School of Electrical & Information Engineering (ETDs)
Permanent URI for this communityhttps://hdl.handle.net/10539/37968
Browse
24 results
Search Results
Item Navigating the Underground: Assessing Vision-Based SLAM Methods in Simulated Subterranean Scenarios(University of the Witwatersrand, Johannesburg, 2024) Steenkamp, Dani¨el Johannes; Celik, TurgayThis dissertation explores the viability of vision-based localization methods in subterranean environments, employing a variety of feature extraction techniques including traditional methods and advanced deep learning approaches. A unique dataset was generated using an autonomous exploration UAV within a simulated subterranean environment. This dataset served as the testing ground for evaluating various feature extraction methods. The ORB-SLAM3 was modified to integrate these methods, adapting its feature extraction module to accommodate alternative approaches while retaining its core pose optimization and backend components. The study includes detailed experiments and analyses of different sensor configurations and feature extraction methods, providing insights into their applicability and performance in subterranean settings.Item Investigating Design Parameters for Practical Load Forecasting of Grid-Interactive Buildings Using LSTM(University of the Witwatersrand, Johannesburg, 2024) Simani, Kyppy Ngaira; Yen , Yu-chieh; Genga , YuvalThe dissertation presented contributes to research in residential load forecasting using machine-learning (ML) models to optimise energy management within grid- interactive efficient buildings (GEB) and addresses the challenge of implementing forecasting models for practical applications. The use of accurate load forecasting ML models has been shown to extend and im- prove the field of energy management towards the resource coordination of GEB. Developing these models is a time-, data- and compute-intensive process that re- quires selection and tuning of various design parameters. In practice however this has costs and limitations: collection and storage of real-time data, measurement device costs and finite compute power. To simplify model development and reduce these costs, this dissertation investigates to what extent the relationships between model parameters and performance can inform ML design decisions. In the research presented, 3 experiments are performed to investigate the relation- ships between training data size, prediction horizon length and data resolution and their impact on the predictive model fidelity. Several prediction models are devel- oped using various parameter settings. The accuracy and compute intensity are measured to evaluate the influence of each parameter on the model performance. A dataset of 3 years and 11 months of residential power consumption measure- ments at a one-minute resolution is used as the load profile for this investigation. For this load profile it is found that increasing training data size increases compute time linearly, with an exponential decay in model prediction error. This results in a maximum, resource-efficient training data size of 450 days. In addition, the effec- tive prediction (i.e prediction models with R2 scores greater than 0) horizon length is found to be 3 hours. At this length the load profile is suited for short-term load forecast applications such as distributed energy resource (DER) and real-time pric- ing (RTP) management. Furthermore, models trained on low-resolution data (up to 30 minutes) can achieve comparable performance to higher-resolution models with at least 3 additional months of training data. The findings of this investiga- tion therefore represent a contribution towards the development of a ML design tool for more efficient design parameter tuning considering practical load forecast- ing conditions. This shows potential to enable more efficient implementations of load forecasting systems for GEB environments.Item Performance Modeling of Cognitive NOMA-aided IoT Networks with Energy Harvesting(University of the Witwatersrand, Johannesburg, 2024) Selematsela, Neo Edwin; Takawira, Fambirai; Chabalala, ChabalalaIn an attempt to address rocketed connectivity and bandwidth demands in 5G wireless networks, Non-Orthogonal Multiple Access (NOMA) and Cognitive Radio (CR) concepts have been proposed. The former addresses increased connectivity requirements by allowing multiple users in the same NOMA group to utilise the same channel resources. The latter enhances spectrum efficiency by intelligently al- lowing spectrum sharing between primary and secondary networks, if secondary to primary network interference is properly managed. To prolong connectivity/ser- vice life-time of battery capacity constrained Internet of Things (IoT) devices, Energy Harvesting (EH) technique has been identified as the technology that can enable such devices to harvest energy from ambient sources present in the envi- ronment. This research work is motivated by the observed surge in adoption of IoT devices around the globe. The resulting adoption has brought about the need to investigate performance of different IoT system models and hence, understand potential applicability and optimization options for different services. The focus of this dissertation is to model and analyse the performance of an EH Cognitive Radio Non-Orthogonal Multiple Access (CR-NOMA) IoT network. To accomplish this, a simplified energy harvesting CR-NOMA IoT network is considered. The considered network consists of primary and secondary network components. The primary network contains Macro Base-Station (MBS) and Pri- mary Network users (PUs), while the secondary network is made up of Secondary Base-Station (SBS) and multiple CR-NOMA groups containing two Secondary Users (SUs) each. To analytically capture the stochastic nature of energy harvest- ing process and cater for residual energy from one transmission frame to the next, each SU’s energy level in the battery is discretized to represent the state of each SU during each transmission frame; with this, we derive a complete Markovian model for the considered system model using queueing theory and Markovian analysis. Two Markovian models are developed for the considered system model, with one assuming that the SUs are harvesting energy from the SBS (one energy source) and the other, adopting an assumption that energy is harvested from both the SBS and MBS (two energy sources). The considered system performance is analysed in terms of up-link system outage probability and mean capacity. To provide detailed insights, closed-form analytical expressions for up-link outage probability and mean capacity for each user in the CR-NOMA group are derived using the Markovian models as the ba- sis. Produced analytical results are confirmed through simulations using Matlab. Simulation results matched the analytical results, this confirmed the validity of the derived analytical expressions for SUs outage probability and mean capacity. ii Both performance metrics are studied and the impact of varying different network parameters on outage probability and mean capacity is investigated. For out- age probability, results are generated which demonstrate SUs outage performance as we vary Signal-to-Interference-plus-Noise Ratio (SINR), interference threshold, and battery power level. Similarly, mean capacity results are generated to illus- trate each SU mean capacity performance while varying their battery levels, this is done for different values of primary transmit power and interference threshold. Performance results observed as different parameters are varied for outage prob- ability and mean capacity align with the theoretical performance expected when those parameters are changed. The significance of this work lies in providing ana- lytical tools to assess the performance of the CR-NOMA IoT system with energy harvesting (EH). These tools enable easy computation of system performance in- dicators such as outage performance and mean capacity. Attempting the same assessment through simulation would be a cumbersome process.Item Estimating Resistance and Performance of Earthing Systems Electrode in Variably Saturated Soil Conditions(University of the Witwatersrand, Johannesburg, 2024) Nnamdi, Onyedikachi Samuel; Gomes, ChandimaThe design and determination of post-installed resistance of earthing systems are significantly influenced by subsoil resistivity profiles, which are prone to seasonal variations due to environmental and climatic changes. These fluctuations can compromise operational safety and reliability of transmission systems, necessitating periodic monitoring of earthing installations as recommended by national and international standards. However, compliance with these recommendations is often impractical due to the vast number of earthing installations and associated costs. To address this challenge, this thesis proposes a novel multiphysics earthing model that integrates hydraulic and electrical properties of subsoil and earthing enhancement materials (EEMs) with climatic parameters to predict earthing resistance under varying conditions. The model, developed by coupling partial differential equations governing electric current dispersion and fluid retention in porous media, is validated through COMSOL Multiphysics® simulations of vertical earth rods in single and double subsoil layers. The results demonstrate that earthing resistance variation is dependent on subsoil texture, water content, and distribution of soil water potential, which determines subsoil resistivity. The proposed method achieves a relative error range of 2.72% to 6.53% and 1.47% compared to analytical and finite element method solutions, ensuring accuracy and validity. This innovative approach enables site-specific and climate-adaptive assessments of EEM effectiveness, facilitating informed decisions for earthing improvements in diverse conditions, and ultimately optimising material selection and recommendation for various soils and climates.Item Comparative Study on the Accuracy of the Conventional DGA Techniques and Artificial Neural Network in Classifying Faults Inside Oil Filled Power Transformers(University of the Witwatersrand, Johannesburg, 2024) Mokgosi, Gomotsegang Millicent; Nyamupangedengu , Cuthbert; Nixon , KenPower transformers are expensive yet crucial for power system reliability. As the installed base ages and failure rates rise, there is growing interest in advanced methods for monitoring and diagnosing faults to mitigate risks. Power transformer failures are often due to insulation breakdown from harsh conditions like overloading, that leads to prolonged outages, economic losses and safety hazards. Dissolved Gases Analysis (DGA) is a common diagnostic tool for detecting faults in oil-filled power transformers. However, it heavily relies on expert interpretation and can yield conflicting results, complicating decision-making. Researchers have explored Artificial Intelligence (AI) to address these challenges and improve diagnostic accuracy. This study investigates using Machine Learning (ML) techniques to enhance DGA for diagnosing power transformers. It employs an Artificial Neural Network (ANN) with Feed Forward Back Propagation, a Bayesian Regularizer for predictions, Principal Component Analysis (PCA) for feature selection and Adaptive Synthesizer (ADASYN) for data balancing. While traditional DGA methods are known for their accuracy and non- intrusiveness, they have limitations, particularly with undefined diagnostic areas. This research focuses on these limitations, to demonstrate that ANN provides more accurate predictions compared to conventional methods, with an average accuracy of 76.8% versus lower accuracies of 55% for Dornenburg, 40% for Duval, 38.4% for Roger and 31.8% for IEC (International Electrotechnical Commission) Methods. The study findings prove that ANN can effectively operate independently to improve diagnostic performance.Item Breakdown Strength Influences of Titanium Dioxide Nanoparticles on Midel Canola-Based Natural Ester oil: A Comparison Between the Anatase and Rutile Phases of Titanium Dioxide(University of the Witwatersrand, Johannesburg, 2024) Miya, Mabontsi Koba; Nixon, KenNatural ester oils are an alternative solution for sustainable transformer insulation. They offer good dielectric properties and in addition improve safety of equipment and sustainable environment. They have higher fire resistance than the widely used mineral oil and are less prone to explosions. They are also highly biodegradable and renewable. However, some challenges such as inconsistent breakdown voltage at higher temperatures and higher streamer speeds hinder the wide use of natural esters. Nanotechnology has been found to improve the properties of the oil, including the breakdown voltage. Different nanoparticles have been previously studied, giving varying results. This dissertation presents a study of the use of two phases of TiO2 nanoparticles, namely rutile and anatase, to improve the breakdown voltage of natural ester oil at higher temperatures. The study seeks to find the effects of the nanoparticle phases on the oil under uniform and non-uniform electric fields. Nanofluids of different loading concentrations (0.01 vol%, 0.03 vol%, 0.05, vol%) were created in each nanoparticle phase for the purpose of the study. The findings are that both phases of the nanoparticles improve the breakdown voltage under uniform fields. The anatase portrayed an impressive improvement of 85% at ambient temperature, while the rutile phase enhanced by 61%. At higher temperatures however, the rutile phase had better improvement. Rutile TiO2 nanoparticles consistently outperformed the anatase phase in improving the breakdown voltage at higher temperatures. Under non-uniform electric fields, the rutile TiO2-based nanofluid was found to be superior to the anatase-based fluid. Rutile TiO2 resulted in a significant 10% improvement in the average breakdown voltage and streamer acceleration voltage. An overall decrease in the streamer speeds was observed with the addition of the rutile TiO2 nanoparticles. In contrast, anatase TiO2 resulted in decreased breakdown voltage and increased streamer speeds when compared to both the rutile nanofluid as well as the pure natural ester oil. The rutile phase of TiO2 can be regarded as a feasible solution for breakdown voltage improvement of natural ester oil in both cases of uniform and non-uniform electric field. The effects are attributed to the electron capture phenomenon and the good thermal stability of rutile TiO2. A stable composite is formed between the rutile nanoparticles and the host natural ester. The resultant morphological structure enables stable interfacial regions even at higher temperatures. In conclusion therefore, rutile TiO2 nanocomposite natural fluid is a possible solution to the current limitations in ester oils as power transformer insulation oil alternative.Item Characterization of high-frequency time-domain e↵ects arising from the transmission line substitutions of reactive components in a buck converter(University of the Witwatersrand, Johannesburg, 2024) Maree, JohnThe work presented in this dissertation is a continuation of a line of research that suggests that the energy storage components within a DC-DC converter may be a source of high frequency e↵ects in power converter circuits. It is shown that for physically large energy storage compo- nents, conventional models are insucient for modelling the e↵ects of these components and that a transmission line approach is required. Very little work has been done within switching circuits using transmission line theory for the primary components themselves, specifically re- garding the time-domian e↵ects of these components. A significant finding of this work is that it is shown that both in simulation and experimental results these components do indeed have a measurable e↵ect on the output of the converter. Furthermore, this dissertation explores time-domain quantification methods for these distributed e↵ects, and shows that the delay ratio between the transmission lines is a key parameter in determining the magnitude of the e↵ects. This work provides strong experimental evidence for the existence of distributed e↵ects occurring from energy storage components within a DC-DC converter, and indicates that this area of research is worth further investigation. Advancements into our understanding of the high-frequency operation of DC-DC converters have become increasingly rare, necessitating a new perspective. This work focusses on using transmission line theory to model energy storage components within a DC-DC converter, and investigating the e↵ects of doing so. The research firstly introduces the design, simulation and experimental evidence for inductors and capacitors using transmission line theory. In fact, it is shown that in order to accurately model a physically large reactive component, transmission line modelling is required. Thereafter, these components in a physically large form are then applied to a DC-DC buck converter circuit where it is shown that the converter manifests high frequency e↵ects that are not predicted by conventional models, but is adequately shown using transmission line models. The e↵ects of these components are then investigated, and it is shown that the delay ratio between the transmission lines is a key parameter in determining the magnitude of the e↵ects. This work provides strong experimental evidence for the existence of distributed e↵ects occurring from energy storage components within a DC-DC converter, and indicates that this area of research is worth further investigation.Item A Longitudinal Study on the Effect of Patches on Software System Maintainability and Code Coverage(University of the Witwatersrand, Johannesburg, 2024) Mamba, Ernest Bonginkosi; Levitt, SteveIn the rapidly evolving landscape of software development, ensuring the quality of code patches could potentially improve the overall health and longevity of a software project. The significance of assessing patch quality arises from its pivotal role in the ongoing evolution of software projects. Patches represent the incremental changes made to the code-base, shaping the trajectory of a project’s development. The identification and understanding of factors influencing patch quality could possibly contribute to enhanced software maintainability, reduced technical debt, and ultimately, a more resilient and adaptive code-base. While previous research predominantly concentrates on analysing releases as static entities, this study extends an existing study of patch testing while incorporating an examination of quality from a maintainability point of view, thereby filling a void in patch-to-patch investigations. Over 90, 000 builds spanning 201 software projects written in 17 programming languages are mined from two popular coverage services, Coveralls and Codecov. To quantify maintainability, a variant of the SIG Maintainability Model, a recognised metric designed to assess the maintainability of incremental code changes is employed. Additionally, the Change Risk Anti-Patterns (CRAP) metric is utilised to identify and measure potential risks associated with code modifications. A moderate correlation of 0.4 was observed between maintainability and patch coverage, indicating that patches with higher coverage tend to exhibit improved maintainability. Similarly, a moderate correlation was identified between the CRAP metric and patch coverage, suggesting that higher patch coverage is associated with reduced change risk anti- patterns. In contrast, patch coverage demonstrates no correlation with overall coverage, underscoring the distinctive nature of patches. However, it is noted that relying solely on patch coverage lacks comprehensive overview of coverage patterns. Thus, it is recommended to supplement it with overall system coverage for a more comprehensive understanding. Moreover, patch maintainability also exhibits no correlation with overall coverage, again, highlighting the unique nature of patches. In conclusion, the study offers valuable insights into the nuanced relationships between patch coverage, maintainability, and change risk anti-patterns, contributing to a more refined understanding of software quality in the context of software evolution.Item Modelling OAM Crosstalk with Neural Networks: Impact of Tip/tilt and Lateral Displacement(University of the Witwatersrand, Johannesburg, 2024) Makoni, Steven Gamuchirai; Cheng, LingThis research focuses on a critical challenge within Free Space Optical ( FSO) commu- nication systems, specifically those utilizing Mode Division Multiplexing (MDM) with Orbital Angular Momentum ( OAM ) modes of a limited transmission range. Despite these systems’ potential to significantly enhance spectral efficiency and transmission capacity, their effectiveness is hindered by the limited range caused by atmospheric turbulence-induced aberrations. Atmospheric turbulence and mis- alignments distort the optical wavefront, causing degradation in orthogonal spatial modes and resulting in power spreading into adjacent modes, known as crosstalk in MDM systems. This research presents a simple neural network model for estimating OAM crosstalk in FSO systems, specifically focusing on atmospheric turbulence-induced aberrations. Firstly, we generated datasets through simulation and experimentation for validation purposes. We then develop and evaluate the neural network model, assessing its accuracy under various turbulence aberrations. The simple neural network, trained solely on tip/tilt and displacement inputs and without retraining, accurately estimated OAM spectra using approximated inputs in turbulent condi- tions, closely matching experimentally measured spectra. Despite the presence of turbulent aberrations, the model showed a minimal decrease in the coefficient of determination, indicating its ability to generalize well to unseen measurements. Our findings indicate that a simple neural network trained solely on tilt and displacement inputs can accurately estimate OAM crosstalk amidst many turbulence aberrations for ℓ ∈ [-5, 5] as a proof of concept. This implies that simple detectors such as cameras can be used to implement or optimize digital signal processing for error detection and correction utilizing the knowledge of crosstalk, offering promising avenues for improving system efficiency and quality of service for MDM systems. In summary, this research leveraged neural networks to model OAM crosstalk induced by misalignments and turbulence. The model’s ability to estimate OAM crosstalk due to misalignments and atmospheric turbulence shows potential for use in real-time predictive systems. With further refinement, neural network models could indicate the evolution of OAM crosstalk in FSO communications that em- ploy OAM multiplexing schemes in atmospheric turbulence. The demonstrated efficacy of the neural network model positions it as a valuable tool for enhancing the robustness of FSO communications employing higher-order OAM modes.Item Assessment of DC-DC Converter Selection Metrics(University of the Witwatersrand, Johannesburg, 2024) Letsoalo, Future Malekutu; Hofsajer, IvanThe exponential growth of Internet of Things (IoT) devices, powered by diverse energy sources, poses significant challenges in power electronics. Despite advances in DC-DC converter topologies, a gap remains in the literature regarding standardized performance metrics for selecting suitable converters, making the selection process complex. This study critically assesses metrics from seminal works of the 1960s to contemporary state-of-the-art, proposing a systematic approach to converter assessment. Two major categories of metrics are identified: averaging metrics and waveform-preserving metrics. Averaging metrics, grounded in Wolaver's foundational work, are effective for high-level comparisons among many converter options, establishing a performance baseline. The study introduces an average modeling tool to reveal core converter characteristics for objective comparison. Waveform-preserving metrics, on the other hand, provide detailed performance insights and are suitable for a narrower set of converter options. The study further categorizes these metrics to assess converter switches and reactive components. A new RMS metric is proposed, refining the existing processed power metric for better accuracy. By integrating both averaging and waveform-preserving metrics at relevant design stages, this study offers a systematic framework for converter assessment. This approach bridges the gap between high-level comparison and detailed performance evaluation, facilitating informed decision-making in converter selection.
- «
- 1 (current)
- 2
- 3
- »