No description available.

No description available.

No description available.

No description available.

One of the weakest links in any communications system is the power amplifier, largely due to the necessary trade-off between linearity and efficiency. The challenge of power amplifier design is therefore to provide maximum power and efficiency while maintaining highly linear operation. Gallium Nitride HFET devices are of great interest in power amplifier applications and are potentially the best device technology to provide high power amplification at high frequencies. At the outset of this work, the development of GaN HFETs was in its early stages and the primary aim of this work is to help the development of the QinetiQ GaN program and to investigate the suitability of GaN HFETs for use as power amplifiers. This was achieved through the use of novel measurements using the custom RF non linear time domain measurement systems at Cardiff University. These QinetiQ devices exhibit the common problem of DC-RF dispersion, which results in much degraded RF performance relative to that predicted from DC measurements. It has been shown that the devices exhibit a direct correlation between dispersion and channel current, and that it is possible to overcome some of the unwanted side-effects of dispersion through suitable design, so that efficiency can be improved to acceptable levels, while maintaining high power operation. It has also been shown that there is a strong correlation between RF waveform measurements and pulsed l-V measurements, and that through careful choice of bias point, pulsed l-V measurements can be used to accurately predict the RF performance of a device. Through large-signal CW and two-tone characterisation, it has been determined that these devices operate in a very similar way to that predicted by simple theory and exhibit little AM-PM behaviour. Their CW operation is also found to be less complex than that of GaAs pHEMTs, and their two-tone operation correlates well with the CW analysis. They should therefore be ideally suited to power amplifier applications in terms of linearity as well as power and frequency performance. Finally, after the development of a new two-tone time domain measurement approach, it was determined that the devices suffer from intrinsic electrical memory effects due to their transit delay but are otherwise inherently free from memory within the measurement bandwidth - the range of measurable modulation frequencies is 200KHz-25MHz. All other memory effects detected are not caused by the device but are extrinsic electrical memory effects caused by the impedance environment that is presented to the device.

Investigations of the behaviour of point discharge currents down an earthed point mounted at the height of surrounding trees are described. Simultaneous measurements of the potential gradient at the ground to windward of the point, point discharge current, and windspeed at the point indicate that the current can be represented by the expression I = k (W + C) (F-M) where I is the point discharge current W is the windspeed F is the potential gradient M is the onset value of the potential gradient and k and 0 are constants. This is in direct agreement with the results of Chalmers and Kirkman (1957) for a point on a high mast and considering what has been said by Chalmers (1957) the results suggest that the current flowing down points in all conditions can be represented by the equation I = A (V -Vo) (W(^2) + C(^2)V(^2)1/2 where I and W are as above. V is the potential of the air surrounding the point Vo is the onset potential and A and c are constants. Measurements of point discharge currents down a tree indicate that for a fixed windspeed the current can be represented by I = a (F - M)and that other things being equal the current down the tree is only one tenth of the current down the point. This suggests that estimations of charge brought down by point discharge based on the equivalence of a point and tree would have to be greatly modified. Laboratory experiments were carried out to test the practicability of measuring currents down trees by a transformer method and currents of 5 microamps were measured with ease. The sensitivity could be greatly improved and it is suggested that this is a possible method of measuring point discharge currents down trees.

No description available.

No description available.

No description available.

Many countries are rolling out smart electricity meters. A smart meter measures the aggregate energy consumption of an entire building. However, appliance-by-appliance energy consumption information may be more valuable than aggregate data for a variety of uses including reducing energy demand and improving load forecasting for the electricity grid. Electricity disaggregation algorithms – the focus of this thesis – estimate appliance-by-appliance electricity demand from aggregate electricity demand. This thesis has three main goals: 1) to critically evaluate the benefits of energy disaggregation; 2) to develop tools to enable rigorous disaggregation research; 3) to advance the state of the art in disaggregation algorithms. The first part of this thesis explores whether disaggregated energy feedback helps domestic users to reduce energy consumption; and discusses threats to the NILM. Evidence is collected, summarised and aggregated by means of a critical, systematic review of the literature. Multiple uses for disaggregated data are discussed. Our review finds no robust evidence to support the hypothesis that current forms of disaggregated energy feedback are more effective than aggregate energy feedback at reducing energy consumption in the general population. But the absence of evidence does not necessarily imply the absence of any beneficial effect of disaggregated feedback. The review ends with a discussion of ways in which the effectiveness of disaggregated feedback may be increased and a discussion of opportunities for new research into the effectiveness of disaggregated feedback. We conclude that more social science research into the effects of disaggregated energy feedback is required. This motivates the remainder of the thesis: to enable cost-effective research into the effects of disaggregated feedback, we work towards developing robust NILM algorithms and software. The second part of this thesis describes three tools and one dataset developed to enable disaggregation research. The first of these tools is a novel, low-cost data collection system, which records appliance-by-appliance electricity demand every six seconds and records the whole-home voltage and current at 16 kHz. This system enabled us to collect the UK’s first and only high-frequency (kHz) electricity dataset, the UK Disaggregated Appliance-Level Electricity dataset (UK-DALE). Next, to help the disaggregation community to conduct open, rigorous,repeatable research, we collaborated with other researchers to build the first open-source dissaggregation framework, NILMTK. NILMTK has gained significant traction in the community,both in terms of contributed code and in terms of users. The third tool described in this thesis is a metadata schema for disaggregated energy data. This schema was developed to make it easier for researchers to describe their own datasets and to reduce the effort required to import datasets. The third part of this thesis describes our effort to advance the state of the art in disaggregation algorithms. Three disaggregation approaches based on deep learning are discussed: 1) a form of recurrent neural network called ‘long short-term memory’ (LSTM); 2) denoising autoencoders; and 3) a neural network which regresses the start time, end time and average power demand of each appliance activation. The disaggregation performance was measured using seven metrics and compared to two ‘benchmark’ algorithms from NILMTK: combinatorial optimisation and factorial hidden Markov models. To explore how well the algorithms generalise to unseen houses,the performance of the algorithms was measured in two separate scenarios: one using test data from a house not seen during training and a second scenario using test data from houses which were seen during training. All three neural nets achieve better F1 scores (averaged over all five appliances) than either benchmark algorithm. The neural net algorithms also generalise well to unseen houses.