There is a trend for offshore wind farms to move further from the point of common coupling to access higher and more consistent wind speeds to reduce the levelised cost of energy. To accommodate these rising transmission distances, High Voltage Direct Current (HVDC) transmission has become increasingly popular. However, existing HVDC wind farm topologies and converter systems are ill suited to the demands of offshore operation. The HVDC and AC substations have been shown to contribute to more than 20% of the capital cost of the wind farm and provide a single point of failure. Therefore, many wind farms have experienced significant delays in construction and commissioning, or been brought off line until faults could be repaired. What is more, around 75% of the cost of the HVDC and AC substations can be attributed to structural and installation costs. Learning from earlier experiences, industry is now beginning to investigate the potential of a modular approach. In place of a single large converter, several converters are connected in series, reducing substation individual size and complexity. While such options somewhat reduce the capital costs, further reductions are possible through elimination of the offshore substations altogether. This thesis concerns the design and evaluation the Hybrid HVDC Transformer, a high power, high voltage, DC transformer. This forms part of the platform-less (i.e. without substations) offshore DC power collection and distribution concept first introduced by the Offshore Renewable Energy Catapult. By operating in the medium frequency range the proposed Hybrid HVDC Transformer can be located within each turbine’s nacelle or tower and remove the need for expensive offshore AC and DC substations. While solid state, non-isolating DC-DC transformers have been proposed in the literature, they are incapable of achieving the step up ratios required for the Hybrid HVDC transformer [1]– [3]. A magnetic transformer is therefore required, although medium frequency and non-sinusoidal operation does complicate the design somewhat. For example, inter-winding capacitances are more significant and core losses are increased due to the added harmonics injected by the primary and secondary converters [1], [2]. To mitigate the impact of these complications, an investigation into the optimal design was conducted, including all power converter topologies, core shapes and winding configurations. The modular multilevel converter in this case proved to be the most efficient and practical topology however, the number of voltage levels that could be generated on the primary converter was limited by the DC bus voltage. To avoid the use of pulse width modulation and hence large switching losses, a novel MMC control algorithm is proposed to reduce the magnitude of the converter generated harmonics while maintaining a high efficiency. The development and analysis of this High Definition Modular Multilevel Control algorithm forms the bulk of this thesis’ contribution. While the High Definition Modular Multilevel Control algorithm was developed initially for the Hybrid HVDC Transformer, analysis shows it has several other potential applications particularly in medium and low voltage ranges.

The Transformer model is a very recent, fast and powerful discovery in neural machine translation. We experiment with multi-task learning for enriching the source side of the Transformer with linguistic resources to provide it with additional information to learn linguistic and world knowledge better. We analyze two approaches: the basic shared model with multi-tasking through simple data manipulation, and multi-decoder models. We test joint models for machine translation (MT) and POS tagging, dependency parsing and named entity recognition as the secondary tasks. We evaluate them in comparison with the baseline and with dummy, linguistically unrelated tasks. We focus primarily on the standard- size data setting for German-to-Czech MT. Although our enriched models did not significantly outperform the baseline, we empirically document that (i) the MT models benefit from the secondary linguistic tasks; (ii) considering the amount of training data consumed, the multi-tasking models learn faster; (iii) in low-resource conditions, the multi-tasking significantly improves the model; (iv) the more fine-grained annotation of the source as the secondary task, the higher benefit to MT.

Large capacity high voltage power transformers are one of the most expensive items of equipment in an electrical power network. Power utilities can ill-afford breakdown of transformers, especially, in a deregulated scenario. The consequences of such a failure are well known. Under these circumstances, utilities have figured-out that condition-based monitoring and diagnosis is worth pursuing, in spite of increased expenditure. Thus, monitoring and diagnosis is an integral part of operation and maintenance. Mechanical forces generated during short-circuits is the main cause leading to displacement/deformation of windings. Frequency response measurements have attained worldwide acceptance as a highly sensitive monitoring tool for detecting occurrence of such events. This is evident from the fact that customized commercial equipment are available (popularly called FRA or SFRA instruments), and with recent introduction of an IEEE draft trial-use guide for application and interpretation of frequency response analysis. Once a damage is detected, the next task is to identify its location along the winding and, if possible, determine its extent of severity. Understandably, these two tasks are best achieved, without disassembling the transformer and should ideally be based on off-line and on-site terminal measurements. In this regard, literature analysis reveals that recent research efforts have successfully demonstrated possibilities of using frequency response data for localization of discrete change in windings. This is indeed noteworthy, in spite of one major drawback. This pertains to excessive computing time needed to synthesize large-sized ladder-network, which automatically limits its practical use. Keeping these issues in mind, a research was initiated to find alternatives. The primary objective of this thesis is to examine the use of- • Complexnetwork-function-lociforlocalizationofadiscretechangeinasingle,isolatedtransformerwinding,basedonterminalmeasurements It goes without saying that the proposed method should be non-invasive, simple, time-efficient and overcome drawbacks in the earlier approach. A brief summary of the proposed method follows- This thesis presents a different approach to tackle the problem of localization of winding deformation in a transformer. Within the context of this thesis, winding deformation means, a discrete and specific change imposed at a particular position on the winding. The proposed method is based on the principle of pre-computing and plotting the complex network-function-loci (e.g. driving-point-impedance) at a selected frequency, for a meaningful range of values for each element (increasing and decreasing) of the ladder network. This loci diagram is called the nomogram. After introducing a discrete change (to simulate a deformation), the driving-point-impedance (amplitude and phase) is measured again .By plotting this single measurement on the nomogram, it is straightforward to estimate the location and identify the extent of change. In contrast to the earlier approach (wherein the entire ladder-network had to be synthesized for every new measurement), the proposed method overcomes the drawbacks, is non-iterative and yields reasonably accurate localization. Experimental results on a model coil and two actual transformer windings (continuous-disc and interleaved-disc) were encouraging and demonstrate its potential. Further details are presented in the thesis.

Investment decisions on electric power networks have developed to balance network functionality and cost efficiency by analyzing the main risks associated with network operation. Ageing infrastructures, like large power transformers in particular, aggravate the stress of management, because the failure of a power transformer could cause power supply interruption, network reliability reduction, large economic losses and also environment impacts. Transformer asset management is therefore aimed to develop a cost-efficient replacement strategy to get the most usage of transformers. The main objective of this thesis is to understand how UK National Grid transformer assets failure trend can be used, as the engineering evidence to help make financial decisions related to transformer replacements. The studies in this thesis are implemented via two main approaches. First statistical analyses methods are undertaken. This approach is realized to be non-optimal, because the transformer failure mechanism at the normal operation stage is different from that when transformers are aged. Secondly, the transformer physical ageing model is used to estimate thermal lifetimes under the ageing failure mechanism. In conjunction with the random hazard rate obtained by statistical analyses, the actual National Grid transformer population failure hazard with service age is derived. Statistical analyses are carried out based on the ages of National Grid failed and in-service transformers. Transformer lifetime data are fitted into various distribution models by the least square estimator (LSE) and maximum likelihood estimator (MLE). Statistics are however powerless to suggest the population future failure trend due to their intrinsic limitations. National Grid operational experience actually indicates a stable and low value of the random failure hazard rate during the transformer early operation ages. The engineering knowledge however suggests an ageing failure mechanism exists which corresponds to an increasing hazard in the future. Transformer lifetime under ageing failure mechanism is conservatively indicated by its thermal end-of-life corresponding to a specific level of insulation paper mechanical strength. By analyzing National Grid scrapped transformers’ lowest degree of polymerization (DP), these transformers are estimated to have deteriorated at different rates and their thermal lifetimes distribute over a wide age range. The limited number of scrapped transformers cannot adequately indicate the ageing status of the whole population. A transformer’s thermal lifetime is determined by its loading condition, thermal design characteristics and installation site ambient temperature. However, these input data are usually incomplete for an individual transformer.A simplified approach is developed to predict the National Grid in-service transformer’s thermal lifetime by using information from scrapped transformers. The in-service transformer population thermal hazard curve under ageing failure mechanism can thus be obtained.Due to the independent effect from transformer random failure mechanism and ageing failure mechanism, the National Grid transformer population actual failure hazard curve with age is therefore derived as the superposition of the random failure hazard and the thermal hazard. Transformer asset managers are concerned about the knee point age, since aged transformer assets threaten network reliability and the transformer replacement strategy needs to be implemented effectively.

Power transformers are a key element in the transmission and distribution of electrical energy and as such need to be highly reliable and efficient. In power system networks, transformer core saturation can cause system voltage disturbances or transformer damage or accelerate insulation ageing. Low frequency switching transients such as ferroresonance and inrush currents, and increasingly what is now known as geomagnetic induce currents (GIC), are the most common phenomena to cause transformer core saturation. This thesis describes extensive simulation studies carried out on GIC and switching ferroresonant transient phenomena. Two types of transformer model were developed to study core saturation problems; one is the mathematical transformer magnetic circuit model, and the other the ATPDraw transformer model. Using the mathematical transformer magnetic circuit model, the influence of the transformer core structure on the magnetising current has been successfully identified and so have the transformers' responses to GIC events. By using the ATPDraw transformer model, the AC system network behaviours under the influence of the DC bias caused by GIC events have been successfully analysed using various simulation case studies. The effects of the winding connection, the core structure, and the network parameters including system impedances and transformer loading conditions on the magnetising currents of the transformers are summarised. Transient interaction among transformers and other system components during energisation and de-energisation operations are becoming increasingly important. One case study on switching ferroresonant transients was modelled using the available transformer test report data and the design data of the main components of the distribution network. The results were closely matched with field test results, which verified the simulation methodology. The simulation results helped establish the fundamental understanding of GIC and ferroresonance events in the power networks; among all the influential parameters identified, transformer core structure is the most important one. In summary, the five-limb core is easier to saturate than the three-limb transformer under the same GIC events; the smaller the side yoke area of the five-limb core, the easier it will be to saturate. More importantly, under GIC events a transformer core could become saturated irrespective of the loading condition of the transformer.

Geomagnetically Induced Currents (GIC) have been a growing concern within power system operators and researchers as they have been widely reported to lead to power system related issues and material damage to system components like power transformers. In power transformers, GIC impacts are evidenced by part-wave saturation, resulting in transformers experiencing increased presence of odd and even harmonics. The three-phase three-limb (3p3L) transformer has been found to be the most tolerant to high dc values compared to other core types. The research was based on a hypothesis which reads “transformer laboratory testing results can be used as a guide towards developing suitable Finite Element Matrix (FEM) models to be used for conducting GIC/DC experiments”. This study thus investigates the response of a 15 kVA 3p3L laboratory transformer to dc current, emulating the effects of GICs. GIC and dc current are the same under steady state conditions, and hence mentioned interchangeably. Laboratory tests conducted identified two critical saturation points when the transformer is exposed to dc. The early saturation point was identified to be at around 1.8 A/phase of dc (18% of rated current), while the deep saturation point was at around 15 to 20 A/phase of dc (about 72% of rated current). Further analysis showed that holes drilled on the transformer can lower the transformer knee-point by about 26%, depending on the size and location of the holes. The holes hence end up affecting the operating point of the transformer due to losses occurring around the holes. A transformer FEM model was developed following the laboratory exercise, where it was concluded that a 2D model leads to grossly erroneous results, distorting the magnetizing current by about 60% compared to the laboratory results. A solid 3D model improved performance by about 30% as it took the transformer's topological structure into consideration. The 3D model was then refined further to include joints and laminations. It was discovered that laminations on the transformer need to be introduced as stacks of the core, with each core step split into two, allocating a 4% air gap space between stacks. Refinement of the T-joints proved that the joints have a relatively high influence on the transformer behaviour, with their detailed refinement improving the transformer behaviour by about 60%. The final FEM model was used for dc experiments. The results of such experiments showed close resemblance to the laboratory results, with saturation points identified in FEM lying within 10% of the laboratory identified saturation points. Overall, the various investigation methods explored showed that the hypothesis was satisfactorily proven true. Laboratory results functioned as a guide in developing the model, offering a reference case.

abstract: This dissertation covers three primary topics and relates them in context. High frequency transformer design, microgrid modeling and control, and converter design as it pertains to the other topics are each investigated, establishing a summary of the state-of-the-art at the intersection of the three as a baseline. The culminating work produced by the confluence of these topics is a novel modular solid-state transformer (SST) design, featuring an array of dual active bridge (DAB) converters, each of which contains an optimized high-frequency transformer, and an array of grid-forming inverters (GFI) suitable for centralized control in a microgrid environment. While no hardware was produced for this design, detailed modeling and simulation has been completed, and results are contextualized by rigorous analysis and comparison with results from published literature. The main contributions to each topic are best presented by topic area. For transformers, contributions include collation and presentation of the best-known methods of minimum loss high-frequency transformer design and analysis, descriptions of the implementation of these methods into a unified design script as well as access to an example of such a script, and the derivation and presentation of novel tools for analysis of multi-winding and multi-frequency transformers. For microgrid modeling and control, contributions include the modeling and simulation validation of the GFI and SST designs via state space modeling in a multi-scale simulation framework, as well as demonstration of stable and effective participation of these models in a centralized control scheme under phase imbalance. For converters, the SST design, analysis, and simulation are the primary contributions, though several novel derivations and analysis tools are also presented for the asymmetric half bridge and DAB. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2019

The rise and adaptation of the transformer architecture from natural language processing to visual tasks have proven a useful and powerful tool. Subsequent architectures such as visual transformers (ViT) and shifting window (SWIN) transformers have proven to be comparable and oftentimes exceed convolutional neural networks (CNNs) in terms of accuracy. However, for mobile vision tasks and limited hardware, the computational complexity of the transformer architecture is an impediment. This project aims to answer the question of whether the Swin Transformer can be adapted towards lightweight and low latency classification as a basis for industrial automation, and how it compares to CNNs for a specific task. A case study from the logging industry, binary classification of wooden boards on chain conveyors, will serve as the basis of this evaluation. For these purposes, a novel dataset has been collected and annotated. The results of this project include an overview of the respective architectures and their performance for different implementations on the classification task. Both architectures exhibited sufficient accuracy, while the CNN models performed best for the specific case study.

The low-voltage dc-dc converter (LDC) in electrical vehicles (EVs) is to convert high dc voltage (270V~430V) from traction battery to low dc voltage (12.5V~15.5V) for the vehicle auxiliary systems. Galvanic isolation is required in the LDC due to safety considerations. Three challenges exist in the LDC design: (1) wide regulation range; (2) high output current; (3) thermal management. The single stage solutions, such as phase-shift full-bridge converter and LLC resonant converter, have been widely studied in the past. A matrix transformer is widely adopted in single-stage LDC design to deal with the large current. At last, the low-profile design allows large footprint area for high power density and ease of cooling design. However, the trade-off between wide regulation range and efficiency exists in single-stage LDC design. Recently, a two-stage solution is proposed to achieve high efficiency and wide regulation range at the same time. The fixed turn ratio LLC stage serves as a dc transformer (DCX) to meet the galvanic isolation requirements and PWM dc-dc stage regulates the output voltages. In this thesis, a variable ratio matrix transformer-based LLC converter is proposed for two-stage LDC efficiency improvement. The transformer secondary copper losses are reduced by taking advantage of the adaptive number of element transformers. In addition, the PWM dc-dc stage achieves better efficiency with variable intermediate bus voltage. The operation principle and design considerations are studied in this thesis. The proposed 1600W two-stage LDC prototype achieves 96.82% full load efficiency under 400V input condition which is 1.2% efficiency higher than the fixed ratio LLC based two-stage design. Last but not least, the prototype shows a comparable efficiency to the fixed ratio LLC based two-stage design even under the low input voltage (270V) condition. / M.S. / The electrical vehicle market is growing rapidly in recent years. However, the driving range is one of the bottlenecks which imperils market growth in the future. Thus, efficient power modules in electric vehicles are desired to extend the driving range. Low voltage dc-dc converter is one of the power modules in electric vehicles which is rated at several kilowatts and converts traction battery voltage for the vehicle auxiliary system, such as air conditioner, headlights, power steering and etc. In this thesis, a variable ratio matrix transformer-based LLC converter is proposed for the two-stage low-voltage dc-dc converter efficiency improvement. Consequently, the driving range of electric vehicles is further extended.

This work presents the design and analysis of two Litz wire transformers for a 500 kHz, 18 kW input­parallel output­series partial power processing converter (IPOS PPP). Because the two power paths in the IPOS PPP operate as “dc transformers” (DCX), both transformers are designed with the goal of leakage inductance minimization in order to reduce gain variation around the resonant frequency. The selected winding topology with the lowest leakage inductance results in an impedance mismatch among parallel secondaries used in the majority power path transformer, resulting in poor current sharing. In order to balance the goals of leakage inductance minimization and even current sharing, a new winding technique called “intra­leaving” is presented which reduces current sharing error from 50%, to 5%. A design rule for “intra­leaving” is also established which extends the winding method to different winding configurations and higher numbers of parallel winding. A novel cooling duct designed with computational fluid dynamics is used for transformer thermal management. The cooling duct uses two 30 mm 7.7 CFM fans to cool the transformer winding and achieves a small height of 43 mm and only 6.8 W power consumption. Using the cooling duct, 106 °C peak winding temperature and 76 °C peak core temperature is achieved at 15 kW load, an ∼ 8% reduction compared to using a conventional 120 mm fan 41 CFM fan. The two transformers with the cooling system achieve 635 W/in3 power density, 1U height compliance, and 99.4% peak efficiency. / M.S. / As society moves towards the electric grid of the future, there have been increased calls for the research and development of resonant power converters due to their high efficiency, high power density, and low electromagnetic interference. The high frequency transformer is one of the main components of the resonant converter system as it contributes substantially to the converters volume, power loss, and thermal management risks. This work seeks to address the trade­offs between leakage inductance minimization and transformer current sharing and proposes a winding method called “intra­leaving” which achieves both. Using “intra­leaving” current sharing error was reduced from 50%, to 5%. Operating transformers at high frequency reduces their volume in accordance with Faraday’s law but also increases thermal risks due to decreased core surface area, higher winding fill factor, and higher loss per unit volume. A novel cooling duct designed using computational fluid dynamics is presented using two 30 mm 7.7 CFM fans and achieves a small height of 43 mm and only 6.8 W power consumption. Using the cooling duct, 106 °C peak winding temperature and 76 °C peak core temperature is achieved at 15 kW load, an ∼ 8% reduction compared to using a conventional 120 mm fan 41 CFM fan. The transformers with the cooling system designed in this work achieve 635 W/in³ power density, 1U height compliance, and 99.4% peak efficiency.