Identification of generator loss-of-excitation from power-swing conditions using a fast pattern classification method

This study describes a support vector machine (SVM)-based technique for identifying loss-of-excitation (LOE) condition in synchronous generators from other disturbances such as external faults and power-swing conditions. In this new approach, only one zone of LOE is required and the time coordination is reduced significantly. The proposed method is compared with traditional two-zone impedance method. Several operating conditions within the generator capability are used to verify the generality of the SVM-based classifier. The proposed classifier identifies an LOE condition in all cases before the impedance enters the larger mho impedance zone. Faults and power-swing conditions are identified correctly, thereby preventing incorrect operation of the LOE impedance zone.