A Support Vector Machine-Based Power Transformer Differential Protection
support vector machine, power transformer, differential protection, CT saturation, wavelet transform.
Power transformers play a major role to power systems operation, interconnecting
generation, transmission, and distribution systems. Typically, the phasor-based differential
protection is used as primary protection to this equipment. However, conventional
power transformer differential protection can still face issues dealing with some type of
events, such as faults during current transformer (CT) saturation, evolving faults, and
inrush currents during an internal fault. Over the last years, many works based on digital
signal processing techniques and machine learning algorithms have been proposed
in the literature to deal with these problems. In this research, a comprehensive support
vector machine (SVM)-based power transformer differential protection is described. Using
as inputs high- and low-frequency information, extracted with the real-time boundary
stationary wavelet transform (RT-BSWT) from the current signals, three different SVM
models were derived with simulation data comprising various events, including internal
faults in the transformer windings, overexcitation conditions, external faults during CT
saturation, evolving external-to-internal faults, and faulty transformer energizations. By
adopting a trigger-based approach, the proposed protection scheme can accurately distinguish
simultaneous events in a power transformer. To validate the proposed protection
scheme, an extensive comparative analysis with the conventional method was carried out,
and the proposed method has outperformed the conventional one on most of the evaluated
metrics, such as accuracy, dependability, safety, as well as relay operating time. In
addition, the proposed SVM-based method was designed to be embedded in a dedicated
hardware and run in real-time, thus being a promising solution to enhance power transformer
differential protection.