Electrical and Electronic Engineering - Theses
Permanent URI for this collection
Search Results
1 - 10 of 22
-
ItemOperating converter interfaced microgrids( 2016)Microgrids have potential applications for both developing as well as developed countries. In developed countries, microgrid deployment can help in lowering carbon-dioxide emissions and increasing resiliency of the electricity network. In developing countries, microgrids can be used to improve rural electrification levels. The microgrid trend is going to disrupt the traditional energy generation paradigm of monolithic generation. This thesis addresses some challenges that arise with the evolution of microgrids. Microgrids are small sections of the electrical network that comprises of local generation and load. Microgrids, by definition, must be able to operate in isolation from the main grid. In isolated operation, the coordination between sources is very important to achieve stability and ensure operational longevity. Two control hierarchies are popular in microgrid design: 1) master-slave control, and 2) master-less control. These controls are implemented on the sources' power electronic converters. In master-slave control based microgrids, sizing and operating storage systems in combination with local generation is deemed beneficial for reasons like, facilitating pre-meter consumption and leveraging the time-of-use price arbitrage. In this thesis, a model of radial master slave microgrid network with distributed generation and storage is developed. The model developed can be readily adopted to DC master-slave microgrids as well as AC master-slave microgrids with high power factor. Using this model, a central optimization algorithm is proposed. This algorithm allows users to size, position and operate batteries in an intelligent manner where all the network constraints are satisfied and the network costs (capital + operational) are minimized. On solving the problem as a mixed integer linear program (MILP) we obtain appropriate battery sizing decisions at each house in the network and their intended temporal charging and discharging profiles. We extend our results using Monte-Carlo simulation based analysis to address forecasting errors in generation and demand. On the other hand, the design and operation of master-less converter-based microgrids depend on various factors. Little explored is the effect of component mismatches and parameters drifts on the stability and power sharing properties of these systems. On this subject, this works also aims to comprehend the stability and steady state behaviour of multi-master converter based microgrid systems under the presence of non-identical components and parameter drifts. It is shown that microgrid wellness is very sensitive to such changes. We propose a set of coordination controls based on sparse (inter-node) communications to improve the stability margin and ensuring desired power sharing properties. Stability conditions are developed using Lyapunov's indirect method. The performance of the proposed microgrid design is verified using simulation results. For master-slave microgrids, the proposed optimization algorithm improves the economics of microgrid storage while ensuring better power quality. For master-less microgrids, the robustness of power sharing is made independent of parameter drifts, load and sources changes using the proposed methods. Scalability and source modularity are also well-preserved for the master-less scenario using the distributed control laws introduced.
-
ItemCoherent optical short-reach communications( 2016)Optical communication using high-speed on-off-keying by direct modulation (DM) and direct detection (DD) was one of the most inspiring breakthroughs for telecommunication in 1970s. The wide deployment of 2.5-Gb/s per wavelength submarine fiber links in 1990s helped drive the emergence of the Internet as a global phenomenon. However, the evolution of optical coherent detection during the last decade brought a thorough transformation for long-haul transmission, which completely substitutes the role of DM-DD and contributes a 10-time scaling of the fiber channel capacity. Now, DM-DD still holds its position in optical short-reach applications, due to its natural advantage – the simplicity. With the ever-increasing Internet traffic demand, short-reach links require a capacity upgrade in line with the long-haul progression. There is no doubt that coherent detection will gradually penetrate to shorter distance to offer higher data-rate and better system performance than the conventional DM-DD. This thesis investigates how the coherent detection can be transformed to short reach communications cost-efficiently. There are two fundamental transmission impairments for DM-DD systems: (i) chromatic dispersion (CD); (ii) DM laser frequency chirp. The CD-induced frequency selective fading generates nulls within the signal spectrum after DD, which brings severe inter-symbol interference (ISI). This thesis proposes a set of self-coherent detection subsystems to linearize the DD channel, which effectively overcomes the dispersion by digital compensation. DM frequency chirp used to be regarded as the performance barrier for DM-DD, because it expands the optical spectrum, making the signal much more sensitive to the dispersion impairment. For the first time, we show that the detrimental chirp can be converted to an advantage by coherent detection, which significantly improves the system OSNR sensitivity compared with the conventional DM. Moreover, coherent detection enables powerful digital signal processing (DSP) to optimize system performance. This thesis involves a variety of DSP schemes for short reach applications, such as the OFDM modulation for high electrical spectral efficiency, maximum likelihood sequence estimation for optical channel with memory, adaptive equalization for ISI mitigation, digital polarization demultiplexing for multi-dimension direct detection, and advanced forward error corrections. The DSP offers economical solutions to overcome these impairments (or limitations) in short reach systems. The thesis bridges the gap between DM-DD and standard coherent systems, aiming to accelerate the coherent detection application to short reach optical communications.
-
ItemBig data clustering for smart city applications( 2016)The Internet of Things (IoT) infrastructure for the creation of smart cities consists of internet connected sensors, devices and citizens. This IoT infrastructure generates an enormous amount of data in the form of city-scale physical measurements and public opinions, constituting big data. Smart cities aim to efficiently use this wealth of data to manage and solve the problems faced by modern cities for better decision making. However, interpretation of the massive amount of smart city generated big data to create actionable knowledge is a challenging task. Aggregation and Summarization (data clustering) is a useful tool to create knowledge from raw data from different sources. However, traditional data clustering algorithms are not suitable for unlabelled smart city data owing to its high volume and generation velocity and limited experience about generating phenomenon. This thesis presents a novel framework for clustering tendency assessment for big data: clusiVAT, which provides an aggregated view of the big data to create actionable knowledge. clusiVAT intelligently selects a small number of samples from the data such that the samples retain the approximate geometry of the big dataset. The reordered dissimilarity image of the samples generated using single linkage minimum spanning tree (MST) suggests the number of clusters in the data, which is required as an input for most popular clustering algorithms. The cluster labels are then extended to the non-sampled points using the nearest prototype rule. The clusiVAT framework was applied to two real life smart city applications to understand the underlying patterns hidden in the huge volumes of data to generate knowledge. The first application used clusiVAT for clustering and anomaly detection from the pedestrian and vehicle trajectories obtained from a video surveillance system. Experiments were performed on a real-life MIT trajectories dataset of vehicles and pedestrians from a parking lot scene. The trajectory clusters and anomalies thus obtained were helpful in the high-level interpretation of a scene (crowd behavior modeling), as feedback for a low-level (individual) tracking and activity prediction system and as an alarm for human supervisor. For the second application, clusiVAT was used to cluster large scale (of the order of millions) vehicular trajectories obtained from the GPS traces of taxis in the city of Beijing and Singapore using a novel Dijkstra-based dynamic time warping distance measure. The results facilitated the understanding of spatial and temporal patterns in trajectories and were of great significance for decision-makers to understand road traffic conditions and to propose metro bus corridors and light rail systems for better public transport. Another prominent data generated by smart city IoT infrastructure are high-velocity data streams. Automatic interpretation of these evolving big data is required for timely detection of unusual events. This thesis presents a computationally efficient 'hot' update approach for incremental visualization of evolving cluster structure in streaming data. The new algorithms were demonstrated for two applications: online anomaly detection and sliding window based clustering of time series data. Numerical experiments on weather monitoring data from great barrier reef and the city of Melbourne provided visual clues to the onset of the new structure in streaming data.
-
ItemNovel applications of static micro-scale interdigitated electrodes for energy harvesting and biosensing( 2016)Microsystems or Micro-electromechanical Systems (MEMS) refer to the integration of sensors, actuators and signal processing electronics on the single unit. Today microsystems technology has grown rapidly due to its advantages compared to the macroscale counterparts. In 2015, the total market for commercial microsystems products has reached US\$10 billion and has been forecasted to keep growing. This thesis investigates microsystems for energy harvesting and biological sensing applications. Interdigital electrode (IDE) capacitor is one of the most widely used structures for both sensing and actuation. In this thesis, it is applied to both biosensing and energy harvesting applications. To optimize the design and fabrication of these devices, the thesis investigates various analytical models of the IDE capacitor structure. Understanding the effect of different parameters of this structure is critical in the design and optimization of the system. The designs are then verified with a commercial microsystems design and simulation tool (CoventorWare). Simulations were carried out for both microscale IDE capacitor structures, which are used for energy harvesting, and nanoscale IDE capacitor structures, which are used for biological sensing. The thesis contributes advances to microsystems based energy harvesting and biological sensing. In the case of microsystems applied to energy harvesting, a new type of device is proposed. An electrostatic power generator converts mechanical energy to electrical energy by utilizing the principle of variable capacitance. This change in capacitance is usually achieved by varying the gap or the overlap between two parallel metallic plates. This thesis proposes a novel electrostatic micro power generator where the change in capacitance is achieved by the movement of an aqueous solution of NaCl. A significant change in capacitance is achieved due to the higher than air dielectric constant of water and the Helmholtz double layer capacitor formed by ion separation at the electrode interfaces. The proposed device has significant advantages over traditional electrostatic devices which include low bias voltage and low mechanical frequency of operation. This is critical if the proposed device is applied to harvest energy from the environment. A figure of merit exceeding 10000 $\frac{10^8\mu W}{(mm^2HzV^2)}$, which is two orders of magnitude greater than previous reported devices, is demonstrated for a prototype operating at a bias voltage of only 1.2 V and a droplet frequency of 6 Hz. The second contribution of this thesis is in the area of new microsystems based biological sensing. Specifically, this work demonstrates the detection of antigen/antibody. A nanoscale sensing device is implemented to detect glial fibrillary acidic protein (GFAP) antibody for early detection and monitoring of brain tumour. Glioma is the most common primary brain tumour with its early detection remaining a challenge. Autoantibodies against GFAP have shown the highest differential expression compared with the other glioma expressed autoantibodies. Here a prototype of immunosensor to detect GFAP antibody levels is developed using an interdigital coplanar waveguide (ID-CPW). The ID-CPW is fabricated on a SiO\textsubscript{2}/Si substrate with the CPW and interdigital electrode conducting layers made using Cr/Au. The sensor is functionalized, and protein extracted from astrocytes is immobilized on the surface. Sensitivity and dynamic range are ascertained using varying the concentrations of a commercial, polyclonal antibody to GFAP. The electrical detection of antigen-antibody binding is performed in both dry and wet environments across the 1–25 GHz range. The results show that the proposed sensor can detect antibodies against GFAP to a minimum concentration of 2.9 picograms per milliliter with a turnaround time of less than 3 hours. The electrical measurements indicate an improved sensitivity compared with the state-of-the-art optical detection methods. The proposed sensor, developed to detect antibody against GFAP, is the first to show the applicability in the detection of glioma using the GFAP antibodies.
-
ItemReturns on investment: considerations on publicly funded ICT research and impact assessment( 2016)In recent years there have been increasing calls for research to justify public funding by demonstrating how it benefits wider society. This has led to an accompanying search for meaningful metrics and indicators of impact for use in assessment, decision making and policy and program formulation. This thesis examines issues to be considered when attempting to determine the impact of publicly funded research, with particular reference to information and communications technology (ICT). Drawing on many disciplines and often having its biggest impacts outside the corresponding industrial sector, ICT provides a useful lens through which to examine issues surrounding impact and its assessment. Before exploring the idea of impact itself, the question of why governments should support research, particularly ICT research, is looked at, from wide-ranging economic benefits to national security. The main types of research impact are identified, along with common difficulties that arise when attempting to assess them. For research to have impact it predominantly needs the intervention of other parties, such that the question arises as to just how much control researchers actually have over the ability of their research to have impact. In which case, should we be more interested in assessing the potential for impact? If so, what are appropriate indicators and metrics to use? Where does peer review, the traditional method of assessing research, fit in this framework? For government funded research to have an impact, it must first be clear about what it is trying to achieve before determining what indicators are best suited for the demonstration of success. The thesis concludes by posing some questions that should be considered when funding programs are being constructed. Included in the thesis are three case studies, examining models for having impact via doctoral training, trans-disciplinary research and sectorial engagement within the Australian higher education sector.
-
ItemThe systems biology of eukaryotic transcription( 2016)Understanding the regulatory systems that govern eukaryotic transcription is a central challenge in the life sciences. All cellular phenotypes and behaviours emerge from complex programs of gene transcription, and even subtle disruption of their epigenetic regulatory layer can drive debilitating human disease. Many developmental, autoimmune, neurological, inflammatory and neoplastic disorders have been associated with aberrant epigenetic behaviour, in addition to breast, prostate and other cancers. Prognosis and treatment success for these diseases can be monitored through sets of epigenetic biomarkers, and several chemotherapies that intervene at the epigenetic level (epidrugs) have been approved for human treatment. Despite the early promise of epidrugs, understanding of how cellular phenotypes and behaviours emerge from the complex systems of epigenetic interactions (namely those of transcription factors, histone modifications and DNA methylation) remains fragmentary and incomplete. Thanks to recent consortia including ENCODE, FANTOM and NIH Roadmap Epigenomics, there is now a wealth of publicly-available data that match transcriptomic and epigenetic sequencing across hundreds of organisms, cell-types and diseases. The bottleneck that limits further understanding of transcriptional regulation has now shifted to the effective integration of these large and multi-factorial data-sets. This thesis evaluates, develops and improves upon several computational frameworks for transcriptomic and epigenetic data integration. Recent advancements in information theory are employed to improve unsupervised inference of both symmetric and directed gene regulatory networks. Despite these improvements, pairwise gene-level interactions prove insufficient for capturing complex regulatory activity at the epigenetic level. Focus is instead placed on an emerging class of predictive gene expression models, wherein genes are treated as independent observations of epigenetic activity to explore specific regulatory processes. Several limitations of previous modelling methods are identified and addressed, including previously-unexplained redundancy within epigenetic data-sets. This limitation is leveraged as a source of valuable insight regarding conditional and synergistic regulatory activity at methylated and bivalent gene promoters. These findings are unified into a conceptual framework that explains the general machinery of transcriptional regulation and allows for phenotypic data to be captured effectively. The power of this systems-level approach is demonstrated by predicting novel, epigenetically-regulated drivers of invasive breast cancer from previously-unpublished data.
-
ItemPatient-specific neural mass modelling of focal seizures( 2016)Patient-specific computational modelling of epileptic seizures may make the models more useful for diagnosis, management and treatment of epilepsy. In this thesis, patient-specific models from electroencephalogram data from patients are presented. The models can potentially be used as a starting point in the design of a seizure control system based on electrical stimulation. In this thesis, generalised versions of the Jansen and Rit neural mass model to emulate underlying generators of seizures are presented. This model is generalised to slow-fast neural mass models to replicate deterministic biological mechanisms of seizure initiation and termination. To resemble stochastic mechanisms of seizure initiation, a Duffing neural mass model is developed by introducing perturbations to the linear time invariant dynamics model of the synapse of a typical neural population in the Jansen and Rit model. Parameter identification of a slow-fast neural mass model is explored using a genetic algorithm and unscented Kalman filter. Using the genetic algorithm, we show that parameter estimation can lead to quantitative reproduction of the pattern of seizure initiation in EEG data. The unscented Kalman filter is employed to enhance the parameter estimation of the slow-fast neural mass model using genetic algorithm-based identification. Identification of the Duffing neural mass model of seizures is performed by simultaneous usage of the unscented Kalman filter and genetic algorithm to optimise the likelihood function with respect to the parameters to be estimated. The results shows that these methods are useful for parameter identification of neural mass models of focal seizures. Patient-specific modelling of an animal model of focal seizures is explored by fitting the slow-fast and Duffing neural mass models to animal EEG data. The unscented Kalman filter is employed to infer hidden dynamics of neural populations for EEG data. For a given EEG data, a model likelihood function is employed for model selection.
-
ItemDistributed demand side management and tariff design in distribution networks( 2016)As a non-storable commodity with time-varying demand, the `logistics' of electricity has always been challenging. To ensure reliability, electricity systems are required to be built with redundant generation, transmission and distribution capacity such that they are able to withstand the maximum demand under expected fault conditions. On the one hand, structuring the electricity system in such a way is not efficient. A significant portion of network capacity is under utilised or sitting idle for most of the time and the situation is getting worse over time as shown by increasing peak to average ratios. On the other hand, emerging technologies such as Electric Vehicles (EV), photovoltaic (PV) and storage devices will fundamentally change the existing paradigm of the existing electricity network. The potential peak surge caused by large numbers of EVs and possible over-supply from PVs may harm the networks but the large flexible battery capacity of EVs, inexpensive and clean energy from PVs together with storage devices could, on the other hand, greatly benefit the networks. This is especially the case with smart measuring devices being widely installed and affordable sensors being ubiquitous. Therefore, it is hugely important to implement the technologies appropriately and prepare the electricity networks for upcoming changes. Demand Side Management (DSM) and tariff design are proposed, from two complementary directions, as promising concepts to increase the efficient use of electricity grid assets, to enable a smooth transition towards smarter grids and to promote fairness among network users. Among the rich literature on DSM algorithms, advanced electricity grid information and communication technology (ICT) is usually assumed to exist. In this thesis, however, the requirement of communication is relaxed and solutions for networks with different ICT infrastructures are developed and compared. It is shown that satisfactory DSM can be achieved even without communication via the stochastic voltage-demand model we develop. Network tariff design focusing not only on revenue recovery and social welfare maximisation, but also fairness among electricity users especially in the presence of PVs has also been studied. A utility based model to quantify the benefit and cost for an individual user using the network is proposed. The model is adopted to assess efficiency and fairness of existing tariff structures as well as designing optimal tariffs.
-
ItemWired and wireless services over next generation WDM/TDM PON systems( 2016)Passive optical networks (PON) have been the most popular networking approach for fibre-to-the-home network deployments. The incorporation of wavelength division multiplexing (WDM) along with time division multiplexing (TDM), delivered over a PON with longer feeder network, has been studied as a promising solution. This is widely known as the Long Reach WDM/TDM PON and is considered as a suitable option for a cost effective access network offering higher bandwidth, potentially to a larger customer base compared to PON. Future access networks like this will be expected to support the increasing level of fixed mobile convergence in wireless communications by offering cost-effective backhaul for such access networks. Future access networks will also need to offer better support for peer-to-peer (P2P) networking between customers. Increasingly, these network alternatives need to offer improved energy efficiency to help lower the carbon footprint of future access networks. As future networks will support the increasing mix of fixed and mobile wireless access, network needs to consider better mechanisms for the management of medium access control (MAC) layer functions to deal with the complexity of delivering wired and wireless services over a WDM/TDM PON architecture. This thesis explores long-reach WDM/TDM PON networks with a view to answering the following key questions: (a) how to support wireless networks through MAC layer design, (b) how to provide P2P networking interfaces, (c) how to support energy efficient operation, and (d) how to offer low latency in such networks. To understand the performance of optically networked wireless base stations, the MAC layer performance of wireless networks was investigated using radio-over-fibre (RoF) and picocell based architectures. Through mathematic analysis and simulations, it is shown that picocell solutions offer better performance compared to RoF solutions. The performance effect of contention-based bandwidth request mechanisms in mobile networks (using WiMAX as an example) is studied. A rigorous analysis of MAC layer performance is formulated using an accurate Markov chain mode, including new parameters for maximum bandwidth request retries and three waiting states during this procedure. Traditional P2P overlay networks, which are built on top of the network hierarchy, lead to long transmission distance for P2P traffic, crossing core networks and several Internet Services Providers systems, even though the peer node is located very close or even in the same passive optical network. In this case, to improve the performance of P2P traffic, a new P2P network architecture employing channel combine/splitting module is proposed to implement a low latency, flexible and scalable P2P overlay network over long-reach WDM/TDM PON systems. P2P traffic is localized at a remote node, to avoid long roundtrip delays due to long-reach system. Through the experiment and simulation, it is shown the P2P overlay network has negligible physical performance degradation and P2P data transmission delay is reduced. Furthermore, the overlay network shows good flexibility in constructing optical virtual private networking. To enable energy-saving design of the optical line terminal (OLT) by activating the ‘sleeping mode’, a tunable grating-based monitoring technique in RSOA-based WDM PON systems is proposed to wake up the sleeping OLT when the subscriber is active and connected to the network. The experimental evaluation shows good feasibility, and the tunable grating-based technique shows a 10dB distance gain compared to the amplified spontaneous emission (ASE)-based technique. Moreover, to support cost-effective and energy efficient network deployment and operation, a remote channel combine/split (CCS) module for long-reach WDM/TDM PON systems was proposed to achieve a higher energy saving by adopting subscriber take-up rate and traffic adaptive power management. Through experimental evaluation and simulation, it is shown the proposed CCS module only causes 0.15dB negligible physical link performance degradation. Also, simulation shows up to 60% investment saving at the initial network deployment stage, and over 20% energy saving compared to traditional full operating systems. To achieve less registration latency during network transition, a contention-free low-latency handover scheme is proposed to implement the smooth transition of optical network units between different PON groups. Through the simulation, it is shown the proposed handover scheme produces only 3.2-ms handover delay, which is far below the stringent delay requirement for voice services and shows a negligible degradation on QoS performance and ONU’s buffer size. Moreover, to support an energy-efficient networks operation, a globalized OLT cluster structure with energy-aware dynamic wavelength and bandwidth allocation (EA-DWBA) algorithm was proposed to save the power consumption of the OLT cluster itself and the edge router in the central office. The simulation shows up to 40% power consumption saving of the OLT at the lower traffic load, and up to 30% power consumption saving on the edge router based on realistic proportional edge router technology.
-
ItemSystem identification and control of rivers( 2016)Water resource management has immense importance in the modern world. A large amount of water is wasted due to inefficient management of rivers, lakes and other water bodies associated with them. In order to achieve improved management of water resources, control and decision support systems can be employed. To design control systems for a river, a river model is required. Traditionally, the Saint Venant equations have been used for modelling purposes. The equations describe river flows accurately, however, they are complex, non-linear and require many unknown parameters. It is therefore difficult to use them for control design purposes. On the other hand, data-based models have proven to be very useful in control design for rivers. In this thesis, different data-based modelling methods are explored, and they are applied to the data from the upper part of Murray River in Australia. For each method, the thesis analyses the ease with which available prior knowledge can be incorporated in the modelling procedure and the ability of the obtained models to describe the river well. For efficient river control, forecasts of future water demands and flows in the unregulated tributaries are required to be taken into account. A Stochastic Model Predictive Control (S-MPC) or a randomised version of it can not only accommodate such forecasts, but it can also handle physical and environmental constraints well. However, due to uncertainties in the forecasts, the feasibility of optimisation problems cannot always be guaranteed in the presence of constraints. This thesis proposes an S-MPC based river control schemes, that not only incorporate the forecasts, but also ensure feasibility of the optimisation problems. The schemes are successfully applied in simulations to the past data from the upper part of Murray River. Another important aspect of river management is to mitigate flood risks. An ideal strategy is to reduce the risk of severe floods, and at the same time not being overly cautious while performing normal river operations. This thesis uses Value-at-Risk (VaR) as a risk measure and incorporates it into the river control problem, forming a Multiple Chance-Constrained optimisation Problem (M-CCP), to be solved in an S-MPC setup. A computationally tractable Optimisation and Testing algorithm is developed to find solutions to M-CCPs, with probabilistic guarantees on the solution. The algorithm is successfully applied to the historical data from the upper part of Murray River. The simulation results show better regulation and flood avoidance.