ADENUGBA, FAVOUR TOLUWANIMI (2020) ISLANDING DETECTION FOR GRID-CONNECTED DISTRIBUTED GENERATION SYSTEMS USING CONVOLUTIONAL NEURAL NETWORK. Masters thesis, Covenant University Ota..
![[img]](http://eprints.covenantuniversity.edu.ng/style/images/fileicons/application_pdf.png) |
PDF
Download (965kB) |
Abstract
In the world today the lack of adequate supply of electricity is still a major problem especially in developing and underdeveloped countries. The global electrification rate is put at 75% and this figure has to go up in the coming years in other to promote sustainable development and eliminate world poverty. Distributed generation (DG) integration with the grid has been increasing worldwide due to the advantages it can provide to the electrical power systems, such as the possibility of reducing transmission and distribution losses, environmental benefits, the increase in the reliability of the power supply and the deferral of transmission and distribution investments. This makes it a suitable mechanism to improve electrification rate all over the world. Unintentional islanding is a major technical challenge that bedevils this system. Some researchers have developed islanding detection models to detect islanding and regard all other events that take place while the grid is still connected as Non-Islanding events while others have developed islanding detection models to detect islanding as well as identify Nonislanding disturbances when they occur (Islanding and Non-Islanding disturbance detection). Both system types are developed in this research. This research presents image-based islanding detection models using convolutional neural network. These models utilize scalogram images obtained from the aggregated phase voltages at point of common coupling (PCC). Therefore the models utilize the PCC voltage as the islanding detection parameter. The power system, islanding, and non-islanding events are simulated in MATLAB, wavelet transform is applied to the voltage signals obtained from the PCC for the different events to obtain the scalogram representation of the event. In both models developed a portion of this image data generated is used to train the classifier while the other part is used to test the classifier. The immunity of the developed models to noise is also investigated, the noise introduced did not have an adverse effect on the models. The results obtained from the simulation proves the ability of the proposed classifiers to detect islanding. The proposed models compare favourably with existing techniques and methods. For the first model, detection accuracy of 99.83% was obtained while for the second system detection accuracy of 99.2% was obtained.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Convolutional Neural Network (CNN), Distributed Generation, Scalogram, Unintentional Islanding, Islanding Detection, Non-Islanding Disturbance. |
| Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering |
| Divisions: | Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science |
| Depositing User: | Mrs Patricia Nwokealisi |
| Date Deposited: | 17 Aug 2021 14:30 |
| Last Modified: | 17 Aug 2021 14:30 |
| URI: | http://eprints.covenantuniversity.edu.ng/id/eprint/15299 |
Actions (login required)
 |
View Item |
