Improving Fault Classification in Differential Protection Schemes with Machine Learning Based Methods

The use of machine learning techniques in classifying between external faults and internal faults in differential protection schemes is proposed. The SVM Gaussian classifier is found to be suitable for such applications. Model training and testing results show that it is possible to develop methodologies that can be applied universally in similar system setups with different degrees of CT saturation while minimizing modeling and training efforts. By applying the proposed methodology on a real-world transformer differential mis-operation event, the clear advantages and potentials of machine learning techniques over conventional slope-based fault classification methods are demonstrated.