O trabalho desenvolvido tem como objetivo analisar o impacto da Produção em Regime Especial nos Preços do MIBEL, assumindo cenários em que a PRE é nula para possível comparação de preços de mercado. Para tal, serão analisadas várias tecnologias de Produção em Regime Especial e estudado o seu impacto nos preços do Mercado Ibérico de Eletricidade.

No description available.

Os preços de mercado da energia elétrica são obtidos através do cruzamento de propostas de compra e de venda apresentadas ao Operador do Mercados pelos diversos agentes registados. O aumento do número de veículos elétricos irá necessariamente aumentar o preço da energia elétrica em algumas horas. Esses aumentos dependerão do número de veículos elétricos considerados e igualmente do perfil de carregamento considerado. Neste trabalho pretende-se construir uma aplicação que utilize informação obtida no site do MIBEL e que, considerando um número especificado de veículos elétricos e respetivo perfil horário de carregamento, permita estimar os novos valores dos preços da energia elétrica.

No description available.

Todos nós tomamos contacto diário com distritos ou setores, direta ou indiretamente. O carteiro que todos os dias entrega correspondência na nossa residência possui uma área de entrega pela qual ele é responsável. Ou ainda, no caso de queda de neve, também as operações de limpeza e remoção da mesma encontram-se sujeitas a uma divisão de acordo com setores.Todos estes exemplos têm algo em comum: uma grande área geográfica é particionada em regiões mais pequenas. Por setorização entende-se assim esta divisão de um todo, em elementos de menor dimensão designados por setores, distritos, zonas, regiões ou áreas de responsabilidade.

A classificação automática de imagens é uma área científica que se encontra na fronteira entre aVisão por Computador eMachine Learning. A classificação de imagens é um processo que, muitosimplesmente, consiste na atribuição de imagens a uma ou mais categorias.Até muito recentemente, os sistemas tipicamente utilizados para fazer a classificação au-tomática de imagens, eram constituídos por duas camadas distintas. A primeira camada é tipica-mente constituída por umFeature ExtractoreFeature Detector. Sendo que estes subsistemas sãotipicamente constituídos por complexos algoritmos já existentes. A segunda camada é usualmenteconstituída por um classificador. Nos últimos anos, asconvolutional neural networks(CNNs)têm-se mostrado capazes de obter melhores resultados que estes sistemas mais tradicionais.Os carros têm um papel muito importante no mundo de hoje e a classificação automática dedanos é, também por isso, bastante relevante para a indústria seguradora, entre outras. As se-guradoras automóveis têm que lidar frequentemente com inspeções a veículos danificados. Esteprocesso é normalmente demorado e ineficiente, com inconvenientes e custos quer para a em-presa, quer para os clientes. Embora a automatização total destes processos possa ainda não serpossível, a utilização de sistemas de classificação automática de danos pode acelerar e melhorarestes processos manuais com a tecnologia de que dispomos hoje.Não existe, do meu conhecimento, nenhuma investigação feita, que aplique os avanços maisrecentes na área das CNNs à deteção automática de danos em veículos. Isto embora exista jáalguma investigação que indica uma boa performance deste tipo de sistemas em tarefas de classi-ficação relativas a imagens de veículos.Este documento ilustra a aplicação dos melhores e mais recentes avanços na área de classi-ficação de imagem, nomeadamente os últimos avanços em CNNs, à classificação de imagens deveículos danificados. Para isso foi desenvolvido um protótipo de um sistema capaz de identificar eclassificar a severidade dos danos evidenciados em imagens, bem como localizar, com algum graude precisão, a área do veículo que se encontra danificada. Serão usadas as arquiteturas de CNNsmais promissoras, baseando o critério de escolha das arquiteturas em aspetos relacionados com apercentagem de erro de classificação, bem como na rapidez com que a classificação é feita.Para o treino e teste do protótipo, foi criado um conjunto de dados que permite treinar e avaliara performance do sistema, uma vez que tal conjunto de dados não se encontra disponível. Para tal,foram recolhidas imagens de pesquisas efetuadas em motores de busca, bem como recolhidas dewebsitesde agências de segurança automóvel. / Visual image classification is a research area that involves both computer vision and machinelearning. The task of visually classifying an object consists in assigning an object to a category, orset of categories the object belongs to.Traditionally, visual classification tasks are performed using a two layered system, made upof a first layer featuring an out-of-the-shelf feature extractor and detector, and a second classifierlayer. In most recent years, convolutional neural networks have been shown to outperform suchpreviously used systems.Cars have a paramount role in today's world, and being able to automatically classify damagesin cars is of great interest specially to the car insurance industry. Car insurance companies dealwith car inspections on a daily basis. Such inspections are a manual, lengthy and sometimes faultyprocesses. Processes that bring costs and inconveniences to costumers and insurance companiesalike. Even though the total replacement of such manual inspection processes might still be faraway, developing systems to aid, accelerate or enhance the process might be possible with today'stechnology.There isn't, to my knowledge, much work developed in automatic visual car damage classi-fication, and none of it employs these recent performance improvements in image classificationmade possible through the use of CNNs. This happens in spite of some recent research pointing atthe fact that modern CNN technology does in fact, outperform traditional methods in non damage,car related image classification tasks.I hope to successfully apply state-of-the-art Convolutional Neural Network technology to solvethe problem of automatically identifying, distinguishing and locating damages in car images. Iintend to develop a working prototype of a system that will be able to tell if a given photographexhibits a car with damages or not, and possibly identifying, to some extent, the damaged areaswithin the car. The most promising CNN architectures will be used, taking in account both itsclassification accuracy as well as training and classification times.In order to be able to develop such system, a suitable dataset was gathered. The dataset is veryunbalanced in terms of the represented classes. Such imbalances have important effects that werecorrected with suitable techniques to prevent a significant performance degradation. The datasetis used to both train and measure the performance of the system. Since no car damage datasetsare freely available, the used dataset is composed of images gathered using search engines and carcrash agencies galleries.

No description available.

The desire of improving and developing new technologies targeting the ocean's supervision is continuously increasing. Since underwater tasks might involve hostile environments far too hazardous for human, it is typical to resort to system based on Remotely Operated Vehicles (ROVs) and/or Autonomous Underwater Vehicles (AUVs). During the extraction of information, the position control of the vehicle is critical. Specifically, the distance between the vehicle and the sea floor must be warily controlled to ensure its safety and the reliability of the missions that require proximity to the object of interest. Commonly, to deal with the altitude control, a system based on sonar technology is used. Although this solution simplifies the problem and is effective in most cases, it carries a lot of disadvantages in some underwater conditions and in some vehicles with certain specifications. Particularly the sensors based on acoustic waves, like the sonar, might present difficulties on the interpretation of the signals received when the vehicle is too close to the obstacle, requiring a minimum distance to retrieve valuable and reliable information. Furthermore, the inclusion of the sonar sensor demands an increase on the energetic cost of the system that in the case of vehicles powered by an external source through a cable like the ROVs is not a problem, but in AUVs, it might be valuable to avoid it since these vehicles are powered by batteries. Lastly, sometimes the space occupation of the sonar sensor represents a problem in some vehicles with meticulous limits relative to space usage, a common problem found in AUVs.In order to overcome these problems, the acquirement of the distance measurement can be accomplish through image processing using a system based on a camera and laser pointer devices. Since several underwater vehicles already have an embedded camera and it is common the existence of laser pointer devices as a scale, this approach is opportune and can accomplish the task with high reliability and efficiency.In this work it is presented a module capable of measuring the distance based on computer vision (Sensor module) and a module able to filter the data gathered though the use of Kalman Filter and capable of using this data to control the distance of the vehicle using a velocity and position controller that are adaptable to the mission characteristics (Filtering and Control module). The vehicle used in order to test the modules created was a profiler developed in \cite{Monteiro}. The Sensor module was implemented based on two laser pointer devices placed parallel to one another beside a CCD camera. In order to calculate the distance of the vehicle towards the obstacle was used the laser triangulation principle. Furthermore, the Sensor module is capable of retrieving information about the quality of the measurements and apply mathematical operations like circular average. It allows the user to fully configure the information that is gathered and what type of operations are performed through an configuration file. The communication with the Sensor module is made through UDP. In order to characterize and test the module, the laser triangulation principle was analyzed and a series of experimental tests were performed to know the error induced through the utilization of non ideal components and possible software limitations. The Filtering and Control module is responsible for the interface between the Sensor module and the vehicle, and the control of the thruster's actuation. It receives the data gathered, filters it through a Kalman filter that is tuned using the quality factor of the measurements, and then makes the information available through a shared memory block to the vehicle's software. The solution adopted regarding the control stands on the switching of two controllers, a velocity controller (based on a PI controller approach), and a position controller (based on a PID controller approach). The mathematical model of the vehicle was used in order to design the parameters of the controllers to accomplish certain temporal demands of the mission. The designed controllers were validated using the simulink toolbox from Matlab.In order to validate the systems created in a real environment, a series of operational tests were performed where the profiler is commanded to different altitudes. These tests were realized on a tank where the environment conditions are controllable and the results can be compared to the exact values. / The desire of improving and developing new technologies targeting the ocean's supervision is continuously increasing. Since underwater tasks might involve hostile environments far too hazardous for human, it is typical to resort to system based on Remotely Operated Vehicles (ROVs) and/or Autonomous Underwater Vehicles (AUVs). During the extraction of information, the position control of the vehicle is critical. Specifically, the distance between the vehicle and the sea floor must be warily controlled to ensure its safety and the reliability of the missions that require proximity to the object of interest. Commonly, to deal with the altitude control, a system based on sonar technology is used. Although this solution simplifies the problem and is effective in most cases, it carries a lot of disadvantages in some underwater conditions and in some vehicles with certain specifications. Particularly the sensors based on acoustic waves, like the sonar, might present difficulties on the interpretation of the signals received when the vehicle is too close to the obstacle, requiring a minimum distance to retrieve valuable and reliable information. Furthermore, the inclusion of the sonar sensor demands an increase on the energetic cost of the system that in the case of vehicles powered by an external source through a cable like the ROVs is not a problem, but in AUVs, it might be valuable to avoid it since these vehicles are powered by batteries. Lastly, sometimes the space occupation of the sonar sensor represents a problem in some vehicles with meticulous limits relative to space usage, a common problem found in AUVs.In order to overcome these problems, the acquirement of the distance measurement can be accomplish through image processing using a system based on a camera and laser pointer devices. Since several underwater vehicles already have an embedded camera and it is common the existence of laser pointer devices as a scale, this approach is opportune and can accomplish the task with high reliability and efficiency.In this work it is presented a module capable of measuring the distance based on computer vision (Sensor module) and a module able to filter the data gathered though the use of Kalman Filter and capable of using this data to control the distance of the vehicle using a velocity and position controller that are adaptable to the mission characteristics (Filtering and Control module). The vehicle used in order to test the modules created was a profiler developed in \cite{Monteiro}. The Sensor module was implemented based on two laser pointer devices placed parallel to one another beside a CCD camera. In order to calculate the distance of the vehicle towards the obstacle was used the laser triangulation principle. Furthermore, the Sensor module is capable of retrieving information about the quality of the measurements and apply mathematical operations like circular average. It allows the user to fully configure the information that is gathered and what type of operations are performed through an configuration file. The communication with the Sensor module is made through UDP. In order to characterize and test the module, the laser triangulation principle was analyzed and a series of experimental tests were performed to know the error induced through the utilization of non ideal components and possible software limitations. The Filtering and Control module is responsible for the interface between the Sensor module and the vehicle, and the control of the thruster's actuation. It receives the data gathered, filters it through a Kalman filter that is tuned using the quality factor of the measurements, and then makes the information available through a shared memory block to the vehicle's software. The solution adopted regarding the control stands on the switching of two controllers, a velocity controller (based on a PI controller approach), and a position controller (based on a PID controller approach). The mathematical model of the vehicle was used in order to design the parameters of the controllers to accomplish certain temporal demands of the mission. The designed controllers were validated using the simulink toolbox from Matlab.In order to validate the systems created in a real environment, a series of operational tests were performed where the profiler is commanded to different altitudes. These tests were realized on a tank where the environment conditions are controllable and the results can be compared to the exact values.

Atualmente as empresas têm necessidade em reagir em tempo real a eventos que ocorram durante o seu funcionamento. Estes eventos surgem sob a forma de streams de eventos que ocorrem a um dado instante de tempo.Uma forma popular de processar essas streams de eventos é utilizar a tecnologia de Complex Event Processing. Esta permite processar streams de eventos em tempo real e construir janelas temporais sobre essa stream, podendo depois aplicar agregações sobre as mesmas. Normalmente esta funcionalidade é obtida através da adição de funcionalidades à linguagem SQL por parte de um motor de CEP, permitindo que se possa utilizar SQL para processar streams através da construção de queries e criar janelas temporais sobre as mesmas.Infelizmente muitos sistemas de CEP requerem conhecimento à priori do tipo (schema) de eventos que terão de processar bem como do tipo de queries que irão ser executadas sobre eles.Pretende-se com esta dissertação implementar um sistema de CEP que possa funcionar sem ter qualquer tipo de conhecimento à priori do tipo de eventos que possam surgir, mas mantendo a capacidade de criar queries que possam executar esses eventos.O motor de CEP utilizado nesta dissertação foi o Microsoft StreamInsight. / Nowadays business needs to react in real time to events that happen during their work. These events appear as streams of events, where each event has occurred during a point in time.A popular way to process those event streams, is by using Complex Event Processing. This technology allows the processing, in real time, of event streams, the creating of time windows over those streams and the use of aggregations on those windows. Usually this functionality is gained by using a CEP engine that extends the SQL language allowing the latter to process streams by constructing queries and create temporal windows on them.Unfortunately before using a CEP system, many require à priori knowledge regarding the type (i.e. schema) of events that can appear on their streams and what queries it can run. This dissertation implements a CEP system that can work without knowing the type of events that may appear, but still has the ability to create queries over those event.The CEP engine used in this dissertation was Microsoft's StreamInsight.

Sendo a televisão o meio de comunicação e de entretenimento mais popular no mundo, foi com naturalidade que surgiu a ligação entre a Internet os conteúdos televisivos que chegam diariamente a mais de um bilião de pessoas a nível global. As estações de televisão estão sempre à procura de novas formas de cativar e captar mais audiência e os fabricantes responderam com as Smart TVs, capazes de providenciar aplicações dinâmicas que sejam passíveis de interagir com o utilizador.Outro mercado que está em forte ascensão é o das casas inteligentes, recheadas de sensores e atuadores que visam centralizar o controlo e a realização de tarefas domésticas e, desta forma, aumentar o conforto dos seus habitantes.Esta dissertação assenta sobre estas duas temáticas, propondo a criação de uma solução que seja capaz de receber a informação de vários sensores instalados numa casa inteligente, injetá-la no sinal de televisão doméstico e culminar no desenvolvimento de uma aplicação para Smart TVs capaz de fazer a leitura dos dados e de os mostrar no ecrã de forma simples, intuitiva e dinâmica.O contributo principal envolve a injeção de dados personalizados no sinal de televisão, exigindo a criação de um perfil de metadados específico para a situação. / Being the television the most popular electronic media and entertainment device in the world, naturally was created a bond between the Internet and the television content that daily to the homes of more the a billion people globally. TV stations are always looking for new ways to engage and grab the audience and the manufacturers replied with the Smart TVs, able to provide dynamic applications with interactive content.Another market in strong ascension is the smart homes, filled with sensors and actuators aiming to centralize the domestic tasks, increasing the user's confort.This dissertation relies on these two themes, proposing the creation of a solution capable of receiving information from several sensors installed in a smart home, inject it in the domestic television signal and culminating with the development of an Smart TV application able to read the data from the sensors and show it on the screen in a simple, intuitive and dynamic way.The main contribution involves the injection of personalized data in the television signal, demanding the creation of a specific metadata profile for this case.