit images. Finally, those features are used to train a residual convolutional neural network, and then, classifying images as authentic or forged with a Support Vector Machine Classifier. A result of the proposed model provides enthusiasm, mainly when applied to a dataset with a small number of images.
Programação inteira binária com técnicas bio-inspiradas para o planejamento otimizado de redes de transporte ópticas
(Universidade Federal de Goiás, 2020-08-25) Oliveira, Bruno Quirino de; Vieira, Flávio Henrique Teles; http://lattes.cnpq.br/0920629723928382; Vieira, Flávio Henrique Teles; Sousa, Marcos Antônio de; Rocha, Flávio Geraldo Coelho; Dantas, Maria José Pereira; Cardoso, Alisson Assis
In telecommunications systems, data traffic continues to grow at a high speed, and the increase in both the amount of services offered and the required transmission rate are responsible for this scenario. Clearly, this growth in data traffic is posing serious challenges for optical transport networks in terms of improving their capacity efficiency in order to meet new traffic requirements. This work presents optimization models for the design of optical transport networks. The optical network planning problem is considered, in which a traffic interest matrix between demand nodes is specified. This traffic interest matrix can be modeled in terms of the required transmission rate or the number of channels required for a standardized modular service. The optical transport network is modeled as a graph, using the arc-path approach. Models of integer linear programming (ILP) and mixed integer linear programming (MILP) with variables 0-1 are developed with guidance to minimize costs. Restrictions on guaranteeing demand compliance, specific technical capacity of equipment and exclusivity in the allocation of transmission link modularity are also contemplated. In order to ensure more flexible and realistic decision support systems regarding the application scenarios they intend to portray, artificial intelligence techniques, such as fuzzy logic, genetic algorithms and firefly, are incorporated into the modeling and resolution processes of the models. In this sense, a Hybrid Firefly-Genetic (HFA) optimization method is used to solve the ILP problem, for the planning of the optical transport network (OTN), considering cost minimization. The method combines the Firefly discrete algorithm (FA) with the standard genetic algorithm (GA). Computational results of scenarios that contemplate: medium and large networks, different optical transmission technologies and diversity of traffic matrices are presented and discussed. The results achieved are encouraging, with emphasis on the ease of adapting the MILP and ILP models to meet new requirements and/or specificities of the network and technology to be evaluated.
Modelo baseado em redes neurais profundas com unidades recorrentes bloqueadas para legendagem de imagens por referências
(Universidade Federal de Goiás, 2020-09-28) Nogueira, Tiago do Carmo; Vinhal, Cássio Dener Noronha; http://lattes.cnpq.br/9791117638583664; Cruz Júnior, Gélson da; http://lattes.cnpq.br/4370555454162131; Cruz Júnior, Gélson da; Ferreira, Deller James; Santos, Gilberto Antonio Marcon dos; Vinhal, Cássio Dener Noronha; Lemos, Rodrigo Pinto
Describing images using natural language has become a challenging task for computer vision. Image captioning can automatically create descriptions through deep learning architectures that use convolutional neural networks (CNNs) and recurrent neural networks (RNNs). Image captioning has several applications, such as object descriptions in scenes to help blind people walk in unknown environments, and medical image descriptions for early diagnosis of diseases. However, architectures supported by traditional RNNs, in addition to problems of exploding and fading gradients, can generate non-descriptive sentences. To solve these difficulties, this study proposes a model based on the encoder-decoder structure using CNNs to extract the image characteristics and multimodal gated recurrent units (GRU) to generate the descriptions. The part-of-speech (PoS) and the likelihood function are used to generate weights in the GRU. The proposed method performs knowledge transfer in the validation phase using the k-nearest neighbors (kNN) technique. The experimental results in the Flickr30k and MS-COCO data sets demonstrate that the proposed PoS-based model is statistically superior to the leading models. It provides more descriptive captions that are similar to the expected captions, both in the predicted and kNN-selected captions. These results indicate an automatic improvement of the image descriptions, benefitting several applications, such as medical image captioning for early diagnosis of diseases.
Complexidade natural de sistemas com base em análise de sensibilidade
(Universidade Federal de Goiás, 2020-09-30) Gomes, Viviane Margarida; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Calixto, Wesley Pacheco; Wainer, Gabriel Andres; Peretta, Igor Santos; Pinheiro Neto, Daywes; Martins, Weber
This work proposes a methodology for analyzing systems based on a particular measure of complexity, called the natural complexity of the system. This measure corresponds to the proper level of complexity of each system, characterized by the region of optimized configurations. Given the optimal or optimized solution, the sensitivity analysis is performed to define the impact generated at the output of the system due to variations in the input parameters. The proposed methodology comprises: i) sensitivity analysis metrics, ii) system complexity metrics based on weighted connections, iii) analysis of the system using natural complexity as a reference and iv) development of models for application of the methodology. The complexity metric uses the sensitivity indices of the parameters to define the relevance values of the connections, in order to establish a relationship between the parties and the whole. The results point to the complexity metric as a mechanism for synthesizing the configuration, arrangement, performance and workload of the system in a single measure. Regarding the measure of natural complexity, it may be used as a reference of the desired level of complexity, since it was significantly different from the measures obtained under overload or idle conditions. Thus the natural complexity may correspond to the minimum complexity value of the system in regular activity. The proposed complexity metric strengthened the assertion that every system exhibits some level of complexity. Thus, it may be said that complexity is the totality of the system in interaction, with its own internal dynamics and its own environmental flows.
Modelo neural recozido para a representação semântica de documentos por meio de vetores contínuos
(Universidade Federal de Goiás, 2020-11-13) Mendonça, Leandro Rezende Carneiro de; Cruz Junior, Gelson da; http://lattes.cnpq.br/4370555454162131; Cruz Junior , Gelson da; Soares Alcalá , Symone Gomes; Oliveira , Marco Antonio Assfalk de; Soares , Fabrízzio Alphonsus Alves de Melo Nunes; Campos , Sérgio Vale Aguiar
As a result of the growing production of unstructured textual data, techniques for representing words and documents in the vector space have emerged recently. The Brazilian Public Ministry has received several textual requests that are send by citizens with different needs, such as those involved in cases of domestic violence against women, others requesting intensive care unit admissions, and more. The time spent in classifying, detecting similar requests and distributing them is essential to optimize and save public resources. Therefore, we adopted the neural model with the Simulated Annealing (SA), a classic global optimization algorithm with low computational complexity, because of the need to reduce the daily training time, providing a more friendly graphic visualization of data in high dimensions, supporting the judicial decision process. The physical analogy of the SA meta-heuristic associated with the continuous representation of documents in the vector space contribute greatly to the friendly visualization of a high-dimensional dataset, maintaining a comparable score with other deep models and optimization algorithms, such as Covariance Matrix Adaptation Evolution Strategy (CMA-ES) and Bayesian Optimization (BO).
Modelagem e análise da migração da umidade em sistema isolante papel-óleo de transformadores de potência
(Universidade Federal de Goiás, 2020-11-20) Sousa, Felipe Resende de Carvalho; Brito, Leonardo da Cunha; http://lattes.cnpq.br/6660680440182900; Brito, Leonardo da Cunha; Ribeiro, Cacilda de Jesus; Domingues, Elder Geraldo; Marques, André Pereira; Paula, Geyverson Teixeira de
This thesis aims to present an innovative tool designed to study how the phenomenon of moisture migration occurs inside insulating paper-oil systems of power transformers. The tool consists of a two-dimensional mathematical model, representing the dynamics of humidity in the oil-paper set and including the variations in electrical charging, which was developed using mathematical tools, existing studies, and the experience of specialists/analysts. Furthermore, this study includes a classification method developed to investigate some of the harmful effects of moisture in power transformers, considering the possibility of bubbles and free water formation. The novelty of this numerical method consists of a calculation methodology that contemplates the use of the finite element method - in two dimensions - applied to the concepts of mass transfer and thermal modeling of the transformer loading, considering the temperature differences between two windings. Additionally, the classification method includes the creation of: a) a categorization system for the coupling of a transformer and electric charging in five levels, from “A” to “E”, with the respective recommended actions; and b) an analysis that involves varying the different levels of loading applied to the equipment. The contribution of this work is an efficient method, implemented via a computational algorithm, which provides support for the maintenance team's decision making when unfavorable phenomena and conditions occur that can critically affect the useful life of the equipment, especially its insulation. This thesis presents some case studies of transformers in operation, under real load conditions, in which the data validation is performed through the comparation of obtained data with experimental collection and analysis of oil samples and the results of monitoring device. Therefore, this work is useful in supporting the classification and diagnosis of power transformers, through the prediction of humidity levels in the oil-paper isolating system, assisting on failure prevention in this equipment and capital loss to the owners of the transformers and consumers supplied by these.
Potencialização do rendimento do gerador a relutância chaveado empregando técnica de rastreamento associada a controle de tensão otimizado
(Universidade Federal de Goiás, 2020-12-16) Araújo, Wanderson Rainer Hilário de; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Coimbra, António Paulo Mendes Breda Dias; Ribeiro, Luiz Eduardo Bento; Sousa, Marcos Antônio de; Oliveira, Marco Antônio Assfalk de; Calixto, Wesley Pacheco
This work presents the potentialization of the efficiency of the Switched Reluctance Generator (GRC) submitted to the control of the output voltage. The efficiency is enhanced by using tracking technique acting on the switching angles of the power converter. As it is a DC machine, the control of the output voltage is applied to adapt this value to DC-AC conversion systems and load controllers. The PID controller is used and, because it is a controller with only one output, other quantities are not contemplated by the performance of the controller. In this work, there is an interest in enhancing the performance of the GRC to make this type of machine more attractive for generating electricity in distributed systems and in installations without connection to the main distribution network. Therefore, in parallel to the PID controller, a tracking technique is applied to the GRC performance with a disturb and observe algorithm. Other procedures are presented, such as obtaining an inductance surface to improve the mathematical and computational modeling of the generator, in addition to the development of an indirect conjugate detection system. Simulation and experimental results are presented for validation and discussion of the proposed study.
Machine learning methods applied to intraday solar forecasting for power system operation
(Universidade Federal de Goiás, 2021-03-05) Paiva, Gabriel Mendonça de; Marra, Enes Gonçalves; http://lattes.cnpq.br/8463332056679918; Alvarenga, Bernardo Pinheiro de; http://lattes.cnpq.br/9850449311607643; Alvarenga, Bernardo Pinheiro de; Mussetta, Marco; Marra, Enes Gonçalves; Pimentel, Sergio Pires; Brito, Leonardo da Cunha
Prever o recurso solar é uma ferramenta essencial para sua integração com a rede elétrica. Esta tese foca em previsão solar intra-diária, com uma análise robusta de previsão de irradiância testada em múltiplas localidades e uma proposta de implementação de previsão de potência fotovoltaica (FV). Dois algoritmos de aprendizagem de máquinas são avaliados para previsão intra-diária da irradiância solar: programação genética multigene (PGMG) e redes neurais artificiais do tipo multilayer perceptron (MLP). PGMG é um algoritmo evolucionário e um método tipo "caixa branca" e é uma nova técnica na área. Os algoritmos de aprendizagem de máquinas também são comparados com um modelo de persistência inteligente (smart persistence) para prever a irradiância solar com dados de seis localidades. Os horizontes de previsão considerados são 15-120 minutos à frente. Os resultados das simulações mostram um aprimoramento consistente das previsões quando variáveis climáticas exógenas são adicionadas como entrada aos modelos, sendo 5.68% o aprimoramento pelo cálculo de erro médio absoluto (MAE) e 3.41% o aprimoramento pelo cálculo de raiz do erro quadrático médio (RMSE). Os resultados também mostram que localidade, horizonte de previsão e métrica de erro escolhida influenciam a dominância de acurácia dos modelos. Dois modelos de irradiância de céu claro foram implementados, mas os resultados indicam para uma baixa influência dos modelos na acurácia de previsão para previsões multivariadas por aprendizagem de máquinas. Em uma perspectiva genérica, PGMG apresentou resultados mais precisos e robustos que MLP em previsões individuais, provendo soluções mais rápidas. Entretanto, MLP apresentou mais precisão em previsões do tipo ensemble, porém estas apresentam também maior complexidade e maior custo computacional. A implementação de previsão de potência FV mostrou resultados consistentes, aprimorando valores de RMSE de previsões de persistência em 9.79%-23.75% para horizontes de 15-120 minutos.
Otimização da força contra eletromotriz e do torque eletromagnético mediante a modificações nos parâmetros estruturais de uma máquina síncrona de ímãs permanentes
(Universidade Federal de Goiás, 2021-08-02) Jesus, Luiz Henrique Reis de; Brito, Leonardo da Cunha; http://lattes.cnpq.br/6660680440182900; Paula, Geyverson Teixeira de; http://lattes.cnpq.br/0140145167826333; Paula, Geyverson Teixeira de; Pereira, William César de Andrade; Almeida, Thales Eugenio Portes de; Suetake, Marcelo; Brito, Leonardo da Cunha
This work aims to present an approach to the optimization process of a surface mounted permanent magnet synchronous machine (SM-PMSM) using the Tensors method, as well as methods and techniques for parametric optimization and topological optimization. It evaluates the behavior of the Back-Electromotive Force (Back-EMF) operating at no load and under load, as well as the behavior of the electromagnetic torque. The optimized parameters directly relate to the saturation of the machine as the electromagnetic load is increased (nominal load). The optimization process allows a reduction in the electromagnetic torque ripple, keeping the back-electromotive force unchanged. The results for the optimization of the Back-Electromotive Force presented variations lower than 0,5% when compared to its behavior operating at no load, with its behavior operating under load. For the electromagnetic torque optimization process, the results achieved provided to the SM-PMSM project a reduction in the electromagnetic torque ripple of approximately 37% and a reduction in the structural volume of 38,23%. Finally, optimization techniques and methodologies were compared regarding the results and objectives proposed for the optimization of the permanent magnet synchronous machine under study.
Desenvolvimento de métrica de avaliação bilíngue por estrutura frasal aplicada à tradução automática
(Universidade Federal de Goiás, 2021-10-08) Silva, Alan Henrique Ferreira; Wainer, Gabriel Andres; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Calixto, Wesley Pacheco; Wainer, Gabriel Andres; Reis, Márcio Rodrigues da Cunha; Marques, Leonardo Garcia; Araújo, Wanderson Rainer Hilário de
This work proposes a metric capable of evaluating automatic bilingual translations, through the use of phrasal structures between the source and target languages. Existing methods currently provide a dependent assessment of large amounts of data as texts examples. The proposed methodology uses as a database the association of generic phrasal structures between the source and target language to assess the text quality, resulting given by an automatic translation machine. The method validation is performed using comparative studies between the evaluations generated by the proposed metric and the BLEU and METEOR metrics. The results obtained shows greater accuracy of evaluation of the proposed metric compared to the other metrics, with an improvement of approximately $5\%$ compared to the results of the evaluation of human translators. The results obtained in the tests validate the capability of the metric as an auxiliary tool in the evaluation of machine translations.
Análise e otimização da rede de distribuição de energia utilizando conceitos de redes inteligentes
(Universidade Federal de Goiás, 2021-10-21) Caetano Neto, João; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Calixto, Wesley Pacheco; Reis, Márcio Rodrigues da Cunha; Lemos, Rodrigo Pinto; Marques, Leonardo Garcia; Magalhães, Alana da Silva
The main objective of this work is to develop a methodology for analyzing the quality of the voltage level in the distribution power grid to identify and re-duce the violations of voltage limits through the proposition of optimal points for the allocation of photovoltaic distributed generation. The methodology uses the geographic location of the power grid and its consumers to perform the grouping and classification in spatial grids of 100×100 m using the average annual consumption profile. The generated profiles, including the grid infor-mation, are sent to the photovoltaic distributed generation allocation algo-rithm, which, using an optimization process, identifies the geographic location, the required installed capacity, and the minimum number of photovoltaic gen-eration units that must be inserted to minimize the violations of voltage limits, respecting the necessary restrictions. The entire proposal is applied in a real feeder with thousands of bars, whose model is validated with measurements carried out in the field. Different violations of voltage limits scenarios are used to validate the methodology, obtaining grids with better voltage quality after the optimized allocation of photovoltaic distributed generation. The proposal presents itself as a new tool in the work of adapting the voltage of the distri-bution power grid using photovoltaic distributed generation.
Aplicação da inteligência artificial, ontologia e mineração de dados para classificação de sentenças judiciais
(Universidade Federal de Goiás, 2021-12-20) Castro Junior, Antonio Pires de; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Peretta, Igor Santos; Araújo, Wanderson Rainer Hilário de; Soares, Fabrizzio Alphonsus Alves de Melo Nunes; Gomes, Viviane Margarida; Calixto, Wesley Pacheco
The objective of this work is to apply together ontology with bag-of-words models, similarity learning, and document classification in texts with uttered decisions. The objective is to improve the results of data mining in a database of court decisions. An automatic method of searching sentences in judicial processes related to the one under judgment is developed using the frequency term-inverse of frequency in documents model together with the Jaccard similarity coefficient, establishing weights on the co-occurrence of terms in legal texts of the same category. A dataset with document vectorization is used for supervised training of machine learning algorithms, aiming to classify new justice processes. The proposed methodology provides flexibility to the Judiciary, simulating the role of legal advisors in preparing court decisions with less time and efficiency in the search for jurisprudential standards. The results obtained show that, through accuracy metrics, the proposed model is effective and efficient, and can be applied in the process of identification of court decisions. Thus, the application of artificial intelligence, ontology, and data mining is indicated for information retrieval in court decisions.

Navegar

Navegando Doutorado em Engenharia Elétrica e da Computação (EMC) por Data de Publicação

Resultados por página

Opções de Ordenação