DETECTION AND CLASSIFICATION FAULT IN POWER TRANSMISSION LINE BASED ON 1D CONVOLUTION NEURAL NETWORK

• Main Author: ASH SHIDDIEQY - NIM: 23216066, HASBI
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/27766
• Institution: Institut Teknologi Bandung

<p align="justify">Protection system on power system need to improve reliability. Early detection can be used to prevent failure in power transmission line. And fault classification system is required to protect from error detection and improve protection system to perform analysis and decision making. <br /> <br /> <br /> Each transmission line signal has a continuous pattern. When fault occurs this pattern changes. This fault pattern can be detected and classified by conventional machine learning model using wavelet feature extraction and classify type of fault using Artificial Neural Network (ANN). Hence a machine learning is proposed ini this thesis based on Convolutional Neural Network which more suitablke for pattern recognition. <br /> <br /> <br /> In this thesis the method begin with making power simulation model in Simulink and matlab which generate dataset fault divided into 45.738 data training dan 4.752 data test. Then proposed design klasifier is made. Each proposed model classifier is training by feed in with the same dataset. Training result best optimal model is ConvNet with raw input/without prerprocessing. <br /> <br /> <br /> Optimal model then implemented on Single Board Computer Raspberry Pi 3. In Implementation and validation test using hardware in loop obtained ConvNet model capable to do a realtime classification with average response time 4.34 ms for model classifier and 6.2 ms for full system.<p align="justify"> <br /> <br />