<p>This work is focused on direct-driven permanent magnets synchronous generators (PMSG) with diode rectifiers for use in offshore wind turbines. Reactive compensation of the generator, power losses and control of the generator are studied. Configurations for power transmission to onshore point of common connection are also considered. Costs, power losses, reliability and interface with the PMSG are discussed. The purpose of the laboratory tests and simulations are to learn how a PMSG with diode rectifier behaves. A 55kW PMSG is tested in "Vindlabben", with and without reactive compensation. The generator current and generator voltage are measured and the total harmonic distribution (THD) of the current and the voltage are calculated. The results are compared to simulations on an equal generator in PSCAD/EMTDC. A 2MW PMSG is also simulated to compare parallel and series compensation and to find how the generator efficiency varies with the wind speed. The generator is also simulated with constant DC-link voltage and varying local wind to find how much the turbine and generator efficiency decreases when a Cluster step-up configuration is used. The DC-link voltage is in this case equal for parts of the wind farm or the whole wind farms. A 3MW ironless PMSG with very low synchronous reactance is simulated to find how this generator behaves with a diode rectifier. The laboratory tests and PSCAD simulations show that the maximal generator power increases when reactive compensation of the generator is used. The measured and simulated generator voltage and current shapes are found to be approximately equal. Series compensated PMSGs have lower generator current rms and lower current THD than parallel compensated PMSGs when the synchronous reactance is large. Therefore, the generator losses are 2-15% lower and the diode rectifier losses are 0-1% lower, depending on the wind speed. The diode rectifier losses are lower than 1%. The losses can be reduced even more if the diodes are connected in parallel. If a Cluster step-up configuration is used, the turbine efficiency is reduced by 3-4%. %The generator efficiency is more or less the same. The ironless PMSG has a low synchronous reactance and reactive compensation is not needed because the reactive power produced by the generator is low. Parallel connected capacitors have no positive effect and series connected capacitances must be very large and can therefore not be used. The generator current THD is very large when no reactive compensation is used. However, the current THD can be reduced by connecting an inductance to the DC-link. Cluster step-up, two-step DC/DC system, turbine step-up and series connected wind turbine are the most relevant layouts of the wind farms transmission system discussed in this thesis. The cluster step-up system has low power losses since only one large DC/DC converter is used. Also, the power equipment in the turbine is very reliable. However, the turbine efficiency is reduced since the generator torque and generator speed could not be controlled for one specific turbine. The other transmission systems require DC/DC converters in the turbines and they are therefore probably not as reliable. The total cost is crucial for the chose of the transmission system. Further cost accountings for the different DC systems are needed.</p>

<p>This report aims to describe the work done in order to develop a solution for automation of failure detection and performance of rapid voltage surges. The automated system was developed with basis in the experimental circuit developed at NTNU/SINTEF. The experimental circuit relied on manually performing rapid voltage surges and detecting cable failure. Meaning that unknown variables may have been introduced due to the inequality of each grounding. The realized automation system consisted of two parts. The first part was a switch that was able to withstand the voltage applied and able to perform the rapid voltage surges. The second part was a program that controlled the switch and regularly monitored the condition of the cable object, indicating failure when electrical treeing had caused breakdown in the cable insulation. Two solutions for switches were examined. One switch utilized an arm moving between two electrodes for grounding the objects, using textile bakelite as insulation and support frame for the electrodes. The other switch utilized a sphere gap with air as insulation and an electric field strength close to the breakdown strengt of the air. Groundings in the sphere gap switch were performed using a striker to greatly enhance the electric field and cause a spark between the spheres, causing a transition from a highly insulating state to a highly conductive state. As the experimental circuit had a large current limiting resistance, the experimental circuit was sensitive to leakage currents as they set up a voltage over the resistance, instead of the cable objects. Leakage currents in the switches were examined and it was found that the arm switch had leakage current unacceptable for use in the experiment. Similarly, leakage currents in the sphere gap switch were examined and found to be acceptable for use in the experiment. The behaviour of cable objects that, due to electrical treeing had developed insulation failure was examined and measurement parameters for detecting cable failure were found. These parameters were implemented using LabView as a platform for creating a a control program and interface to the instruments connected to the switch. Having fully created the automated system, the system was used in testing cable objects and was found able to inflict and detect a cable failure due to electrical treeing. However, compared to previous results, a lower number of groundings were needed, which may indicate that the introduction of the sphere gap switch has altered the circuit in a manner such that new results will differ from former results.</p>

<p>In this work, a STAtic synchronous COMpensator (STATCOM) with energy storage system for wind power application has been treated. This device was proposed as a mean to improve voltage stability and power transmission by offering reactive as well as active power compensation. The work focuses on the converter topology of the STATCOM part and the control system. Further on, the energy storage system needed for this application was designed, including the choice of energy storage, its size and the interface/control system. The STATCOM, reactive part of the compensator was based on a voltage source converter (VSC), using a vector control. Its purpose was to maintain a stable grid voltage. For active compensation of wind power, a bank of super capacitors for energy storage system, SCESS, was used in this thesis. The super capacitor bank size was estimated, based upon the short term fluctuations in wind power. These fluctuations are results of contstructional factors of the turbines, variations and turbulence in the wind. The super capacitor bank was interfaced with the DC-bus of the STATCOM with a normal half-bridge buck-boost converter, to control the voltage level of the bank while maintaing a constant DC-bus voltage for good switching operation in the VSC. The control system for the active power compensation part was implented as a cascaded PI-control, compromising an inner current control loop, and an outer power control loop. The outermost loop included a dynamical power reference, based on the actual power transfer in the grid. This reference is supposed to assure that the controller is only compensating small fluctuations, while larger changes are left for other means, for instance controlled hydro power. The designed system was implemented in EMTDC/PSCAD. A small model, including one wind turbine, a weak grid and the STATCOM/SCESS was used in the simulations. With regards to the reactive- and active power compensation, the results were promising. However, the dynamical power reference could be of a better quality, as it does not take into account the losses in the STATCOM/SCESS, and thereby is inacurate regarding the amount power fed to/from the super capacitor bank. In addition, a small STATCOM model was realised in the laboratory. The results from the practical work showed the same general patterns as the simulations.</p>

<p>Artificial intelligence (AI) is a collective term for several computing techniques. They have in common that they use non-linear algorithms and that they are inspired from different processes in the nature, in particular how human beings make decisions. The use of AI for optimal operation of a stand-alone power plant has been investigated. This includes prediction, estimation, optimization and control. A presentation of some relevant AI techniques are given. A comparison with classical approaches such as for example PI control was made. The new techniques that were investigated proved to be very powerful and should be used more frequently than it is used today. AI techniques are especially promising for supervisorial control, but can also be used to control converters directly. A controller for a DC/DC boost converter was developed. It proved to be significantly better than a classical PI controller. Whether the computing time is shorter or faster than for classical approaches depends on the application. Compared to PI controllers the AI algorithms have a long computing time. Compared to classical wind power prediction techniques on the other hand AI techniques are very fast. A disadvantage with AI is the lack of rules for deciding the inner structure of the algorithms.</p>

<p>To create the ultimate wireless instrumentation unit for down-hole applications high temperature electronics with very high reliability is needed. It is possible to use ordinary bulk-CMOS devices at temperature up to 175 ⁰C, but the lifetime at these temperatures is to low for a down-hole instrumentation unit. An alternative is to use s Silicon on Insulator process under the fabrication of the semiconductors. The SOI process is a fabrication process where there is buried a oxide layer in the silicon wafer, and thus allowing higher breakdown voltage and/or lower current leakage. The low current leakage allows the semiconductors to be used at higher junction temperature. SOI devices that are commercial available off-the-shelf as a expected lifetime for at least 5 years at 225 ⁰C and thus much lower at junction temperatures below 200 ⁰C. The SOI technology can then be used together with hybrid circuits using ceramic substrate as a replacement for organic PCB and thick-film technology for the passive devices. A package like this gives a system with high reliability both toward high temperature operation and lifetime. The main limitation in the high temperature design is the availability off the larger capacitors; the limitation for high temperature stacked capacitors is 200 ⁰C. The converters designed are the standard step-up and step-down switch-mode power supplies. The converters are designed with current mode control; current mode control is used because of the advantage that comes with it. One off the advantages is the possibility to limit the inductor current; another advantage is the possibility to use constant current charging for the battery. When designing the SOI devices for high temperature operation it is difficult to achieve high enough breakdown voltage. With this in mind, the high temperature converter is designed with series coupled transistors to achieve high enough breakdown voltage for high voltage operation. The transistors have always some small perturbations in their specifications, this has to be considered when connecting transistors in series. This perturbations in for example turn-off speed makes an uneven voltage sharing; this is solved by connecting suitable capacitors in parallel with the switches to maintain an even voltage sharing.</p>

<p>This thesis offers a short overview of the most important stator mounted permanent magnet machines, with a closer look on the FSPM design. A FSPM machine have been built and tested as a generator, to get a better understanding of the machine concept. The focus of the work have been on the well documented 12/10 (Stator teeth/ Rotor teeth) design while the novel 12/14 pole design have also been tested, as a rotor change is the only difference between the two designs. The machine have been simulated in COMSOL, where inductances, back emf and cogging have been found and compared with the measured results.</p>

<p>To support the teaching in digital signal processing and control in power electronics a laboratory setup of a PV (photovoltaic) converter system is currently under development at NTNU. The equipment consists of a general reconfigurable power converter and a DSP (Digital Signal Processor) control card with system software for software development and testing. The finished system is intended for implementation in an African University to be used in teaching of PV systems. The power converter stage will be a dual stage consisting of a DC-DC converter and a DC-AC inverter connected to the grid through a transformer stage. For this particular master thesis the input stage comprising the PV panel and the DC-DC converter will be of main focus. A control design will be developed, comprising voltage mode control (with feedback from the input of the converter) and Maximum Power Point Tracking (MPPT). The DC link voltage level is set to be 48 V, while the input voltage will vary from 0 to 45 V. In the experiments the setup will consist of DC source simulating the PV-panel, a DC-DC converter and an electronic load representing the grid connection through an inverter and a transformer. The DC-DC converter was built and tested in a previous master project and can be configured as a buck, boost or buck-boost converter. For this thesis the boost topology was chosen, as this topology is the one most frequently used in PV systems. The control was implemented through C code programming. A regular voltage mode controller was developed and tuned through utilization of Ziegler-Nichols ultimate sensitivity method. At first a P-controller was implemented, but it was not able to cancel out the error between the reference voltage and the input voltage. This was expected, and an integral part was added to form a PI-controller. Now the closed loop control of the system turned out to be rather good for the whole range of the input voltage. The MPPT algorithm Perturb & Observe was chosen to track the maximum power point of operation. The MPPT was tested for both step changes in irradiance and temperature levels. When varying the irradiance level the current was the parameter most affected. Even though the MPP was tracked rather well there was uncertainty regarding the MPPT algorithm capability since the voltage was only exposed to minor changes. When the temperature was changed, the voltage was affected in higher degree. The MPPT was able to track the MPP rather well, and tracking in the wrong direction only happened right after a step change. In real life the temperature will normally not change in steps, so this test was said to be done under extreme conditions.</p>

<p>Several inverter topologies for use with power conditioning of photovoltaic modules were investigated with both national and international requirements in mind, as well as also practical challenges and the ability to be user friendly for ordinary people. A good inverter topology should also be low cost, have a high efficiency and have a good output power quality. In addition several filter possibilities was investigated, and it was concluded that the LCL-filter was the best for the given conditions, since it attenuated the unwanted frequencies the best with relative small filter parameters. Five different inverter topologies was then presented and investigated: A hybrid multilevel inverter, a full bridge inverter, a series resonant buck-boost inverter, a flyback converter with unfolding H-bridge inverter and a series resonant converter with unfolding H-bridge inverter. After an investigation of the above mentioned criteria, two of the inverter topologies, the H-bridge inverter and the hybrid multilevel inverter, were considered better than the rest for the given requirements and purposes. These were then closer analysed with the computer simulation programs SIMULINK and SPICE in order to find quantitative arguments about which topology was the best under the above mentioned conditions. Filter parameters were also quantified. From this it was found that the hybrid multilevel inverter was 0.5 to 1 percent point more effective than the H-bridge, at the same time the total harmonic distortion was significantly better, approximately five to ten times better than the H-bridge inverter’s total harmonic distortion. This means that the hybrid multilevel inverter may have a considerably cheaper filter. Both of these factors contributed so that the hybrid multilevel inverter was regarded the better topology and this topology was selected for further tests. The last simulation was about finding good switches to equip the hybrid multilevel inverter with. Here it was found that the decisive factor for the low voltage bridge was quick switches, whereas for the high voltage bridge it was more important to have switches with low resistance when turned on. The chosen switches were STY60NK30Z and BSC520N15NS3 G for the low voltage bridge. In addition it was meant to perform a laboratory experiment with the selected topology, but because of a delay with the deliverance, the test object did not arrive at time, so the experiment could not be done.</p>

<p>The General Purpose 2D Electromagnetic Tool, GP2DET, at Voith Siemens Hydro Power Generation GmbH is a tool developed for simulation and calculation of electrical parameters in large hydro power machines. In this tool, the stator windings have been modelled as massive conductor blocks, and current has been impressed on these conductors as current density for load situations. The scope for the work in this thesis has been to implement an automated procedure for a detailed modelling of the stator windings, testing of the procedure and analyses of losses from eddy currents and circulating currents. A group of macros was written in ANSYS APDL for the reworking of existing simulation models. The macros automatically rework models created in GP2DET, and replaces massive stator conductors with detailed stranded conductors. 17 alternatives for which conductors that shall be replaced were implemented. Bar- and coil windings are supported, and bar windings can be modelled with Roebel-transposition. Some detailing alternatives allow replacement of coils or coil groups. For these alternatives, automatic connection of each strand in the stranded coil sides is carried out. All strands are connected uniquely within a coil, and can be connected blockwise between coils and coil groups if desired. Transposition is available within a coil, between coils, and between coil groups. One type of transposition was implemented. The 17 detailing alternatives were tested on three different machines; two with bar windings and one with coil windings. For these machines the procedure executed faultlessly.</p>

<p>This Master thesis studies losses in superconductors. Losses arise when the superconductor carries alternating current or is placed in an alternating magnetic field. As most power applications involve at least one of the two, loss mechanisms and magnitudes are important when examining the possibility of making superconductor systems competitive to conventional power systems. There are two parts to the task at hand. The first part is a literature study on superconductivity and AC losses in superconductors. A division between two general types of superconductors is presented; type I and II. The Bean model for AC losses in type II superconductors is described, together with equations giving the power law between generated losses and applied magnetic field. In the Bean model, the losses are proportional to the applied field cubed below a limit called the penetration field. Above it, the exponent changes to one. Losses due to coupling of filaments are also treated. The measuring setup used in the AC loss experiments is calorimetric, and the principle behind the method is presented. The superconductor used, magnesium diboride (MgB2), is introduced. The literature study is concluded with short résumés on other AC loss studies done on MgB2, and studies done on one type of high temperature superconductor. The second part is measuring AC losses due to an applied alternating magnetic field in two superconductor samples from different manufacturers. Specific information on the two samples, details on the measuring system, preparations and the measuring procedure is described. The logged data and equations used when processing the results are also listed. Measurements have been performed at six different temperatures; 25, 28.5, 30, 31.5, 35 and 45 K. The magnitude of the applied magnetic field was varied between 3 and 150 mT. Both parallel and perpendicular field directions were applied. Generated losses lead to a temperature increase in the superconductor. The rise in temperature was detected as increased resistivity of a thin copper wire glued onto the sample, as the copper resistivity is temperature dependent. The obtained results are examined in double logarithmic (loglog) and normal axis diagrams, where the main aim is to find loss slopes and penetration fields at the different temperature levels, and to compare these to the Bean model loss equations. In addition, the results are compared to theoretical loss equations for cylindrical conductor geometry. This is done in order to look at the accuracy of the fittings and to compare the penetration fields obtained here to the ones found in loglog diagrams. The results have also been compared to various studies on MgB2 and other superconductor types. The measured loss slopes at fields below the penetration field, found from loglog diagrams, do not fit the Bean model. The slopes are here lower than the applied field cubed. At fields greater than the penetration field, losses are proportional to the applied field, as in the Bean model. Two reasons for the deviations have been discussed; measuring errors and losses being coupling losses. Even if the measuring errors may be considerable due to human reading errors, they would have to be systematic for the losses to fit the Bean model. This is the reason why measuring errors are seen as unlikely to be the grounds for the non-fitting results. The results do also not fit the coupling loss slope and as only one field frequency has been used, the obtained results are not enough to support or reject this theory. Due to the deviation from the Bean model loss slopes the curve fittings to the cylindrical conductor loss equations were mostly poor, as they have the same loss slopes as the Bean model. The penetration fields found from loglog and normal axis diagrams and the curve fitting are not equal. It is establish that the ones found from loglog diagrams should be used. Two of the other studies done on MgB2, which have been summarised in the thesis, fit the Bean model, and the last case did not. The authors found no explanation to the non-fitting results, and have ruled out coupling losses as a viable reason. Studies on the other type of superconductor also represented both cases. Here, some non-fitting results were explained by coupling losses. When comparing loss magnitudes, only one of the samples used in these experiments had as low results as found in two other studies.</p>