A new classification model based on SVM for single and combined power quality disturbances

The simultaneous occurrence of power quality (PQ) disturbances is increased in recent times, and the detection of combined disturbances has become a pressing concern. A new model based on support vector machine (SVM) for classification of single and combined PQ disturbances is proposed in this paper. The classification of k disturbances with any of the conventional multiclass SVM approaches demands the utilization of at least k binary SVMs. This increase in the number of SVMs with an increase in classes will affect not only the classification time but also the recognition accuracy. The proposed classification model overcomes this limitation by employing a number of binary SVMs significantly less than the number of classes to be classified. The classification of sixteen disturbances is considered in this paper by utilizing only nine binary SVMs, which facilitates better detection of the combined disturbances with fewer computations. To validate the performance of the proposed SVM model, it has been tested on a wide variety of synthetic signals and a few real signals. Further, the results obtained are compared with the one-against-one approach based multiclass SVM technique.