Fire-Fly Algorithm Based ANN Classification of Power Transformer Faults
P. Lakshmi Supriya1, P.Ram Kishore Kumar Reddy2
1P.Lakshmi Supriya*, Assistant Professor, Department of EEE, Mahatma Gandhi Institute of Technology, Gandipet, Hyderabad, India.
2P.Ram Kishore Kumar Reddy, Professor, Department of EEE, Mahatma Gandhi Institute of Technology, Gandipet, Hyderabad, India.
Manuscript received on March 15, 2020. | Revised Manuscript received on March 30, 2020. | Manuscript published on April 10, 2020. | PP: 1880-1884 | Volume-9 Issue-6, April 2020. | Retrieval Number: F4589049620/2020©BEIESP | DOI: 10.35940/ijitee.F4589.049620
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Monitoring and estimating the states of the transformer during faulted phase condition is essential to continuity of supply. Varied techniques are proposed for faulted phase detection to improve condition assessment. In this paper, we propose a novel method to detect and classify power transformer faults using wavelet transform Multi Resolution Analysis (MRA) as feature extracted parameter vector and Fire-Fly Algorithm (FFA) based Artificial Neural network training as classification method. The observed Dissolved Gas Analysis (DGA) waveform data is analyzed with wavelet transforms (WT) to identify abnormalities which is supported by MRA. In MRA, the current, voltage and temperature of winding and oil are decomposed into high and low frequency components. The magnitude of components, signifies the feature vector, gives a detection criteria. After detecting feature vector, dominant coefficients of WT can be used to train the ANN with FFA based learning algorithm. Different types of faults are created on transformer such as Single Line-Ground (SLG), Line-Line (LL), Double Line-Ground LLG, Three phase fault (LLLG) for the analysis using WT and ANN. The detection and classification of the fault signal are executed and examined in different winding location and different fault conditions. Finally, the presented precise model recognizes the faults based on performance metrics with high classification accuracy for various classes.
Keywords: Artificial Neural Networks, Dissolved Gas Analysis, Fire-Fly Algorithm, Multi Resolution Analysis.
Scope of the Article: Low-Power Design