Classification of power-quality disturbances in noisy conditions

An approach to classify various types of power-quality disturbances in noisy conditions is proposed. The approach is based on the S-transform and support vector machines (SVM), a novel statistical learning method that provides efficient and powerful classification algorithms capable of dealing with high-dimensional input features and with theoretical bounds on the generalisation error and sparseness of the solution provided by statistical learning theory. An SVM-based classification scheme is proposed for power-quality disturbances. It is a two-stage system in which the S-transform is applied to obtain useful features of the nonstationary power-quality disturbance signals in the first stage. The S-transform can be seen either as a phase-corrected version of the wavelet transform or a variable-window short-time Fourier transform that simultaneously localises both real and imaginary spectra of the signal. The features obtained from S-transform analysis of the power-quality disturbance signals are much more amenable for pattern recognition, unlike the currently available wavelet transform techniques. In stage two a SVM classification tree is constructed to classify the various disturbance waveforms generated due to power-quality violations. The approach is easy to implement and classification accuracy is high even in noisy conditions.