Power transformer disturbance classification based on the wavelet transform and artificial neural networks

Abstract

Currently, the differential function has been widely used in transformer protection relay. However, the main issue of this technique is assigned to the relay misoperation during the presence of inrush currents and current transformer (CT) saturation. In the literature, these limitations have been overcome with the use of tools based on artificial intelligence and signal processing, such as the methods based on artificial neural networks and wavelet transform. This paper proposes a method based the ANNs and wavelet transform to detect and classify disturbance in the power transformer accurately. The algorithm uses wavelet-based disturbance detector in order to detect any disturbance related to a power transformer, whereas a neural network-based routine is used to classify the disturbance type (internal fault, external fault and transformer energization) appropriately, as well as to classify the internal faults. Several events were simulated, such as external and internal faults, with variations of fault resistance, fault inception angle, and fault type parameters, as well as transformer energizations. The method presented an excellent success rate regarding the correct classification of the disturbance as well as an accurate fault classification.