Masters Degrees (Electrical and Electronic Engineering)

Permanent URI for this collection


Recent Submissions

Now showing 1 - 5 of 947
  • Item
    Machine learning models for mass appraisals: advancing valuations in the digital era
    (Stellenbosch : Stellenbosch University, 2023-12) de Wet, Dominique; du Preez, Johan; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT: A property appraisal is a professional, unbiased valuation that determines the market value of a property. Traditionally, property appraisals were conducted exclusively by professional appraisers. However, with increased data availability and enhanced computational resources, automated valuation models (AVMs) have gained widespread recognition as efficient tools to assist property appraisers in conducting mass appraisals. This report investigates the suitability of various machine learning (ML) methods as AVMs. The techniques include multiple polynomial regression, random forest regression, support vector regression, and a neural network. In addition to these four models, this report also introduces and assesses the advantages of a new innovative fusion model as an AVM. The fusion model is an ensemble approach that employs a neural network to combine the predictions from the four previous ML models, aiming to achieve improved accuracy and precision. This report uses property sales data from three specific neighbourhoods located within the Western Cape province of South Africa: Edgemead, Pinehurst, and Brackenfell. The results from this study indicate that all the individual ML techniques produce highly accurate property price predictions. However, they also yielded relatively large errors for some predictions. In contrast, the fusion model achieved greater accuracies and minimised most of its errors compared to the standalone models, establishing it as an effective AVM technique. This report provides a comprehensive framework for improving mass appraisal models.
  • Item
    Modelling the broadband impedance of a lithium-ion battery cell using the pseudo-random impulse sequence perturbation
    (Stellenbosch : Stellenbosch University, 2023-12) Moyo, Zandile; Mwaniki, Fred; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT: Lithium-ion (Li-ion) batteries are becoming increasingly important for integrating renewable energy sources such as solar and wind into the electrical grid. Battery Energy Storage Systems (BESS) using Li-ion batteries address challenges posed by renewable energy, such as unpredictability and non-dispatchability, thus enhancing grid reliability. Before implementing Li-ion cells in BESS, thorough investigation and optimisation are essential. Accurate prediction of remaining useful life and State-of-Health (SOH) under varying conditions, such as temperature and State-of-Charge (SOC), is vital for effective battery management. Battery impedance, which is frequency-dependent, significantly impacts SOH and SOC, especially in the 1 Hz to 1 kHz range. Therefore, modelling the broadband impedance of the Li-ion battery helps to improve battery management and reliable operation. Accurate battery cell modelling also enables system integrators to perform effective network simulations. The impedance of a Li-ion battery is often computed from its response voltage and current signals to an injected excitation signal. This thesis uses the novel Pseudo-random Impulse Sequence (PRIS), a broadband bipolar excitation signal, for Li-ion battery impedance measurements. The PRIS excitation source is optimised, by tuning its series resistor, inductor and capacitor (RLC) network components and Direct Current (DC) voltage input, to ensure sufficient spectral energy for impedance testing without damaging the battery. The impedance is determined from the battery’s excitation response voltage and current signals using Fourier techniques. An excitation method is proposed for accurate Li-ion cell impedance measurements. This method involves separating the desired frequency range into multiple bands that are perturbed separately using optimised excitation parameters to improve the measurement signal-to-noise ratio, reduce variance, and minimise bias. Parameter Estimation (PE) techniques, via the physics-based Equivalent Circuit Model (ECM) approach, are used to characterise and model Li-ion battery cell impedance. A simulated PE is performed in MATLAB, while a PE using physical experimental data is performed in ZView software. The SOC of the Li-ion cell is varied through discharge tests, and the cell impedance is measured at each SOC using the proposed excitation method. This research demonstrates that the optimised PRIS source and the proposed excitation method improve the accuracy of the measurement of the Li-ion battery impedance, thus enhancing the performance of battery modelling. The excitation responses of the validated Li-ion battery cell model, from PE, closely match the actual system responses across all relevant frequencies. The influence of SOC on impedance is effectively observed, especially at mid to lower frequencies. This thesis confirms that the PRIS and the proposed excitation method are suitable for Li-ion battery impedance measurements and modelling. It is evident that the PRIS excitation signal, along with the proposed method, results in good Li-ion cell impedance measurements with minimal noise and bias. This leads to reliable estimation of the ECM parameters that replicates the physical responses of the battery at varying SOC, thus benefiting future simulations and system integration.
  • Item
    Grid-connected hybrid energy system modeling and optimization study for green hydrogen production in South Africa
    (Stellenbosch : Stellenbosch University, 2023-12) Mukoni, Esmerelda; Garner, Karen; Van Staden, Chantelle; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT: This study proposes a grid-connected hybrid energy system optimization model for green hydrogen production. The grid-connected hybrid energy system consists of wind and solar PV systems, which supply green hydrogen production facilitated by a proton exchange membrane electrolyzer. With the intermittent nature of wind and solar resources, the electrical grid supplies renewable energy to satisfy the electrolyzer load demand. The proposed grid-connected hybrid energy system optimization model is determined using a constrained multi-objective, non-dominated sorting genetic optimization algorithm implemented in Pymoo, an open-source Python framework. The optimization model aims to minimize the cost of electricity purchased from the electrical grid and maximize efficiency at high reliability. The cost of electricity and reliability are based on the time-of-use tariff structure and loss of power supply probability respectively. The non-dominated genetic algorithm successfully converges to a Pareto front solution set, and the optimal solution is determined, including the optimal performance parameters of the wind turbine and number of solar PV modules. The optimal performance parameters provide a guideline for choosing the optimal wind turbine model and solar photovoltaic module. For evaluation and validation purposes, the developed grid-connected hybrid energy system optimization model is applied to a case study of six renewable energy development zones in South Africa. A grid-connected hybrid energy system optimization model which takes wind and solar resources, as well as load input to calculate the optimal wind and solar energy mix is successfully developed. As a result, the optimal wind turbine, solar PV module and inverter as well as the number of solar PV modules that result in an optimal grid-connected hybrid energy system are successfully obtained.
  • Item
    Development of a frequency synthesizer for a low-cost radio interferometer
    (Stellenbosch : Stellenbosch University, 2023-12) Lotter, Kyle; Steyn, Werner; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT: This document contains a description of the basics of a phase-locked loop when used as a frequency synthesizer. The document also contains an introduction to phase noise and a description of how to model and simulate phase noise in a phase-locked loop. Three different phase-lock loops are then designed, simulated and measured with the goal of generating local oscillator signals for an array of superheterodyne receivers.
  • Item
    Design and optimization of a large scale grid connected wound rotor synchronous machine
    (Stellenbosch : Stellenbosch University, 2023-12) Siphepho, Nosimilo; Garner, Karen; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT: The main focus of this work is the cost-efficient design of a 3 MW, geared, medium speed wound rotor synchronous machine for grid-tied wind applications. Following this concept, a fractional slot non-overlapping winding is employed. To deal with the adverse effects of such a winding, methods of mitigating magneto-motive force harmonics are studied. The effects of phase shifting, flux barrier insertion, and pole shaping on the machine performance of the 18 slots, 16 poles, double layered wound rotor synchronous machine are explored. The machine is also optimized to make it grid compliant. Finally, the effects of phase shifting are tested on a 3 KW prototype.