[pt] Os sistemas de potência são sistemas cyber-físicos responsáveis pela geração e transporte da energia elétrica desde sua fonte geradora até os consumidores finais. Durante este percurso, existem diversos processos que devem ser seguidos para se manter a qualidade do serviço e a segurança e estabilidade do sistema. Um destes processos envolve o controle de diversos equipamentos de maneira que a tensão dos barramentos do sistema se mantenha dentro de faixas pré-estabelecidas. Este controle, normalmente realizado pelos operadores do sistema em tempo real e por equipamentos automáticos de controle, envolve um número muito grande de considerações que dificilmente serão avaliadas no momento da decisão. Para contornar este problema, propõe-se a utilização de uma ferramenta inteligente que seja capaz de escolher as melhores ações a serem tomadas para que a tensão do sistema se mantenha nos níveis adequados levando em consideração as variadas condições do sistema. A metodologia utilizada pela ferramenta consiste na técnica de Deep Reinforcement Learning juntamente com três novas variações: windowed, ensemble e windowed ensemble Q-Learning, que consistem na divisão do processo otimizado em janelas de treinamento, utilização de múltiplos agentes inteligentes para um mesmo processo e a combinação destas duas metodologias. As variações são testadas em circuitos consagrados na literatura e são capazes de obter resultados expressivos quando comparados com a abordagem de Deep Reinforcement Learning tradicional utilizada em outros estudos e com o controle intrínseco do próprio sistema, mantendo a tensão sob controle ao longo do dia. / [en] Electrical Power Systems are cyber-physical systems responsible for the generation and transportation of energy from its generating source to the final customers. During this process many different activities must be conducted in order to keep quality of service and the system s safety and stability. One of these activities regards control of various equipment in order to keep the voltage level on each system bus between specified limits. This control, which is usually conducted by system s operators in real time and by automatic control equipment involves many different constraints and considerations that are hardly ever taken into account during the decision process. In order to mitigate this problem a smart agent capable of deciding which action is best in order to keep the voltages in adequate levels taking into account system s conditions is proposed. The proposed methodology consists on the Deep Reinforcement Learning technique along with three novel variations: windowed, ensemble and windowed ensemble Q-Learning, which consist on the division of the problem in training windows, the usage of multiple learning agents for the same process and on the combination of both these techniques. The variations are tested on academically consecrated test circuits and are capable of attaining expressive results when compared to the traditional Deep Reinforcement Learning approach which is used in other academic studies and also with the systems intrinsic control, keeping voltage under control along the day.

The variable nature of photovoltaic (PV) generation can cause voltage fluctuations in power distribution systems. Feedback control can be used to minimize the voltage fluctuations. This thesis presents the results obtained from comparing the control performance of two types of PV generation feedback control, namely Volt-VAR control and constant power factor control. A three minute PV generation transient is used to evaluate controller performance, where the transient data used originated from one second measurements taken on an actual PV generator. Using the three minute transient, a set of parametric studies are performed on both feedback control strategies. The performance of the control strategies are compared as to voltage control on the distribution feeder and also to the effect that the control may have on transmission system voltage. In considering transmission system voltage, the reactive power drawn from the substation during the transient is evaluated. Simulation results suggest that the choice of control to be implemented should be based on both transmission and distribution system operational concerns. / Master of Science

No description available.

Fuel cells have the potential to provide clean power for a variety of uses including stand-alone residential power. But to increase the acceptance of fuel cells for off-grid generation, the cost of the energy management system must be greatly reduced. Of the many ways to accomplish this, this paper looks at reducing cost through topology changes and elimination of current sensors. A dual 2.5kW non-isolated bidirectional boost converter is designed and analyzed. The various bidirectional boost topologies are compared on cost and ability to meet the specifications. A sensorless average current mode is designed, implemented and verified through testing in a low-cost fixed-point DSP. Both boost and buck modes are accurately modeled and voltage and current controllers are designed for good closed-loop response. The accuracy of the sensorless average current measurement is investigated in both modes of operation. A classical dual-loop controller is implemented in boost mode with the sensorless average current and in buck mode, a dual controller operating in either current or voltage mode is implemented. The design is verified through testing in boost and buck mode and it is shown that the results are acceptable. / Master of Science

[pt] Esta pesquisa está relacionada à avaliação das condições de estabilidade de tensão em pontos de operação do sistema elétrico. A modelagem de geradores / compensadores síncronos e compensadores estáticos são analisadas. Estabeleceram-se os melhores modelos com controle de tensão ativo e com perda do controle. A questão da adequação do controle de tensão foi tratada através da análise dos elementos da matriz de sensibilidade entre as grandezas controladoras e as tensões controladas. Os índices de estabilidade de tensão foram calculados. Em todos os casos, a perda de controle de tensão tem impacto negativo sobre as condições de estabilidade de tensão. Nos casos com ação de controle inadequada, independe da relação inversa entre grandeza controladora e tensão controlada, um aumento da tensão na barra controlada teve um impacto positivo sobre as condições de estabilidade de tensão, e uma diminuição da tensão na barra controlada teve impacto negativo sobre as condições de estabilidade de tensão. Esse é o resultado esperado também quando a ação de controle é adequada. / [en] This research is related to the evaluation of voltage stability conditions of operating points of the power system. It has been studied the modelling of synchronous generators/synchronous compensators and static var compensators. The best models were established with active voltage control and with loss of control. The issue of voltage control was treated through analysis of the matrix sensitivity elements between the controlling variables and voltages at controlled buses. The voltage stability indexes were evaluated. In all cases, control loss proved to have a negative impact on voltage stability conditions. In cases with inadequate control action, results are non-dependent of the opposed relation between the controlling variable and voltage at controlling bus, an increase on voltage at controlled busbar proved to have a positive impact on voltage stability conditions, and a decrease on voltage at controlled busbar proved to have a negative impact on voltage stability conditions. This is also expected when the control action is appropriate.

Submitted by JOSIANE SANTOS DE OLIVEIRA (josianeso) on 2018-12-12T13:10:44Z No. of bitstreams: 1 Felipe de Carli_.pdf: 6857644 bytes, checksum: 51df2faf2ff2365c2443afa4f23a4292 (MD5) / Made available in DSpace on 2018-12-12T13:10:44Z (GMT). No. of bitstreams: 1 Felipe de Carli_.pdf: 6857644 bytes, checksum: 51df2faf2ff2365c2443afa4f23a4292 (MD5) Previous issue date: 2018-10-26 / FATEC – Fundação de Apoio á Tecnologia e Ciência / A geração, transmissão e distribuição de energia elétrica têm papel chave no desenvolvimento econômico de um país. A qualidade e disponibilidade da energia elétrica relacionada a distribuição tem impacto direto na economia e na qualidade de vida das pessoas. Conceitos de Redes de Distribuição Inteligentes tem ganhado cada vez mais força com o intuito de melhorar a qualidade dos serviços prestados. A tensão em regime permanente é um dos principais parâmetros de qualidade de energia elétrica em redes de distribuição e muitas vezes o seu controle requer soluções de elevada complexidade e investimento. Tensões fora de especificação tem impactos negativos para os consumidores finais, uma vez que, podem gerar danos aos equipamentos, interrupção em serviços fundamentais, e inconvenientes para a sociedade. Desta forma, o presente trabalho busca desenvolver um protótipo de controle inteligente e dinâmico aplicado para reguladores de tensão de distribuição com a capacidade de coletar informações do perfil de consumo do alimentador, realizar o auto ajuste de seus parâmetros e enviar informações para o sistema de supervisão. O protótipo dividido em um módulo de controle e módulo fuzzy que foi desenvolvido em escala de laboratório. Baseado no carregamento do alimentador, o dispositivo é capaz de calcular os parâmetros básicos de auto-ajuste, ou seja, a temporização, faixa de insensibilidade e compensações de queda na linha. Os resultados globais desse trabalho mostram que o desenvolvimento de um controle dinâmico e inteligente para reguladores de tensão é viável e pode ser aplicado em redes de distribuição reais. O cálculo dos parâmetros de temporização, tensão de referência, compensação de queda na linha melhora o perfil de tensão de uma rede de distribuição, pois minimiza as oscilações de tensão em baixa e alto carregamento da linha, conforme testes realizados com dados de um alimentador real. Entende-se que os resultados globais deste trabalho servem de fundamento para um estudo técnico-econômico de ampliação desta estrutura de controle para outros reguladores de tensão de redes de distribuição. / The electricity generation, transmission, and distribution play a role key in the economic development of a country. The quality and availability of electricity related to distribution has a direct impact on the economy and people's quality of life. Concepts of smart grids has gained more strength in order to improve the quality of services provided. Steady-state voltage is one of the main quality parameters of electrical power in distribution grids and often its control requires solutions of high complexity and investment. Out-of-specification voltage have negative impacts on end-users, as it can cause damage to equipment, outage in key services, and inconvenience to community. In this way, the present work seeks to develop a prototype of intelligent and dynamic control applied to distribution voltage regulators with the ability to collect information of the feeder consumption profile, perform the self-tuning of its parameters and send information to supervisory system. The prototype divided into a control and fuzzy modules was developed on a laboratory scale. Based on loader readings, the device is able to calculate the basic parameters of self-adjustment, ie time delay, dead band and line drop compensation. The overall results of this work show that the development of a dynamic and smart control for step voltage regulators is feasible and can be applied in real distribution grid. The calculation of the time delay, reference voltage, line drop compensation improves the voltage profile of a distribution grid, as it minimizes voltage variations at low and high line loading, according to tests performed with data from a real feeder. It is understood that the overall results of this work serve as a basis for a technical-economic study to extend this control structure to other step voltage regulators in distribution grids.

The installation of wind power plant has significantly increased since several years due to the recent necessity of creating renewable and clean energy sources. Before the accomplishment of a wind power project many pre-studies are required in order to verify the possibility of integrating a wind power plant in the electrical network. The creation of models in different software and their simulation can bring the insurance of a secure operation that meets the numerous requirements imposed by the electrical system. Hence, this Master thesis work consists in the creation of a wind turbine model. This model represents the turbines installed at Lillgrund wind farm, the biggest wind power plant in Sweden. The objectives of this project are to first develop an accurate model of the wind turbines installed at Lillgrund wind farm and further to use it in different kinds of simulations. Those simulations test the wind turbine operating according to different control modes. Also, a power quality analysis is carried out studying in particular two power quality phenomena, namely, the response to voltage sags and the harmonic distortion. The model is created in the software PSCAD that enables the dynamic and static simulations of electromagnetic and electromechanical systems. The model of the wind turbine contains the electrical machine, the power electronics (converters), and the controls of the wind turbine. Especially, three different control modes, e.g., voltage control, reactive power control and power factor control, are implemented, tested and compared. The model is tested according to different cases of voltage sag and the study verifies the fault-ride through capability of the turbine. Moreover, a harmonics analysis is done. Eventually the work concludes about two power quality parameters.

A novel probabilistic method has been developed for modelling the operation of energy storage in electricity systems with significant amounts of wind and solar powered generation. This method is based on a spectral analysis of the variations of wind speed and solar irradiance together with profiles of electrical demand. The method has been embodied in two Matlab computer programs: Wind power only: This program models wind power on any time scale from seconds to years, with limited modelling of demand profiles. This program is only capable of modelling stand-alone systems, or systems in which the electrical demand is replaced by a weak grid connection with limited export capacity. 24-hours: This program models wind power, solar PV power and electrical demand, including seasonal and diurnal effects of each. However, this program only models store cycle times (variations within a time scale) of 24 hours. This program is capable of modelling local electrical demand at the same time as a grid connection with import or export capacity and a backup generator. Each of these programs has been validated by comparing its results with those from a time step program, making four Matlab programs in total. All four programs calculate the power flows to and from the store, satisfied demand, unsatisfied demand and curtailed power. The programs also predict the fractions of time that the store spends full, empty, filling or emptying. The results obtained are promising. Probabilistic program results agree well with time step results over a wide range of input data and time scales. The probabilistic method needs further refinement, but can be used to perform initial modelling and feasibility studies for renewable energy systems. The probabilistic method has the advantage that the required input data is less, and the computer run time is reduced, compared to the time step method.

[pt] A crescente demanda de energia elétrica, associada a limitações na expansão do sistema de transmissão, resulta na operação dos sistemas elétricos cada vez mais próxima de seus limites, tornando-os vulneráveis a problemas de estabilidade de tensão. Neste contexto, o desenvolvimento de métodos para avaliação do carregamento da rede de transmissão tornou-se fundamental para que se possa compreender o funcionamento do sistema nestas condições, bem como possibilitar sua operação com maior confiabilidade e segurança. No presente trabalho é apresentada a formulação não-clássica para o controle de tensão local, remoto e coordenado, utilizada na solução do problema do fluxo de potência. A partir desta formulação, foi realizado o estudo da teoria associada à avaliação do carregamento da rede de transmissão considerando não só barras de carga e barras de tensão controlada localmente, mas também barras controladoras e de tensão controlada em sistemas com controle de tensão remoto e coordenado. Para essas barras, foram deduzidos índices de avaliação das condições de estabilidade de tensão, os quais indicam a região de operação de cada barra na curva para phi constante no plano SV, a margem em MVA para o máximo carregamento e a importância relativa entre as barras. Exemplos numéricos são apresentados demonstrando a aplicabilidade dos índices propostos. / [en] The growing demand for electricity, coupled with the limitations on the transmission system’s expansion, results in the operation of power systems ever closer to their limits, making them vulnerable to voltage stability problems. In this context, the development of methods to evaluate the loading of the transmission system became essential so that you can understand the operation of the system under these conditions, and to enable its operation with greater reliability and security. This work presents the formulation for the non-classical local, remote and coordinated voltage control, used in solving the problem of load flow. From this formulation, the study of the theory associated with the assessment of the loading of the transmission system was conducted considering not only load buses and locally voltage controlled buses, but also voltage controlling and controlled buses in systems with remote and coordinated voltage control. For these buses were deducted evaluation indexes of voltage stability conditions, which indicate the operating region of each bus in the S-V curve, the margin in MVA for maximum loading and the relative importance between the buses. Numerical examples are presented demonstrating the applicability of the proposed indexes.

[pt] O controle das tensões dos diversos barramentos de um sistema elétrico de potência tem como objetivo garantir a qualidade da energia fornecida aos consumidores. As tensões devem respeitar níveis regulamentados pelo governo. Atualmente, no Sistema Elétrico de Potência (SEP) brasileiro, a tarefa do controle de tensão, realizada pelos operadores de tempo real, se baseia nos valores e tendências de diversas variáveis (tensões, potências reativas e ativas, sensibilidade dos equipamentos, dentre outras). Para a formação de um operador nessa tarefa são necessários de um a dois anos, tempo que poderia ser reduzido caso um sistema de apoio à decisão dedicado ao problema de controle de tensão estivesse à disposição durante o treinamento. Entretanto, em virtude do grande número de grandezas a serem analisadas e de suas não linearidades, é necessário uma ferramenta automática de apoio à decisão que seja capaz de tratar intrinsecamente relações não lineares. Deste modo, neste trabalho optou-se por desenvolver um sistema baseado em Redes Neurais Artificiais (RNA) para a confecção do sistema sugerido, com o objetivo de indicar a necessidade de realizar ações de controle de tensão utilizando-se dos recursos ou equipamentos disponíveis. O sistema desenvolvido é composto de três módulos: Pré-processamento; Análise e Classificação do evento; e Pós-processamento. Tal sistema serve para sugerir a manobra de equipamentos mais adequada para o controle de tensão. No estudo de caso, o sistema proposto foi avaliado nos equipamentos de controle de tensão (reatores, capacitores e tapes) constantes no Sistema de Transmissão em 765 kV, responsável pela interligação dos sistemas Sul e Sudeste do Brasil. Utilizando dados obtidos do sistema de aquisição em tempo real, diferentes configurações de RNAs foram testadas. Os melhores resultados foram obtidos com uma estrutura de duas redes neurais por equipamento a ser controlado, apresentando, em média, 80% de acerto em relação às manobras realizadas em tempo real. Em virtude da complexidade do problema, os resultados foram considerados mais do que satisfatórios, indicando a aplicabilidade desta técnica para a realização do sistema desejado. / [en] The main objective of the voltage control in Electrical Power System (EPS) is to guarantee the quality of the energy supplied to consumers. The voltage must respect government regulated levels. Currently, on the Brazilian EPS, the voltage control task is carried out by system operators based on diverse information, such as current values, and trends of electric variables (voltages, reactive and active powers, their sensitivities in the control devices performance, amongst others). To fully train a operator in this task it is necessary one or two years, period that could be greatly reduced if a decision support system was available during the operator`s training. However, due to the great number of variables that must be analyzed and their nonlinearity, an automatic decision support tool, capable to treat nonlinear relations, is necessary. Therefore, this work proposes a system based on Artificial Neural Networks (ANN), with the objective to identify the necessity or not to use the voltage control resources in the EPS. The developed system is composed of three modules: Pre-processing; Event Analysis and Classification; and Post-processing. Such decision support system suggests the most adequate equipment maneuver in the voltage control task. In the case study, the proposed system was evaluated using the available voltage control equipments (reactors, capacitors and transformer taps) in the 765 kV Transmission System, main responsible trunk for the interconnection of South and Southeastern Brazilian systems. Using real time data, different ANN configurations have been tested. The best results were obtained with a structure composed of two neural networks, for each controlled equipment, presenting, in average, 80% accuracy in relation to maneuvers occurred in real time. Due to the problem complexity, the results were considered more than satisfactory, indicating the applicability of this technique for the development of the desired system.