Electrical and Electronic Engineering - Theses

Permanent URI for this collection

Search Results

Now showing 1 - 10 of 40
  • Item
    Thumbnail Image
    A Modelling Framework for Virtual Power Plants Under Uncertainty
    Naughton, James Ciaran ( 2021)
    The increased integration of renewable energy sources (RES) and distributed energy resources (DER) into electrical networks is causing operational challenges. The reduction in conventional generators, which would traditional provide the reliability and security services for electrical networks, means that these services must now be supplied by other resources. Simultaneously, the intermittency of RES and the lack of visibility of DER means that in some cases these services are required more frequently to maintain a reliable electrical grid. If RES and DER are aggregated and properly controlled in a virtual power plant (VPP) they have the potential to provide network services as well as increase their profitability. The operation of a VPP is a complex problem. While this problem has been examined by numerous authors, no operating framework has been previously proposed that includes consideration of: participation in multiple markets; provision of network and contractual services; modelling of network power flows and voltages; interactions between multiple energy vectors; uncertainty in operational forecasts and; tractability for short dispatch periods. These are key properties for a comprehensive framework that fully captures and unlocks the potential of a VPP. This thesis presents the design and application of a VPP operational framework that incorporates these six key properties. This optimisation-based framework is decomposed into three optimisations to integrate these properties in a tractable manner. This framework is applied to various realistic case studies to prove the efficacy of the proposed approach. The application of this framework demonstrates that the combination of scenario-based optimisation and receding horizon control used is effective at mitigating the effects of uncertainty. The inclusion of short dispatch periods is shown to be key for revenue generation in markets with short dispatch windows. In addition, the application of this framework demonstrates the ability of a VPP to participate in multiple markets and services, and that doing so is essential for maximising VPP revenue. Moreover, the integration of hydrogen resources into the electrical grid provides flexibility that can be assigned to various markets and services. Furthermore, operating in multiple markets fundamentally changes the operational strategy of hydrogen resources, and can increase the amount of hydrogen that can be profitably generated. Additionally, the convex relaxation used for the dispatch of resources is sufficiently accurate to allow a VPP to maintain a network within allowable limits whilst maintaining problem tractability. Lastly, the framework is versatile enough to be utilised by other entities (such as a distribution system operator), or for different purposed (such as techno-economic analysis for business case assessments).
  • Item
    Thumbnail Image
    Distributed multi-agent decision-making for task assignment and collision avoidance
    Khoo, Mitchell Wei Kern ( 2021)
    Multi-agent systems generally involve agents cooperating or coordinating to perform complex tasks. Typically, agents must do so by computing decisions amongst themselves without a centralised decision-maker. In this thesis, a team of agents with different initial locations is tasked with shifting their positions to a set of goal locations. There are two aspects of this type of multi-agent mission that are of interest. The first relates to deciding which agent to assign to which goal location. The second aspect relates to deciding how the agents should move to their goals without colliding with each other along the way. These two aspects can be investigated separately but they are also linked as assigning agents in particular ways can lead to intrinsic collision avoidance between agents. This thesis explores each of these ideas. Assignment problems arise in multi-agent systems when there is a need to allocate a set of tasks to a set of agents. Two types of assignment problems are considered in this thesis. The Bottleneck Assignment Problem (BAP) is an assignment problem where the objective is to minimise the costliest allocation of a task to an agent, while the Lexicographic Bottleneck Assignment Problem (LexBAP) is a related problem with the objective of further minimising the allocation costs of the remaining agents and tasks. These two types of assignment problems are highly applicable in time-critical applications, where there is a need for agents to complete tasks such that the worst-case completion time is minimised. Hence, many centralised algorithms exist to solve them. However, it has become increasingly desirable to produce solutions without relying on a centralised decision-maker. In particular, there is a need for distributed algorithms to solve these problems that do not require a centralised decision-maker having access to all information from each agent. Distributed algorithms for assignment problems with other objectives have been proposed, yet to date no such algorithms for either the BAP or the LexBAP exist. In order to address this gap, novel tools for analysing the BAP and LexBAP are introduced. The introduction of these tools precipitates an analysis of structure in the BAP, where in particular, an investigation into how two separate BAPs can be merged into a combined BAP is conducted. Then by applying these concepts, a distributed algorithm for the BAP is presented, a greedy distributed algorithm for the LexBAP is developed and conditions on exactness of the greedy approach are provided. Numerical results are presented for both distributed algorithms to benchmark them against existing approaches. In specific applications where agents are mobile robots that move towards goal locations to complete tasks, it is subsequently prudent to guarantee inter-agent collision avoidance once the assignment problem is computed. Two approaches for providing collision avoidance guarantees are explored. The first approach shows collision avoidance as an intrinsic property of assigning agents according to the LexBAP, which leads to time-varying safe sets of positions for agents. The second approach considers the use of Control Barrier Functions for collision avoidance, which alternatively leads to time-varying safe sets of accelerations for mobile agents modelled as double integrators.
  • Item
    Thumbnail Image
    Multi-service Supply Adequacy Assessment Framework in Renewables - Dominated Power System
    LIU, GUANCHI ( 2021)
    Many challenges have emerged as the penetration of renewables keeps increasing during the low-carbon transition of power system, among which the most fundamental one could be maintaining power system reliability. In conventional system planning exercises, only system adequacy is considered as the criteria to guarantee reliability. However, the conventional system adequacy concept may not guarantee system reliability anymore due to several challenges brought by the increasing penetration of renewables. Those challenges mainly include concerns about system flexibility issue, as well as system resilience issue considering the wind or solar scarcity for days or weeks. Furthermore, the increasing uptake of distributed energy resources (DER) has also brought great challenges to system reliability and resilience planning due to their various operation paradigms. There are many and far-reaching consequences of planning decisions towards a RES-dominated power system, most notably in terms of system reliability and resilience. When it comes to supply reliability and resilience, there is a vibrant interplay between renewable energy resources (RES), transmission networks, and storage in the context of RES integration and their enabling assets. As the penetration of RES keeps increasing, where large-scale wind and solar farm as well as roof-top PV have accounted for the majority of RES installation capacity, it will be critical to determine the role of different types of storage in supporting system reliability (as well as system resilience to extreme events) coordinating with RES at both transmission and distribution levels. A systematic framework is developed for evaluating the contribution of RES and EES to the adequacy, flexibility, and resilience of RES-dominated system generation. The ability of the system to procure sufficient generation and transmission facilities to support system adequacy, flexibility and resilience is referred as the “multi-service supply adequacy” in this work. Additionally, this thesis has performed a thorough cost-benefit analysis of the effects of distributed energy storage (DES) on enhancing the reliability and resilience of both the distribution community (with a specific focus on rural area communities) and the bulk transmission system. Further, the relationship between electrical energy storage (EES) investment and system reliability and resilience performance is clearly illustrated and explained.
  • Item
    Thumbnail Image
    System dynamics of low-carbon grids: fundamentals, challenges, and mitigation solutions
    Ghazavi Dozein, Mehdi ( 2021)
    Going towards low-carbon grids results in massive grid integration of converter-interfaced technologies and its consequent phase-out of synchronous generators, which could potentially threaten system frequency stability/resilience, associated with low-inertia conditions and inadequate frequency control ancillary services, and voltage stability/control, associated with low system strength and weak-grids. Therefore, system operators need to figure out alternative solutions to provide system dynamic supports of various types (i.e., voltage/reactive power and frequency/active power supports) in order to guarantee a stable and secure system operation. This requires a deep understanding of the dynamics of low-carbon grids, as well as the emerging converter-based technologies, under weak and low-inertia conditions. This thesis presents the principles of low-carbon grid dynamics and the modelling foundations for converter-based resources of various types, i.e., battery, electrolyzer, and photovoltaics, to study their potential benefits and challenges in system dynamic supports. Regarding large-scale batteries, their actual capabilities in multiple dynamic supports are discussed while highlighting the technical challenges/interactions associated with low system strength and frequency-dependent protection schemes. Also, the impacts of reactive power prioritization on the battery converter stability, as well as the external system voltage stability, are studied while highlighting the necessity of specific control design requirements and response characteristics at weak distribution/transmission connections. With respect to hydrogen electrolyzers, the proposed detailed dynamic model takes into account electrolysis stack thermodynamics, power-electronics interface, converter control loops, and operational constraints in hydrogen production. The proposed model is then suitable to study the potential capabilities/challenges of electrolysis units in frequency control. Regarding photovoltaics dynamic modelling, a dynamic equivalent model is proposed based on a novel closed-loop system identification approach. The proposed dynamic equivalencing approach is able to capture with good fidelity the aggregate frequency response from solar units while reducing the complexity in combined transmission-distribution frequency response analysis. The potential stability and resilience benefits from the mentioned resources, and the role of different design setups and operational conditions/requirements, are studied following credible and non-credible extreme events in the context of realistic Australian case studies.
  • Item
    Thumbnail Image
    Optical microspectrometers and chemical classifiers based on silicon nanowires, plasmonic metasurfaces and machine learning
    Meng, Jiajun ( 2021)
    Spectrometers are a workhorse tool of optics, with applications ranging from scientific research to industrial process monitoring, remote sensing, and medical diagnostics. Although benchtop systems offer high performance and stability, alternative platforms offering reduced size, weight and cost could enable a host of new applications, e.g. in consumer personal electronics and field-deployable diagnostic platforms. To contribute to this trend towards miniaturised optical systems including spectrometers, this thesis presents the realisation of a visible spectrum microspectrometer using structurally coloured silicon nanowires and a reconstruction algorithm. We also experimentally demonstrate a plasmonic mid-infrared filter array-detector array microspectrometer that uses machine learning to determine the chemical compositions of a variety of liquids and solids. In this dissertation, we first present a reconstructive microspectrometer based on vertical silicon nanowire photodetectors. The nanowire photodetectors are designed to have absorption peaks across the visible spectrum. The spectral positions of these peaks are controlled by the radii of the nanowires. The nanowire detectors sit on substrate mesas that also serve as photodetectors for the light transmitted through the nanowires. We demonstrate the fabrication of this device, which has a footprint of a few millimetres. We use it as a spectrometer for the visible spectrum by implementing reconstructive algorithms. The identification of chemicals from their mid-infrared spectra has applications that include the industrial production of chemicals, food production, pharmaceutical manufacturing, and environmental monitoring. This is generally done using laboratory tools such as the Fourier transform infrared spectrometer. To address the need for fast and portable chemical sensing tools, we demonstrate the concept of a chemical classifier based on a filter array-detector array mid-infrared microspectrometer and a machine learning classification algorithm. Our device consists of a thermal camera onto which we have added an array of plasmonic filters. We perform simulations to find design parameters to enable the filters to have spectral features covering the wavelength range of interest. We first investigate this concept via a simulation study. We simulate the data that the device would generate when subjected to different chemicals, including noise. The simulated data is collated to train machine learning classification models. Our model predicts that this approach would be able to classify liquid and gas chemicals with very high accuracy. We later verify this concept by experimentally demonstrating a liquid chemical classifier. We design and fabricate a gold plasmonic filter chip containing 20 filters. The chip is integrated into a thermal camera to realise the mid-infrared microspectrometer platform. We train classifiers using the collected readout data of liquid analytes. The trained liquid classifier can accurately identify each type of analyte.
  • Item
    Thumbnail Image
    Bounded Estimation and Stabilization over Channels with Memory
    Saberi, Amir ( 2021)
    In this thesis, we investigate state estimation and stabilization of linear systems over discrete channels with memory, such that the channel output at each time depends on the current and past input and noise signals. By extending the classical zero-error capacity concept, new notions of uniform zero-error capacity (with and without feedback) for such channels are introduced. We show that similar to memoryless channels, uniform zero-error capacity, \(C_0\) has to be larger than the topological entropy of the system, \(h_{lin}\) for achieving uniformly bounded estimation errors. We also show that if \(C_0< h_{lin}\), uniformly bounded estimation is impossible. Furthermore, we investigate the stabilization problem over such channels and show that the uniform zero-error feedback capacity, \(C_{0f}\) of the channel is the tight figure of merit for achieving uniformly bounded states. We then consider finite-state, worst-case versions of the common additive noise and erasure channel models, in which the noise is governed by a finite-state machine without any statistical structure. These models can capture the case when noise patterns are correlated, instead of assuming i.i.d. noise in the channel, and better reflect communication situations in which network congestion or wireless fading can cause bursty error patterns that are difficult to model stochastically or that the underlying probabilities vary significantly over time. We explicitly compute \(C_{0f}\) for both types of channels. We show that the zero-error feedback capacity of the additive noise channel is either zero or \(C_{0f} =\log q -h_{ch}\), where \(h_{ch}\) is the topological entropy of the noise process and \(q\) is the input alphabet size. A topological condition is given to determine when the zero-error capacity with or without feedback is zero. Combining the \(C_{0f}\) formula of additive noise channels with the results on stabilization gives a ``small-entropy theorem'', stating that a linear system can be stabilized over finite-state additive noise channels if the sum of the topological entropies of the linear system and the channel is smaller than \(\log q\). Furthermore, we show \(C_{0f} =(1- \tau) \log q\) for finite-state erasure channels, where \(\tau\) is a topological property of the finite-state machine and called maximal ratio. Several examples, including sliding-window and consecutive-window channels, are considered.
  • Item
    Thumbnail Image
    A Block Coordinate Descent approach for solving Graph SLAM
    Garces Almonacid, Javier Andres ( 2021)
    Simultaneous Localisation and Mapping (SLAM) refers to the problem of estimating the position of a mobile robot navigating in an unknown environment while simultaneously constructing a map of it, using measurements collected by sensors mounted on the robot, such as cameras, lasers, radars, or inertial sensors. SLAM is of particular interest when there is no prior knowledge of the environment nor external sources of localisation (compass, GPS). In this sense, SLAM aims for autonomy of robot motion and environment discovery. The graph-based formulation of the SLAM problem, also commonly referred to as Graph SLAM, maximum a posteriori estimation, factor graph optimisation or smoothing and mapping (SAM), is considered the current de-facto standard formulation for SLAM. This approach defines the SLAM problem as a nonlinear least squares minimisation problem, commonly solved via successive linearisation methods such as Gauss-Newton. However, iterative line search methods have limitations in terms of convergence guarantees and scalability, which suggest the research potential for alternative optimisation algorithms. In our research, we study an alternative numerical method for solving the Graph SLAM problem: the Block Coordinate Descent method. By partitioning the problem into a series of optimisation subproblems, this approach may offer comparatively better performance than iterative linearisation algorithms, such as lower per-iteration computational complexity, scalability and parallel processing capabilities. Importantly, this method is not dependant on linearisation, and under certain conditions, may offer convergence guarantees towards stationary points. We present our Block Coordinate Descent approach by systematically analysing the attributes of the optimisation subproblems originating from the use of this numerical method on a Graph SLAM problem formulation based on particular inertial, bearing and range measurement models: the Affine Motion Model, the Affine Bearing Model and the Squared Range Model. We verify the resulting optimisation subproblems satisfy conditions that offer convergence guarantees and scalability properties. Additionally, we evaluate our Block Coordinate Descent approach by implementing the resulting algorithm in a simulated environment using real-world datasets, comparing its performance to the Gauss-Newton line search method.
  • Item
    Thumbnail Image
    Advanced Direct Detection Receivers for Short-Reach Optical Transmission Systems
    Ji, Honglin ( 2021)
    The advantages of optical field recovery in coherent detection have generated an upsurge of research on direct detection with the same capability. The key reason for the interest in direct detection is that the direct detection receiver has a simple receiver structure without an expensive and narrow-linewidth local oscillator. However, in the conventional direct detection receivers, it can do intensity-only detection and lose the phase information, which generates capacity scaling challenges. Moreover, the conventional direct detection suffers from the fiber chromatic dispersion induced frequency-selective power fading, which significantly degrades the transmission performance and restrains the transmission distance. Therefore, advanced direct detection receivers should be investigated and studied, which aims to combine the advantages of both coherent detection and conventional direct detection. Namely, an advanced direct detection receiver can recover the optical field information by simple front-end but without a local oscillator. To achieve advanced direct detection, self-coherent detection could be investigated to recover the optical field similar to coherent detection. Due to the self-coherent detection, the computational complexity of digital signal processing for carrier and phase recovery can be significantly reduced and the tight requirements on the transmitter laser source regarding the wavelength stability and laser linewidth could be greatly relaxed since the sophisticated wavelength alignment between transmitter and receiver is no longer needed. With the recovered optical field, various channel impairments during transmission, such as chromatic dispersion, imperfect channel response, and fiber nonlinearity, can be mitigated by powerful digital signal processing. Consequently, high-order modulation formats can be utilized to achieve high spectral-efficiency transmission. The transmission reach can also be greatly extended without the intrinsic effects of frequency selective power fading and signal-signal beating interference in the conventional direct detection systems. Furthermore, to achieve higher-capacity advanced direct detection receiver, higher-degree of freedom modulation and multiplexing schemes could be employed for short-reach optical interconnects such as polarization/mode-division-multiplexing transmission systems based on advanced direct detection schemes. In this thesis, we exploit the advanced direct detection techniques for single-polarization, dual-polarization, few-mode transmission of a complex-valued double-sideband signal in short-reach optical transmission systems. To bridge the gap between direct detection and coherent detection, we propose generalized carrier assisted differential detection for the reception of the single-polarization complex-valued double-sideband signals. The concept of carrier assisted differential detection is extended to a general selection of the transfer functions, beyond the originally-proposed delay interferometer. With the silicon photonics micro-ring resonator-based optical filter, both the required carrier-to-signal power ratio and the optical signal-to-noise ratio sensitivity are drastically improved due to the significantly suppressed signal-signal beating interference. For polarization-multiplexed transmission, we firstly propose a hybrid single-polarization coherent receiver and Stokes vector receiver for polarization-diversity self-coherent detection. The required constant optical carrier is remotely delivered from the transmitter without a need for optical polarization control to further reduce the cost, footprint, power consumption of the transceivers for short-reach optical interconnects. To avoid the separate optical fiber for delivering the constant optical carrier for the receiver, we propose a carrier-less four-dimensional direct detection receiver based on differential polarization inner product and Stokes vector. The proposed four-dimensional direct detection receiver can recover the intensities of both polarizations and the differential phase and differential common phase. For mode-division multiplexing transmission in few-mode fibers, we propose, for the first time, a high-dimensional Stokes vector direct detection receiver for the coupled spatial modes without resorting to coherent detection. With the proposed advanced direct detection techniques, this thesis provides some novel approaches for achieving high-spectral-efficiency direct detection transmission and could be promising for short-reach optical transmission systems.
  • Item
    Thumbnail Image
    Sum-Rate Optimization in Wireless Networks
    Dayarathna, Mirinchi Sattambige Shalanika Gangani ( 2021)
    Due to technological enhancements, such as mobile TV, streaming services, smart grids and augmented reality, the need to optimize the overall network throughput becomes one of the most important aspects of communication systems. In this thesis, we analyze the resource allocation problem of four different wireless communication systems with the objective of maximizing the achievable sum-rate. From a theoretical perspective, the achievable sum-rate optimization problem is very hard to solve. As such, the challenge is to solve these non-convex and NP-hard optimization problems with an acceptable level of complexity. We first consider two full-duplex and half-duplex networks and investigate the optimality of binary power allocation. In terms of the transmit power allocation, the simple power allocation method known as the binary power allocation is known to be optimal for special wireless network structures. Therefore, we derive new, necessary and sufficient conditions that would ensure the optimality of binary power allocation for such networks. We further extend our analysis to investigate the sub-optimality associated with binary power allocation and show that the loss of achievable sum-rate due to binary power allocation is negligible. Next, we concentrate on the transmission direction control in flexible duplex networks such that the achievable sum-rate is maximized. We analyze several existing approximation techniques and their feasibility for solving the resulting NP-hard optimization problem. We also develop a novel low-cost heuristic pattern search algorithm based on the direct search of function value. For a flexible network where each node has one potential desired link, we observe that the proposed algorithm has better achievable sum-rate compared to the approximation techniques and other existing resource allocation approaches. In reality, however, a node can have more than one potential desired link. As such, transmitter and receiver scheduling becomes important in the presence of multiple desired links. Therefore, we next investigate the optimum resource allocation in terms of transmit power, link direction and transmitter/receiver selection of flexible half duplex networks when the objective is to maximize the achievable sum-rate. We prove that it is optimum for each transmitting node to only transmit to one desired receiving node and use that insight to develop a novel iterative algorithm. We observe that the proposed algorithm outperforms existing resource allocation techniques specially in denser areas where larger number of nodes are operating. Finally, we consider the achievable sum-rate optimization problem in multi-user multi-hop relay networks. We derive novel closed form expressions for the achievable rate under a single-user multi-hop relay network. We extend this investigation to multi-user network and develop a novel iterative algorithm for joint relay selection and power control with the objective of maximizing the achievable sum-rate. To obtain further insights, we extend our analysis to two-user multi-hop network and prove that optimal power allocation can be analytically obtained by considering binary power allocation for two transmitting nodes. In addition, we develop a novel low-cost relay selection technique based on the average interference estimation for the special case of multi-user dual hop relay networks.
  • Item
    Thumbnail Image
    Model Predictive Controller Tuning by Machine Learning and Ordinal Optimisation
    Chin, Robert Alfred ( 2021)
    While for the past several decades model predictive control (MPC) has been an established control strategy in chemical process industries, more recently there has been increased collaboration in MPC research between academia and automotive companies. Despite the promising work thus far, one particular challenge facing the widespread adoption of MPC in the automotive industry is the increased calibration requirement. The focus of the research in this thesis is to develop methods towards reducing the calibration effort in designing and implementing MPC in practice. The research is tailored by application to offline tuning of quadratic-cost MPC for an automotive diesel air-path, to address the limited time-availability to perform online tuning experiments. Human preferences can be influential in automotive engine controller tuning. Some earlier work has proposed a machine learning controller tuning framework (MLCTF), which learns preferences from numeric data labelled by human experts, and as such, these learned preferences can be replicated in automated offline tuning. Work done in this thesis extends this capability by allowing for preferences to be learned from pairwise comparison data, with monotonicity constraints in the features. Two methods are proposed to address this: 1) an algorithm based around Gaussian process regression; and 2) a Bayesian estimation procedure using a Dirichlet prior. These methods are successfully demonstrated in learning monotonicity-constrained utility functions in time-domain features from data consisting of pairwise rankings for diesel air-path trajectories. The MLCTF also constitutes a plant model, yet there will typically be some uncertainty in an engine model, especially if it has been identified from data collected with a limited amount of experimentation time. To address this, an active learning framework is proposed for selection of the next operating points in the design of experiments, for identifying linear parameter-varying systems. The approach is based on exploiting the probabilistic features of Gaussian process regression to quantify the overall model uncertainty across locally identified models, resulting in a flexible methodology which accommodates for various techniques to be applied for estimation of local linear models and their corresponding uncertainty. The framework is applied to the identification of a diesel engine air-path model, and it is demonstrated that measures of model uncertainty can be quantified and subsequently reduced. To make the most of the limited availability for online tuning experiments, an ordinal optimisation (OO) approach is proposed, which seeks to ensure that offline tuned controllers can perform acceptably well, once tested online with the physical system. Via the use of copula models, an OO problem is formulated to be compatible with the tuning of controllers over an uncountable search space, such as quadratic-cost MPC. In particular, results are obtained which formally characterise the copula dependence conditions required for the OO success probability to be non-decreasing in the number of offline controllers sampled during OO. A gain-scheduled MPC architecture was designed for the diesel air-path, and implemented on an engine control unit (ECU). The aforementioned non-decreasing properties of the OO success probability are then specialised to tuning gain-scheduled controller architectures. Informed by these developments, the MPC architecture was firstly tuned offline via OO, and then tested online with an experimental diesel engine test rig, over various engine drive-cycles. In the experimental results, it was found that some offline tuned controllers outperformed a manually tuned baseline MPC, the latter which has comparable performance to proprietary production controllers. Upon additional manual tuning online, the performance of the offline tuned controllers could also be further refined, which illustrates how offline tuning via OO may complement online tuning approaches. Lastly, using an analytic lower bound developed for OO under a Gaussian copula model, a sequential learning algorithm is developed to address a probabilistically robust offline controller tuning problem. The algorithm is formally proven to yield a controller which meets a specified probabilistic performance specification, assuming that the underlying copula is not too unfavourably far from a Gaussian copula. It is demonstrated in a simulation study that the algorithm is able to successfully tune a single controller to meet a desired performance threshold, even in the presence of probabilistic uncertainty in the diesel engine model. This is applied to two case studies: 1) `hot-starting' an online tuning procedure; and 2) tuning for uncertainty inherent across a fleet of vehicles.