A Single Perceptron Smart Sensor Technique for Pre-fault Monitoring System in an Indoor Substation

• Main Author: Lorothy, Morrison Buah Singkang
• Format: Thesis
• Language: English; English; English
• Published: University Malaysia Sarawak 2024
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://ir.unimas.my/id/eprint/45389/3/DOW_Lorothy%20anak%20Morrison%20Buah.pdf; http://ir.unimas.my/id/eprint/45389/4/Thesis%20PhD_Lorothy%20Morrison%20Buah%20-%2024%20pages.pdf; http://ir.unimas.my/id/eprint/45389/5/Thesis%20PhD_Lorothy%20Morrison%20Buah.ftext.pdf; http://ir.unimas.my/id/eprint/45389/

Single Perceptron Smart Sensor (SPSS) is a new Ultra-High Frequency (UHF) sensor developed to significantly improve the pre-fault monitoring system for early detection, localization, and identification of the Corona and Arcs Electric Discharges (EDs) in an indoor substation. Corona and Arcs ED constitute a significant threat to electrical safety, the apparatus, and the stability of a power system due to the aging and material degradation in the power apparatus. Hence, an early preventive approach must be performed effectively for pre-fault threat detection. In this research, a novel pre-fault monitoring system utilizing SPSS is developed, embedding a novel Signal Identifier Technique for the Corona and Arcs ED detection, localization, and identification. The SPSS formation integrates a 2-element Linear Array Antenna with a Single Perceptron-Artificial Neural Network (SP-ANN). It detects and localizes the Corona and Arc ED signals based on the Direction of Arrival (DOA) angle. The SP-ANN utilizes a single-layer neuron with less complexity, speedy detection, and localization within seconds. The waveform-based signal feature extraction uses the Signal Identifier Technique for signal identification. Since the frequency range of the Corona and Arcs is undecidable, the accuracy of the pre-fault monitoring is tested for the Corona and Arcs ED at a sampling frequency of 300 MHz to 3 GHz. The SPSS has revealed an accuracy of 99.86% for signal identification with minimal computational complexity, thus giving another practical wireless technique for UHF signal interpretation.