<p>Isolationsdiagnostik är ett redskap som är av stor betydelse för underhållsoptimering av elektriska anläggningar. Ett av de möjliga mått på isolationsförsämring som kan användas i diagnosticeringssyfte är partiella urladdningar. Det här examensarbetet beskriver ett modelleringsförsök av ett resistivt-kapacitivt nätverk för simulering av partiella yturladdningar i Matlab. Tyvärr blev försöket misslyckat på grund av ett oväntat stort beroende av högspänningskapacitanser på ytresistiviteten. Ytterligare ett försök genomfördes i COMSOL Multiphysics, ett program baserat på finita elementmetoden ämnat för simuleringar av fysikaliska processer. Den huvudsakliga nackdelen med COMSOL Multiphysics modellen är långa simuleringstider. Det visade sig vara möjligt att simulera urladdningar i COMSOL Multiphysics. Här modellerades ytresistansen med hjälp av ett resistivt skikt. Yturladdningar simulerades genom att ändra det resistiva skiktets konduktivitet. Här upptäcks ytterligare ett problem: mycket långa simuleringstider vid användandet av olinjära konduktivitetsuttryck som beror på det elektriska fältet.</p><p>Alla simuleringar, både i Matlab och COMSOL Multiphysics, utfördes på en dator med Intel dual-core processor: 2.13 GHz, 0.99 GB of RAM.</p> / <p>Insulation diagnostics is a very important tool in optimization of electric installations’ maintenance. One of the possible measures of insulation deterioration that can be used for diagnostic purposes are partial discharges. This thesis work describes an attempt to model a resistive-capacitive network for simulating partial surface discharges in Matlab. Unfortunately this attempt proved to be a failure due to an unexpectedly considerable dependency of high voltage capacitances on surface resistivity. Another attempt described here was performed in COMSOL Multiphysics, a finite-element based program for simulation of physical processes. The main drawback with COMSOL Multiphysics model is long simulation times. It proved to be possible to simulate discharges in COMSOL Multiphysics. Here surface resistance was modeled with the help of a resistive layer. Discharges were simulated by changing conductivity of the mentioned layer. Here another problem was discovered: very long simulation times when using non-linear, electric field dependent expressions for conductivity.</p><p>All the simulations, both in Matlab and COMSOL Multiphysics, were performed on a computer with Intel dual-core processor: 2.13 GHz, 0.99 GB of RAM.</p>

Isolationsdiagnostik är ett redskap som är av stor betydelse för underhållsoptimering av elektriska anläggningar. Ett av de möjliga mått på isolationsförsämring som kan användas i diagnosticeringssyfte är partiella urladdningar. Det här examensarbetet beskriver ett modelleringsförsök av ett resistivt-kapacitivt nätverk för simulering av partiella yturladdningar i Matlab. Tyvärr blev försöket misslyckat på grund av ett oväntat stort beroende av högspänningskapacitanser på ytresistiviteten. Ytterligare ett försök genomfördes i COMSOL Multiphysics, ett program baserat på finita elementmetoden ämnat för simuleringar av fysikaliska processer. Den huvudsakliga nackdelen med COMSOL Multiphysics modellen är långa simuleringstider. Det visade sig vara möjligt att simulera urladdningar i COMSOL Multiphysics. Här modellerades ytresistansen med hjälp av ett resistivt skikt. Yturladdningar simulerades genom att ändra det resistiva skiktets konduktivitet. Här upptäcks ytterligare ett problem: mycket långa simuleringstider vid användandet av olinjära konduktivitetsuttryck som beror på det elektriska fältet. Alla simuleringar, både i Matlab och COMSOL Multiphysics, utfördes på en dator med Intel dual-core processor: 2.13 GHz, 0.99 GB of RAM. / Insulation diagnostics is a very important tool in optimization of electric installations’ maintenance. One of the possible measures of insulation deterioration that can be used for diagnostic purposes are partial discharges. This thesis work describes an attempt to model a resistive-capacitive network for simulating partial surface discharges in Matlab. Unfortunately this attempt proved to be a failure due to an unexpectedly considerable dependency of high voltage capacitances on surface resistivity. Another attempt described here was performed in COMSOL Multiphysics, a finite-element based program for simulation of physical processes. The main drawback with COMSOL Multiphysics model is long simulation times. It proved to be possible to simulate discharges in COMSOL Multiphysics. Here surface resistance was modeled with the help of a resistive layer. Discharges were simulated by changing conductivity of the mentioned layer. Here another problem was discovered: very long simulation times when using non-linear, electric field dependent expressions for conductivity. All the simulations, both in Matlab and COMSOL Multiphysics, were performed on a computer with Intel dual-core processor: 2.13 GHz, 0.99 GB of RAM.

This thesis deals with the modeling, analysis and control of a novel brushlessgenerator for wind power application. The generator is named as rotatingpower electronic brushless doubly-fed induction machine/generator (RPEBDFIM/G). A great advantage of the RPE-BDFIG is that the slip power recoveryis realized in a brushless manner. This is achieved by introducing an additionalmachine termed as exciter together with the rotating power electronicconverters, which are mounted on the shaft of a DFIG. It is shown that theexciter recovers the slip power in a mechanical manner, and delivers it backto the grid. As a result, slip rings and carbon brushes can be eliminated,increasing the robustness of the system, and reducing the maintenance costsand down-time of the turbine. To begin with, the dynamic model of the RPE-BDFIG is developed andanalyzed. Using the dynamic model, the working principle of the generatoris understood and its operation explained. The analysis is carried out atspeeds, ±20% around the synchronous speed of the generator. Moreover, thedynamics of the generator due to external load-torque disturbances are investigated.Additionally, the steady-state model is also derived and analyzed forthe machine, when operating in motor mode. As a next step, the closed-loop control of the generator is considered indetail. The power and speed control of the two machines of the generator andthe dc-link voltage control is designed using internal model control (IMC)principles. It is found that it is possible to maintain the stability of thegenerator against load-torque disturbances from the turbine and the exciter,at the same time maintain a constant dc-link voltage of the rotor converter.The closed-loop control is also implemented and the operation of the generatorwith the control theory is confirmed through experiments.In the third part of the thesis, the impact of grid faults on the behaviourof the generator is investigated. The operation of the generator and its responseis studied during symmetrical and unsymmetrical faults. An approachto successful ride through of the symmetrical faults is presented, using passiveresistive network (PRN). Moreover, in order to limit the electrical and mechanicaloscillations in the generator during unsymmetrical faults, the dualvector control (DVC) is implemented. It is found that DVC to a certain extentcan be used to safeguard the converter against large oscillations in rotorcurrents. Finally, for completeness of the thesis, a preliminary physical design ofthe rotating power electronic converter has been done in a finite elementsoftware called ANSYS. The thermal footprint and the cooling capability,with estimates of the heatsink and fan sizes, are presented. Besides, another variant of a rotating electronic induction machine whichis based on the Lindmark concept and operating in a single-fed mode is also investigated. It’s steady-state model is developed and verified through experiments. / <p>QC 20151006</p>

This thesis deals with the steady-state, dynamic and control aspects of new type of brushless configuration of a doubly-fed induction machine in which the slip rings and carbon brushes are replaced by rotating power electronics and a rotating exciter. The aim is to study the stability of this novel configuration of the generator under mechanical and grid disturbances for wind power applications. The derivation, development and analysis of the steady-state model of the brushless doubly-fed induction machine with a rotating excitor and the power electronic converters mounted on the shaft and rotating with it, is studied. The study is performed at rated power of the generator between ±20% slip range. Moreover unity power factor operation between ±20% speed range is also discussed. Furthermore dynamic modeling and control aspects of the generator are also analyzed. The controllers were designed using Internal Model Control principles and vector control methods were used in order to control the generator in a closed-loop system. It is shown that through the use of proper feedback control, the generator behaves in a stable state both at super-synchronous and sub-synchronous speeds. Moreover Low Voltage Ride Through of the generator during symmetrical and unsymmetrical voltage dips is also investigated. Passive Resistive Network strategy is employed for Low Voltage Ride Through of the generator during symmetrical voltage dips. On the other hand, Extended Vector Control is used in order to control the negative sequence currents during unsymmetrical voltage dips. Suppression of negative sequence currents is important as they cause extra heating in the windings and affects the lifetime of the mechanical and electrical components of the generator and system due to oscillations in power and torque. In addition to the above studies a steady-state model of a single-fed induction machine is also developed and investigated where the rotating exciter is removed and the rotor windings are short-circuited through the two rotating power electronic converters. In this way the slip power circulates in the rotor and with the help of the two rotating electronic converters, rotor current is used to magnetize the induction machine thereby improving the power factor. The steady state model is verified through experimental results. / <p>20120914</p> / Brushless Wind Generator with Rotating Power Electronic Converters

Au cours de la dernière décennie, l'apprentissage profond est devenu une composante majeure de l'intelligence artificielle, ayant mené à une série d'avancées capitales dans une variété de domaines. L'un des piliers de l'apprentissage profond est l'optimisation de fonction de coût par l'algorithme du gradient stochastique (SGD). Traditionnellement en apprentissage profond, les réseaux de neurones sont des fonctions mathématiques différentiables, et les gradients requis pour l'algorithme SGD sont calculés par rétropropagation. Cependant, les architectures informatiques sur lesquelles ces réseaux de neurones sont implémentés et entraînés souffrent d’inefficacités en vitesse et en énergie, dues à la séparation de la mémoire et des calculs dans ces architectures. Pour résoudre ces problèmes, le neuromorphique vise à implementer les réseaux de neurones dans des architectures qui fusionnent mémoire et calculs, imitant plus fidèlement le cerveau. Dans cette thèse, nous soutenons que pour construire efficacement des réseaux de neurones dans des architectures neuromorphiques, il est nécessaire de repenser les algorithmes pour les implémenter et les entraîner. Nous présentons un cadre mathématique alternative, compatible lui aussi avec l’algorithme SGD, qui permet de concevoir des réseaux de neurones dans des substrats qui exploitent mieux les lois de la physique. Notre cadre mathématique s'applique à une très large classe de modèles, à savoir les systèmes dont l'état ou la dynamique sont décrits par des équations variationnelles. La procédure pour calculer les gradients de la fonction de coût dans de tels systèmes (qui dans de nombreux cas pratiques ne nécessite que de l'information locale pour chaque paramètre) est appelée “equilibrium propagation” (EqProp). Comme beaucoup de systèmes en physique et en ingénierie peuvent être décrits par des principes variationnels, notre cadre mathématique peut potentiellement s'appliquer à une grande variété de systèmes physiques, dont les applications vont au delà du neuromorphique et touchent divers champs d'ingénierie. / In the last decade, deep learning has become a major component of artificial intelligence, leading to a series of breakthroughs across a wide variety of domains. The workhorse of deep learning is the optimization of loss functions by stochastic gradient descent (SGD). Traditionally in deep learning, neural networks are differentiable mathematical functions, and the loss gradients required for SGD are computed with the backpropagation algorithm. However, the computer architectures on which these neural networks are implemented and trained suffer from speed and energy inefficiency issues, due to the separation of memory and processing in these architectures. To solve these problems, the field of neuromorphic computing aims at implementing neural networks on hardware architectures that merge memory and processing, just like brains do. In this thesis, we argue that building large, fast and efficient neural networks on neuromorphic architectures also requires rethinking the algorithms to implement and train them. We present an alternative mathematical framework, also compatible with SGD, which offers the possibility to design neural networks in substrates that directly exploit the laws of physics. Our framework applies to a very broad class of models, namely those whose state or dynamics are described by variational equations. This includes physical systems whose equilibrium state minimizes an energy function, and physical systems whose trajectory minimizes an action functional (principle of least action). We present a simple procedure to compute the loss gradients in such systems, called equilibrium propagation (EqProp), which requires solely locally available information for each trainable parameter. Since many models in physics and engineering can be described by variational principles, our framework has the potential to be applied to a broad variety of physical systems, whose applications extend to various fields of engineering, beyond neuromorphic computing.