Proteção diferencial de transformadores de potência baseada em máquinas de vetor de suporte e transformada wavelet

Abstract

Power transformer is one of the main equipment of the electrical system which is responsible for the interconnection of electrical networks at different voltage levels. The protection of this equipment must work quickly and efficiently in cases of internal fault in the transformer, because depending on the severity of the fault can lead to the loss of the transformer. The differential function has been widely used in protection for power transformer, because it is a quick and selective function. However, this protection functionmay work unduly in cases of transformers energizing, as well as external faults with distortions in the current waveform due the saturation of the current transformers. Another problem it is no sensitivity for cases of internal fault in the transformer windings involving few turns. Therefore, this work proposes an alternative to improve the differential protection of transformers using the combination of Machine Learning Algorithms and the wavelet transform. The proposed method recreates the differential function using a disturbance detector, by means of the energies of the wavelet coefficients, which enables the phase and negative sequence neuro-wavelet differential functions, which are based on support vector machines. The proposed method also has an internal fault classifier, also based on support vector machines, to perform fault type classification and assist in the relay trip logic. The proposed method send trip signal in 100% of the evaluated internal fault cases, providing no trip to energization of transformers and external faults. The proposed differential protection scheme can work in conjunction with other system protections by sending warning signals, for example in external fault conditions of the transformer, so it is doing the protection more reliable and intelligent.