Faculty of Engineering and Built Environment

Permanent URI for this communityhttp://ir-dev.dut.ac.za/handle/10321/9

Browse

Now showing 1 - 2 of 2

Kernel estimation modelling and optimization of hybrid power system for a typical South African rural area
(2023-09) Magenuka, Thand’uxolo Kenneth; Kabeya, Musasa; Akindeji, Kayode Timothy
To increase the accessibility of electricity even to those rural sparsely scattered isolated rural regions, renewable energy seems to be a viable and sustainable option. Before investing in renewables in these areas, a feasibility study is of paramount importance starting with assessing and determining the amount of available solar irradiance and wind speeds for the area. In addition, a techno-economic feasibility study is of paramount importance to determine the most economical and sustainable standalone hybrid system. This research presents a study using a nonparametric kernel density estimation method to determine solar irradiance and wind speeds. In addition to this kernel determination method, the study performs a feasibility analysis using a hybrid renewable energy system that consists of two renewables with biodiesel and battery backup to supply the energy demands of a rural household in South Africa. The research commences with a literature review of several probability distribution functions (pdfs) commonly used in testing both solar irradiance and wind speeds. It established that not all sites can be defined by the same pdf and there is no science in selecting a distribution function but rather random testing of a range of functions. The parametric probability functions tested in this work are Gamma, Weibull, and Lognormal. The work then compares the performance of these parametric pdfs with the nonparametric kernel density estimation method which this study advocates for its application. In judging the performance and correctness of these pdfs, mean bias error (mbe) and root mean square error (rmse) are used as performance test criteria for the parametric probability distribution function. As for the nonparametric pdf which this research advocates for its use, an integral squared error, ISE is used for the presentation assessment with the conventional parametric normal distribution. From the results, it is observed with the proposed nonparametric kernel density estimator gives precise estimation and improved adaptableness, as opposed to the widely used conventional parametric distribution for both the use in solar irradiation and wind, speeds estimations. In addition, the research results demonstrated that the commonly used Epanechnikov and Gaussian KDE methods were the most adjustable methods for all seven tested stations. The second aspect of the study applies the tested data to design and perform a feasibility study of using a hybrid renewable energy system that consists of two renewables with biodiesel and battery backup to supply energy demands for a typical rural household. Thus, the study makes use of a simulation to design and determine an optimized hybrid renewable energy system for application in rural households. The energy resources considered for this standalone hybrid system are solar PV, wind, diesel generator, and a storage battery system. In performing the system simulation and optimization concerning economic viability, sustainability, energy efficiency, and environmental impact is carried out using the Hybrid Optimization Model for Electric Renewables (HOMER) simulation and optimization software tool. Concerning the results obtained, HOMER gave seven best-optimized systems. In breaking down the seven optimized results, four of the results were hybrid energy systems and three with only one energy resource. Moreover, from these results, three systems were pure green energy supplied and not utilizing any diesel generator (DG). The best-optimized system for this rural household consisted of PV/DG with an NPC of $ 72,720, while the system which utilized all resources available was second-ranked with an NPC of $ 79,272. The use of only renewable resources for this region was fourth-ranked with NPC of $ 86,760. The study demonstrates the feasibility and viability of having rural areas benefit from electricity access. Moreover, this study will contribute towards the strides of just energy transition envisaged by the country in solving the energy crisis currently being experienced.
Modeling and performance analysis of artificial intelligence (AI) based controllers for AVR of a synchronous generator (SG)
(2023-05) Mazibuko, Ntombenhle; Akindeji, Kayode Timothy; Sharma, Gulshan
An automatic voltage regulator (AVR) is an electronic device used to control, adjust, and maintain a constant voltage level at the stator terminals of a synchronous generator (SG). Hence, the voltage stability of a power system network is affected by AVR’s performance. Maintaining constancy and stability of the nominal voltage level in power systems remains a major control problem. Another critical reason for effective control of the generator's terminal voltage is that real line losses are determined by the real and reactive power flows and variation in terminal voltage has a large effect on reactive power flow and thus on these losses. A large power system consists of several synchronous generators that operate in synchronism; the terminal voltage and frequency are to be kept constant with minimal variation to ensure the stability of the power system. The voltage stability of a synchronous generator is highly affected when the terminal voltage varies above the nominal acceptable range. To maintain a constant voltage at a SG’s terminal, an AVR is used. The performance of an AVR is highly dependent on efficient controller design, which improves the output of the AVR by restoring the voltage of the synchronous generator to its nominal value in the presence of disturbances. The selection of a suitable controller is one of the most challenging aspects of AVR system design. This study presents the design, modeling, and performance analysis of an AVR system employing a Proportional Integral and Derivative (PID) controller, a Fuzzy Logic controller (FLC), and a Model Predictive Controller (MPC) for the performance enhancement and transient response of the AVR system with these controllers. Initially, a transfer function is used to develop a mathematical model of an AVR in order to observe its step response when the terminal voltage of a generator is disturbed. A PID controller is then added to the system and tuned to enhance the step response of an AVR. The third model develops and implements an AVR system based on MPC, while the final model implements an FLC for an AVR system. Simulating the models in Matlab Simulink 2021a, the results have demonstrated the need for a controlling mechanism to enhance the dynamic performance of the AVRS, and MPC has shown to be the most effective controller.