Advanced detection and discrimination of power transformer internal faults from other abnormal condition using DWT-based feature extraction and ANN classification

Abstract

Power transformers are one of the most critical elements of the electrical Power System because of the aforementioned function in the voltage regulation and power supply. It is very important in the field of power engineering to be able to differentiate the inrush currents caused due to the energization of the transformer from the internal fault currents created in the transformer. This paper represents an efficient approach to solving this issue by employing DWT feature extraction and ANN classification. This approach is based on the determination of waveforms by distinguishing the D4 and D5 coefficients of instantaneous differential currents using DWT. These coefficients present much useful information related to the waveform type, making it possible to differentiate between the inrush and the internal fault currents. This is a key factor when making classification in that these criteria are related to the energy content involved within these coefficients. This energetic approach forms the basis for the ANN controller to determine particular decisions about the quality of the current. This proposed approach is supported with simulation to represent empirical data in supporting the use of this approach. The results always confirm the efficiency of such an approach to the differentiation between inrush and internal fault currents with a high percentage of accuracy. The effectiveness of this method goes beyond accuracy as it is reliable, responds quickly to abnormal conditions, and can be applied to a variety of power transformer types. Applying this concept in real grid power systems can lead to increased reliability and less downtime thereby strengthening the electrical system as a whole. The reliability and safety of power transformers remain a critical concern in power engineering. The present paper proposes a new method to improve power transformer protection for differentiating internal faults from other abnormal situations. The method described herein utilizes advanced signal processing and machine learning with the help of DWT and ANN to reach higher standards of accuracy and reliability.