The LHCb detector is one of the four experiments being built to harness the proton-proton collisions provided by the Large Hadron Collider (LHC) at the European Organisation for Nuclear Research (CERN). The data rate expected, when the LHC experiments are fully operational, eclipses that of any previous scientific experiments and has motivated the adoption of a grid computing paradigm to store and process the data. Managing PetaBytes of data in a distributed environment provides a rich set of challenges related to scalability, reliability and performance. This thesis will present the data management requirements for executing the workload of the LHCb collab- oration. We present the systems designed that support all aspects of the grid data management for LHCb, from data transfer, to data integrity, and efficient data access. The distributed computing environment is inherently unstable and much focus has been made on providing systems that are ro- bust and resilient to observed failures.

Denna rapport behandlar behovet av lokalkraft i Vattenfalls transformatorstationer. Tre vanligt förekommande transformatorstationer - stamstation, regionstation och fördelningsstation - har studerats. Alla transformatorstationer behöver lokalkraft för att fungera. Begreppet lokalkraft syftar till den matning som förser en transformatorstations utrustning med el. Apparater som försörjs av lokalkraften är bland annat motorer för brytare, transformatorkylning, kontrollutrustning, värme och belysning. Ok-lindning är ett exempel på en lokalkraftslösning, vilket är det alternativ som Vattenfall vanligtvis använder. Denna lindning är placerad i huvudtransformatorn och består av ett fåtal varv runt järnkärnan. Spänningen från ok-lindningen är aldrig exakt 400 V och därför krävs en matchningstransformator. Rapporten beskriver även kort några andra lokalkraftslösningar samt några tillfälliga effektuttag. Studien undersöker hur stort det teoretiska lokalkraftsbehovet är och vad det är för utrustning som har störst effektförbrukning i de tre stationstyperna. Effekterna delas upp i tre kategorier: ställverk, kontrollanläggning och lokal. Det beräknade lokalkraftsbehovet är i stamstationen 140 kVA, i regionstationen 54 kVA och i fördelningsstationen 25 kVA. Resultaten visar att det teoretiska lokalkraftsbehovet i samtliga studerade stationer är lägre än de standardstorlekar Vattenfall tillämpar som är på 400 kVA, 250 kVA och 100 kVA. / This thesis studies the auxiliary power requirement in Vattenfall's substations. Three common substations - national grid substation, regional grid substation and distribution grid substation - have been studied. All substations need auxiliary power to function properly. Auxiliary power is what powers the substation equipment, for example motor drives, transformer cooling, control equipment, heating and lighting. Auxiliary winding is one solution to supply auxiliary power, which is the option that Vattenfall commonly uses. The winding is placed in the main transformer and consists of a few turns around the iron core. The voltage from an auxiliary winding is never exactly 400 V and a matching transformer is therefore required. The report also briefly describes a few other auxiliary power solutions and a few temporary power consumptions. The study examines the theoretical auxiliary power requirements and what kind of equipment that has the highest power consumption in each of the three types of substations. The power consumption is divided into three categories, switchgear, control equipment and premises. The calculated auxiliary power requirement is 140 kVA for the national substation, 54 kVA for the region substation and 25 kVA for the distribution substation. The results show that the theoretical auxiliary power requirement in all of the studied substations is lower than the standard auxiliary winding sizes used by Vattenfall, which is 400 kVA, 250 kVA and 100 kVA.

Doctor en Ciencias, Mención Computación / Esta tesis apunta a entregar las bases para el desarrollo de los algoritmos de balance de carga para el modelo de objetos activos definido por ProActive en el contexto de las redes a gran escala (grillas). ProActive es un middleware implementado en lenguaje Java, de código abierto, para la progra­mación concurrente, en paralelo, distribuido, y emóvil, poniendo el paradigma de objeto-activo en ejecución. En ProActive, cada objeto activo tiene su propio hilo de control y puede decidir indepen­dientemente en qué orden servir los métodos entrantes, las cuales se almacenan automáticamente en una cola de peticiones pendientes. Para agregar eficacia al paradigma de objetos activos, ProActive proporciona un mecanismo del migración, obteniendo localización automática y trans­parencia mediante el uso de forwarders. La migración viene con un costo de comunicación: un objeto activo debe emigrar con su estado completo, consistiendo en sus peticiones pendientes (lla­madas de método), futuros, y sus objetos pasivos. Por lo tanto, las aplicaciones implementadas con ProActive son sensibles a la latencia. Cuando varios objetos activos con funcionalidad idéntica se despliegan, un algoritmo de bal­ance de carga puede ser utilizado para mejorar el funcionamiento de la aplicación utilizando esa funcionalidad. La carga de trabajo puede ser equilibrada a través de varios objetos activos enviando objetos activos de un procesador altamente cargado a uno menos cargado, o bien robando objetos activos de un procesador altamente cargado por uno menos cargado. El ambiente donde funcionan las aplicaciones implementadas usando el modelo de objetos activos se compone generalmente de grupos múltiples de recursos, por ejemplo, un sistema de máquinas interconectadas por una red local de alta velocidad. Por lo tanto, un algoritmo de balance de carga para objetos activos que pertenecen a una apli­cación paralela fue desarrollado y estudiado, fijando las bases para el desarrollo de los algoritmos de balance de carga para el middleware ProActive. Este primer acercamiento se llama el algoritmo Robin-Hood + Nottingham Sheriff. Este algoritmo fue validado en el contexto de una red de alta escala (sobre 1.000 nodos) mediante simulaciones, utilizando nuestro modelo de las grillas basa­dos en la observación y la medición de lo que consideramos las características dominantes para el balance de objetos activos: capacidad de procesamiento y latencia entre recursos. Finalmente, presentamos los contratos de acoplamiento para el despliegue de aplicaciones par­alelas, su forma de utilización en el contexto de balance de carga, por ejemplo, elegir entre un planificador local y el balanceador de carga de ProActive.

Buildings account for approximately 40% of the energy demand and 33% of the total greenhouse gas emissions in the European Union. Accordingly, there are several efforts that target energy efficiency in buildings both at the European and Swedish levels. The role of buildings in climate change mitigation, however, is not limited to energy savings. Buildings are expected to become key elements of the future smart energy systems by supplying and using energy in a more flexible way. Reducing the energy demand in buildings effectively and shifting the role of buildings in energy systems from ‘passive’ consumers to ‘active’ prosumers, however, require close interaction and cooperation between the energy and buildings sectors. Based on the data collected from interviews and a web survey, this doctoral thesis investigates the relationship between the energy and buildings sectors in Sweden at the inter-company level, presents key stakeholder views on smart energy features in buildings and investigates the opportunities and barriers for their adoption in Sweden and Hong Kong. The results of this thesis suggest a potential for improving the cooperation between the Swedish energy and buildings sectors, which was identified to be influenced by the following factors: district heating monopolies; energy efficiency efforts in the buildings sector; unsuccessful technology-neutrality of the building regulations; self-generation systems in buildings; and energy use patterns. Shifting the focus from self-gains to mutual gains appears crucial to strengthen the inter-sectoral cooperation, as there are several opportunities for achieving mutually beneficial solutions for the two sectors. This would, however, require significant changes in current practices and business models as well as the introduction of new technologies, which would allow for a more flexible energy supply and use. Accordingly, technologies that target flexible energy use in buildings are considered the most important smart energy features in buildings. The current high costs of technologies, such as home automation and smart electrical appliances, however, create the strongest barrier to adoption. Therefore, the introduction of new business and ownership models and the elimination of the institutional and regulatory barriers are crucial to achieve a wide-scale development of smart energy features in buildings. The results from Hong Kong suggest that institutional and regulatory barriers can particularly create strong hinders to the adoption of technologies. It is possible to achieve more sustainable energy systems, where buildings are active elements of networks that supply and use energy in a more flexible and ‘smarter’ way. Cooperation between the energy and buildings sectors can play a key role in the adoption of smart energy features in buildings and pave the way for the smart built environment of the future.

The relationship between structure, dynamics, and function of neural networks in nervous systems is still an open question in the neuroscience community. Nevertheless, for certain areas of the mammalian nervous system we do have sufficient data to impose constraints on the organisation of the network structure. One of these areas is the medial entorhinal cortex which contains cells with hexagonally repeating spatial receptive fields, called grid cells. Another intriguing property of entorhinal cortex and other cortical regions is a population oscillatory activity, with frequency in the theta (4-10 Hz) and gamma (30-100 Hz) range. This leads to a question, whether these oscillations are a common circuit mechanism that is functionally relevant and how the oscillatory activity interacts with the computation performed by grid cells. This thesis deals with applying the continuous attractor network theory to modelling of the microcircuit of layer II in the medial entorhinal cortex. Based on recent experimental evidence on connectivity between stellate cells, and fast spiking interneurons, I first develop a two-population spiking attractor network model that is capable of reproducing the activity of a population of grid cells in layer II. The network was implemented with exponential integrate and fire neurons that allowed me to address both the attractor states and the oscillatory activity in this region. Subsequently, I show that the network can produce theta-nested gamma oscillations with properties that are similar to the cross-frequency coupling observed in vivo and in vitro in entorhinal cortex, and that these theta-nested gamma oscillations can co-exist with grid-like receptive fields generated by the network. I also show that the connectivity inspired by anatomical evidence produces a number of directly testable predictions about the firing fields of interneurons in layer II of the medial entorhinal cortex. The excitatory-inhibitory attractor network, together with the theta-nested gamma oscillations, allowed me to explore potential relationships between nested gamma oscillations and grid field computations. I show, by varying the overall level of excitatory and inhibitory synaptic strengths, and levels of noise, in the network, that this relationship is complex, and not easily predictable. Specifically, I show that noise promotes generation of grid firing fields and theta-nested gamma oscillations by the model. I subsequently demonstrate that theta-nested gamma oscillations are dissociable from the grid field computations performed by the network. By changing the relative strengths of interactions between excitatory and inhibitory neurons in the network, the power and frequency of the gamma oscillations changes without disrupting the rate-coded grid field computations. Since grid cells have been suggested to be a part of the spatial cognitive circuit in the brain, these results have potential implications for several cognitive disorders, including autism and schizophrenia, as well as theories that propose a cognitive role for gamma oscillations.

Master of Science / Department of Electrical and Computer Engineering / Anil Pahwa / This research project builds upon previous work related to intelligent and energy efficient lighting in modern street and outdoor lighting systems. The concept of implementing modern smart grid technologies such as the proposed Street & Outdoor Lighting Intelligent Monitoring System (SOLIMS) is developed. A random sample of photocells from two municipal electric power systems is used to collect data of the actual on/off times of random photocells versus Civil Twilight (sunrise/sunset) times. A business case was developed using the data collected from the observations to support an electric utility company’s implementation of SOLIMS as an alternative to current operations. The goal of the business case is to demonstrate energy and capacity savings, reduced maintenance and operating costs, and lower carbon emissions.

Alpha matting is an important topic in areas of computer vision. It has various applications, such as virtual reality, digital image and video editing, and image synthesis. The conventional approaches for alpha matting perform unsatisfactorily when they encounter complicated background and foreground. It is also difficult for them to extract alpha matte accurately when the foreground objects are transparent, semi-transparent, perforated or hairy. Fortunately, the rapid development of deep learning techniques brings new possibilities for solving alpha matting problems. In this thesis, we propose a residual convolutional grid network for alpha matting, which is based on the convolutional neural networks (CNNs) and can learn the alpha matte directly from the original image and its trimap. Our grid network consists of horizontal residual convolutional computation blocks and vertical upsampling/downsampling convolutional computation blocks. By choosing different paths to pass information by itself, our network can not only retain the rich details of the image but also extract high-level abstract semantic information of the image. The experimental results demonstrate that our method can solve the matting problems that plague conventional matting methods for decades and outperform all the other state-of-the-art matting methods in quality and visual evaluation. The only matting method performs a little better than ours is the current best matting method. However, that matting method requires three times amount of trainable parameters compared with ours. Hence, our matting method is the best considering the computation complexity, memory usage, and matting performance.

Com o aumento no uso de aglomerados e grades de computadores, cresce o interesse no estudo de comunicações entre processadores. Em um computador paralelo dedicado, ou em uma rede local homogênea, o tempo de comunicação é geralmente modelado de forma similar, independente de quais processadores estão se comunicando. Em uma rede onde os links entre os computadores são heterogêneos, computadores mais próximos tendem a apresentar menor latência e maior largura de banda do que computadores distantes. Além disso, a largura de banda agregada é diferente dependendo do número de conexões simultâneas existentes entre dois aglomerados distantes. Neste trabalho estudaremos a troca completa de mensagens de tamanhos diferentes entre aglomerados interligados por backbones. Proporemos um novo algoritmo de comunicação baseado em algoritmos conhecidos, apresentaremos simulações de escalonamentos dos algoritmos estudados para esta rede multi-aglomerado e analisaremos os resultados destas simulações. / The growth in popularity of clusters and computational grids caused an increase in the interest in studying interprocessors communications. The comunication time in a dedicated parallel computer or in a local homogeneous network is modeled in a similar way, regardless of which processors are communicating. In a network with heterogeneous links, closer computers generally have lower latency and larger bandwidth than wide area computers. Besides, the aggregated bandwidth depends on the number of simultaneous connections between two wide area clusters. In this work we study the complete exchange of messages of different sizes between interconnected clusters using a backbone. We propose a new comunication algorithm based on known algorithms, we present some scheduling simulations of the studied algorithms in this multi-cluster network and we present the results analysis of these simulations.

Doctor of Philosophy / Department of Electrical and Computer Engineering / Anil Pahwa / Power distribution systems are transitioning from traditional centralized-control distribution grids to the modern distribution grids that are more customer-interactive and include microgrids (MGs) as well as various unpredictable and multi-scale distributed energy resources (DERs). However, power fueled by renewable DERs such as wind and solar is highly variable and high penetration of renewable DERs in distribution system may potentially degrade the grid reliability and power quality. Moreover, the growth of such generation sources will increase the number of variables and cause scalability concerns for distribution system operators (DSOs) in handling system optimization problems. Further, with development of MGs, DSO and MG may have different owners and schedule renewable and non-renewable DERs based on their own economic rules and policies while secure and economic operation of the entire system is necessary. The widespread integration of wind and solar and deployment of MGs in distribution system make the task of distribution system operation management quite challenging especially from the viewpoint of variability, scalability, and multi-authority operation management. This research develops unique models and methodologies to overcome such issues and make distribution grid operation, optimization and control more robust against renewable intermittency, intractability, and operation complexity. The objectives of this research are as follows: 1) to develop a three-phase unbalanced large-scale distribution system to serve as a benchmark for studying challenges related to integration of DERs, such as scalability concerns in optimization problems, incremental power losses, voltage rise, voltage fluctuations, volt/var control, and operation management; 2) to develop a novel hierarchical and multilevel distributed optimization for power loss minimization via optimal reactive power provisioning from rooftop PVs which addresses the scalability issues with widespread DER integration in large-scale networks; 3) to develop a dynamic operational scheme for residential PV smart inverters to mitigate the fluctuations from rooftop PV integration under all-weather-condition (fully sunny, overcast and transient cloudy days) while increasing network efficiency in terms of power losses, and number of transformer load tap changer (LTC) operation; 4) to develop a stochastic energy management model for multi-authority distribution system operating under uncertainty from load and wind generation, which is able to precisely account interactions between DSO and MGs.

Tese de mestrado integrado. Engenharia Informática e Computação. Faculdade de Engenharia. Universidade do Porto. 2010