A comprehensive study of the motivation and methods of implementing electrical clusters of up to five wind turbines is presented. The merits and pitfalls of three potential technologies for implementing such clusters are investigated through the extensive simulation of each, from which it is concluded that passively rectifying the outputs of PM generators and connecting them in parallel, using a multi-terminal DC collection network, is the preferred option; allowing a degree of independence to occur between the turbines, whilst maintaining a high energy transfer efficiency. The process of generator output rectification is investigated thoroughly, demonstrating that the distortion of the generator terminal voltages allows the generator speeds to slip by up to 33% from the cluster synchronous speed (derived from the cluster common point voltage). The maximisation of the rectifier commutation overlap length to increase the generator slip capability is demonstrated to reduce the generator power factor, but also reduce the winding current and torque total harmonic distortion at maximum power output. In addition it is found that the use of the commutation overlap voltage distortion to provide generator slip, allows a significant energy transfer efficiency benefit to be achieved over the use of additional resistance for the same purpose. The resulting generator slip capability is proven to benefit the energy capture efficiency of the clustered turbines and also provide sufficient damping of the wind turbine drive-train natural oscillations. The operation of the system and the capability of the wind turbine rotational speeds to slip from the cluster synchronous speed are verified experimentally using a bench scale test-rig, incorporating two small wind turbine generators.

Hybrid DC circuit breaker switching techniques have been proposed for use in multi-terminal HVDC networks, to create an efficient, reliable and controllable system. There are two DC hybrid switching techniques. The first is arc interruption based on current oscillation. The second is interruption without an arc, like with a voltage commutation approach. The main difference is whether or not the interruption mechanism deals with the arc. Interrupting the arc not only causes erosion of the electrode surface, therefore reducing useful lifetime, but also introduces uncertainty to the process of successful interruption. This thesis therefore focuses on determination of the best switching technique for HVDC applications. To highlight the advantages of interruption without an arc, this thesis uses a vacuum circuit breaker (VCB) as the main breaker to investigate arc effects on interruption performance, when employing a forced current commutation method. The vacuum arc characteristics are detailed, thereby giving the VCB operation understanding needed. The VCB opening time is reduced from approximately 50ms to 13ms by employing a coil DC activation method. This allows the coil counter-EMF to be observed and utilised to optimally trigger the auxiliary commutation circuit. An active commutation test circuit is proposed, and in simulation and experimentally, the VCB interruption properties are investigated in terms of varied interruption current, di/dt, dvVCB/dt, and gap distance. Experimental results elicit that the key parameters determining arc interruption probability is di/dt, where even if dvVCB/dt is low, successful interruption become impossible if di/dt is above a certain level. By analysing the post-arc current based on its electrical behaviour, the reason why di/dt and dvVCB/dt dominate successful interruption probability is explained. A cascaded commutation circuit is proposed, and its validity is confirmed by simulation and experimentally, showing the interruption probability is improved compared to the test circuit, even with small circuit inductance.

Partial discharge (PD) is recognized as an effective insulation diagnostic tool. PD measurement on high voltage (HV) cables is performed to evaluate the cable insulation conditions and high frequency current transformers (HFCTs) are widely used as the PD sensor. Detected PD signals can be subject to significant attenuation caused by propagation in the cables. How to relate the detected signals to the PDs actually occurring at PD-producing defects and moreover find out the nature of the defects is a problem. To investigate this problem experimentally is both difficult and limited since some defect types cannot be easily reproduced. Exploring this problem through modelling can potentially provide useful insights. A hybrid modelling framework is proposed to simulate HFCT-based PD detection in HV cables regarding PD excitation and propagation in the cables and PD detection using HFCTs. The finite difference time-domain (FDTD) technique is used to simulate PD in the cables and transfer function theory is used to consider effects of HFCTs on the PD signals. The framework can be used to excite a PD pulse anywhere in a cable model and predict the output from HFCTs some distance away. Implementing the framework requires parameters such as cable material/propagation properties, currents owing on conductors in FDTD models and HFCT transfer functions. Based on available cable samples and HFCTs, methods to obtain these parameters were described or developed and implementation of the framework was demonstrated. Although only cable modelling was carried out, the framework is expected applicable to modelling of PD detection in cable accessories. Furthermore, some experience gained in implementing FDTD modelling of the cables was summarised to facilitate future applications.

The electricity network in the UK is facing a challenging future where environmental and efficiency goals to be met require significant changes in the industry. The rethinking of how distribution networks are arranged and operated is essential due to increasing connections of generation at this level. This also gives significance to the issue of arranging and operating these assets in the most optimal manner. The saturated state of the transmission network has led to the popularity of connecting generation to distribution networks. It becomes evident that, when increasing generation from renewable sources, a means of storing energy is required. One area of storage being investigated for this aim is Electrical Thermal Storage (ETS), which is a form of flexible heating demand. This type of energy storage is cost effective, however it is limited in its application due to the natural dissipation of heat. It is preferable that this heat loss be of use and therefore ETS devices are used in the domestic setting. These devices can be in the form of space storage heaters or hot water storage tanks, which are able to store heat or hot water for up to several days, until the user wishes to make use of the resource. Although decentralised generation and energy storage (Distributed Energy Resources - DER) at first glance appear to be an attractive means for the energy and carbon targets to be met, it can introduce further problems to the electricity network. It is possible for the addition of DER to lead to voltage rise or voltage drop, negatively impact on the protection systems and affect power quality. In order for distribution networks to successfully operate it is important that any new network assets are connected in an optimal manner. This research has modified and used a multi-objective network planning framework to determine the optimal arrangement of ETS devices in distribution networks. The framework is built around the Strength Pareto Evolutionary Algorithm 2 (SPE A2), which takes its inspiration from evolution in nature, and is able to take into consideration the constraints of the network and multiple, conflicting objectives. The framework has been adapted to optimise the number of ETS devices, their location and operation, whilst ensuring that the network will operate within its constraints. The results generated by the planning framework illustrate the potential benefits offered by the inclusion of ETS in distribution networks, as well as demonstrate that the method and tools used are valuable and appropriate.

To ensure that infrastructure owners, operators and planners have sufficient information readily available to them regarding the state and condition of their asset, numerous advances and developments have been demonstrated in the field of Non Destructive Evaluation (NDE). The process of detailed imaging and examination of structures and components in a sensitive, safe and inherently non-intrusive manner has numerous advantages in operational, financial and safety terms. Only by delivering NDE sensors directly to all points of interest on a structure, can the desired coverage of the structure or component be undertaken. This vision has driven research and developments in the area of automated collection of NDE data. This thesis identifies and addresses key challenges and technical boundaries currently preventing wider industrial uptake of automated NDE. This has been accomplished through novel research and developments in key areas related to platform kinematics and dynamics, localisation and registration to the structure under inspection, path planning and physical NDE sensing. These advances provide a unified framework to both establish and progress remote automated NDE forward for future commercial deployment. A contribution to knowledge was presented on the positional performance characteristics of a custom rotary wing aerial inspection platform after a series of hover tests within a tracking volume. An additional contribution was presented on the results of a comprehensive evaluation of a laser based positioning sensor operating under a wide range of engineering materials, along with the development of a novel distance correction algorithm. A new custom approach for robotic NDE path planning was researched and developed based on conventional CAD/CAM machining, while utilising a novel custom agile ultrasonic thickness mapping crawler to highlight proof of principle. Both of these approaches provide an additional contribution to knowledge, both individually in their respective fields and also together in highlighting the unique advantages of their combined approach. Finally a new NDE sensor concept based on the vibrissae of rodents such as rats and mice, for both surface roughness and surface profiling measurement, was adapted and investigated for NDE sensing applications. The results document a contribution to knowledge in the area with discrimination down to 14 micrometre Roughness Average presented.

Complex multi-domain engineering systems, where for example mechanical and thermal (sub)systems are connected to each other in some way, have increasingly become a vital part of our society. An example of such a system is the Integrated Full Electric Propulsion (IFEP) concept for the marine shipping industry. With this IFEP concept, as opposed to the more conventional marine power system, the power for the ship's propulsion and ship's services is provided by a common power plant. This offers advantages including fuel efficiency and design flexibility. However, due to its system complexity and capital costs, it is important that the overall dynamic behaviour of these systems can be predicted in the early stages of the design. Predicting the overall system behaviour can be obtained by employing an integrated end-to-end model, which combines detailed models of for example the mechanical and electrical (sub)systems. This allows for example ship designers to investigate disturbances and the primary and higher order responses across the system. However, present existing simulation tools do not easily facilitate such employment of a holistic approach. In this thesis the focus is on how advanced modelling and simulation techniques can be used to de-risk the design and in-service of complex IFEP systems. The state-of-the-art modelling and simulation techniques as well as the IFEP application area are considered. An integrated-model of an IFEP vessel was developed under the EPSRC collaborative AMEPS (Advanced Marine Electric Propulsion System) research project, which forms a major part of this thesis. In order to reduce the computational burden, due to a wide variety of time constants in the IFEP system, a multi-rate simulation technique was proposed. It was demonstrated that a reduction in simulation execution time between 10-15 times can be achieved. However it was conceptually argued that multi-rate simulation could introduce errors, which propagates itself across the system thereby provoking potential unrealistic responses from other subsystems. Several case studies were conducted based on this model, which shows that such an integrated end-to-end model may be a valuable decision-support tool for de-risking the design and in-service phases of IFEP vessels. For example, it was demonstrated that a disturbance on the propeller could provoke a saturation of the gas turbine governor. Different power system architectures were proposed for IFEP power systems such as radial and hybrid AC/DC. For this thesis, an initial study was conducted to assess the relationship between the type of power system architecture and the vessel survivability. For this assessment an existing vessel survivability theory was further developed into a quantitative method. It was concluded that based on a comparative short circuit study and the proposed survivability method that the IFEP-hybrid AC/DC architecture offers the best vessel survivability.

Although recent advances in wireless system technologies have provided ever increasing throughputs, end user demand continues to increase unabated. The research investigates the performance of a system harnessing two relatively new but powerful technologies, Multiple-Input and Multiple-Output (MIMO) and Ultra Wideband (UWB) transmission as a possible solution to meet the growing demand for capacity. Each of these technologies in its own right has been subject to a large volume of research and has been proven to bring an increase in throughput. Nevertheless the predicted future demand will outstrip what each strategy can provide individually. MIMO-UWB systems are thus an emerging wireless solution with, in particular, the potential to satisfy short distance, high speed transmission requirements within indoor environments. Before any system is deployed it is important to characterise performance within representative operating environments. The study therefore emulates appropriate indoor environments, defines an experimental protocol to execute a range of measurements that provide robust evidence of the behaviour of the combined system within indoor scenarios. The application scenario dictates that the transmitter represents a gateway device attached to the ceiling and the receiver, a user device set on a table. The sequence of measurements relate to different positioning of the user device, with different angles and ranges to the gateway device, the layout of antenna placements being important. The output of the study is an accurate model for engineers and, the foundation for the design of MIMO-UWB systems for indoor services.

The cognitive interference channel extends the classical two-user interference channel to have unidirectional cooperation at the transmitters. In this model, the cognitive transmitter is assumed to have noncausal knowledge of the other transmitter's current message (primary message). This a priori knowledge is used by the cognitive user to accomplish its two main purposes, i.e., to relay the primary message in order to boost the primary user's data rate and to maximise its own data rate by cancelling the interference at its receiver. The cognitive interference channel is well studied in the literature and capacity results are available for the weak and very strong interference regimes, amongst others. A general solution is still elusive. In this thesis we study the capacity region of cognitive structures that are based in their core on the cognitive interference channel but with the aggregate that an additional node is considered, e.g., an additional receiver node, an additional transmitter node or a relay node. The cognitive broadcast interference channel consists of the cognitive interference channel with an additional receiver. The cognitive side serves either one or two receivers and the interference goes from the cognitive transmitter to the primary receiver only. In this model we provide a general achievable rate region when the cognitive side serves two receivers. We analyse the discrete memoryless channel and show that the region simplifies to existing results in the literature when certain assumptions are made. An achievable rate region for the Gaussian channel is also provided for the case where the cognitive side sends common information to both receivers. When the cognitive side serves only one receiver, we provide an achievable rate region and an outer bound and show the gap graphically. The cognitive interference channel with a relay consists of the cognitive interference channel with an additional relay node. In this model we show that as in the interference channel with a relay, interference forwarding is also beneficial. We describe analytically achievable rate regions and show the benefits of interference forwarding. We also provide an achievable rate region with generalised interference forwarding, i.e., the relay forwards the intended message and the interference simultaneously, and show that allowing the relay to allocate part of its power to forward interference is beneficial when we are in the strong but not in the very strong interference regime. The cognitive interference channel with causal unidirectional destination cooperation is formed by transferring the relaying capabilities of the relay node in the previous model to the cognitive receiver and its operation is causal rather than strictly causal. In this model we show that instantaneous amplify and forward is good enough to achieve the capacity region of the Gaussian channel. We derive analytically an inner and outer bounds and show that they coincide for certain values of the antenna gain at the relay in the very strong interference regime. We also analyse the cognitive interference channel with a relay for the case where the relay operates causally. The capacity region is obtained for a special case of very strong interference. The cognitive multiple access interference channel consists of the cognitive interference channel with an additional cognitive transmitter. In this model the interference goes from the primary user to the cognitive receiver only. The cognitive users form a MAC channel. We show for this scenario that dirty paper coding achieves the capacity region in the Gaussian case. In the analysis we make use of encoding techniques first utilised for the MAC with state available noncausally at the encoder.

This thesis proposes and discusses aspects of a low-cost wireless network called "Hopscotch" as a potential solution to the rural broadband problem. Providing broadband internet access to rural locations is challenging due to the long distances between internet backbone and households, the sparse population density and difficult terrain. Hopscotch uses a network of renewable powered base stations, termed "WindFi", connected by point-to-point links, to deliver internet access to rural communities. A combination of frequency bands are used within Hopscotch. Standard IEEE 802.11 5 GHz WiFi access technology is used for high capacity links, and an ultra high frequency TV "white space" spectrum overlay in the 600-800 MHz band provides long distance coverage. The advantages of "white space" spectrum are demonstrated for a rural wireless scenario; reducing the number of base stations required to cover a community and decreasing the transmit power required to create long distance links over challenging terrain. The use of renewable power allows WindFi base stations to be well placed to serve a community, irrespective of available infrastructure. The power system is the biggest cost component of the base station therefore the system must be carefully sized. The design of the WindFi base station is presented and the specification of the renewable power generation system validated with operational data. To reduce the energy required, and therefore the demand on the renewable power system, aspects of energy use within the base station are considered. Models of the power consumption and data rate selection for radios used in Hopscotch are presented. Hopscotch trials have been running on the Scottish islands of Bute and Tiree. Measurement based models of household distribution, daily network internet traffic and large-scale path loss for a rural community are presented based on trial results, which are useful for simulating rural broadband networks. To minimise the power consumption of the WindFi base station, an energy optimisation is presented for a Hopscotch scenario. Dynamically altering the assignment of users between two overlay radio access networks, based on the instantaneous capacity required, is shown to reduce power consumption. The optimum assignment between the networks to maximise individual user throughput is also presented.

Loudspeaker-based sound systems, capable of a convincing reproduction of different audio streams to listeners in the same acoustic enclosure, are a convenient alternative to headphones. Such systems aim to generate "sound zones" in which target sound programmes are to be reproduced with minimum interference from any alternative programmes. This can be achieved with appropriate filtering of the source (loudspeaker) signals, so that the target sound's energy is directed to the chosen zone while being attenuated elsewhere. The existing methods are unable to produce the required sound energy ratio (acoustic contrast) between the zones with a small number of sources when strong room reflections are present. Optimization of parameters is therefore required for systems with practical limitations to improve their performance in reflective acoustic environments. One important parameter is positioning of sources with respect to the zones and room boundaries. The first contribution of this thesis is a comparison of the key sound zoning methods implemented on compact and distributed geometrical source arrangements. The study presents previously unpublished detailed evaluation and ranking of such arrangements for systems with a limited number of sources in a reflective acoustic environment similar to a domestic room. Motivated by the requirement to investigate the relationship between source positioning and performance in detail, the central contribution of this thesis is a study on optimizing source arrangements when strong individual room reflections occur. Small sound zone systems are studied analytically and numerically to reveal relationships between the geometry of source arrays and performance in terms of acoustic contrast and array effort (related to system efficiency). Three novel source position optimization techniques are proposed to increase the contrast, and geometrical means of reducing the effort are determined. Contrary to previously published case studies, this work presents a systematic examination of the key problem of first order reflections and proposes general optimization techniques, thus forming an important contribution. The remaining contribution considers evaluation and comparison of the proposed techniques with two alternative approaches to sound zone generation under reflective conditions: acoustic contrast control (ACC) combined with anechoic source optimization and sound power minimization (SPM). The study provides a ranking of the examined approaches which could serve as a guideline for method selection for rooms with strong individual reflections.