Detection and classification of power quality disturbances in time domain using probabilistic neural network

This paper proposes a new approach for detection and classification of power quality (PQ) disturbances in time domain. Most research in this field employs frequency domain analysis tools to analyse the features of PQ disturbances, such as Fourier transform and wavelet transform. However, the transient and steady-state characteristics of PQ disturbances are originally reflected on the waveforms of PQ disturbances, i.e., in time domain. In order to detect and classify the PQ disturbances in time domain, mathematical morphology (MM) and Teager energy operator (TEO), which are excellent analysis tools in time domain, are used for feature extraction in this paper. The features compose a feature vector. After that, a probabilistic neural network (PNN), which is more effective as a classifier than other neural network, is used to classify PQ disturbance signals. The feature vector composed of features extracted by MM and TEO is considered as the input of PNN. The proposed approach is tested by PQ disturbance signals, which are simulated according to the IEEE 1159-2009 standard, including swell, sag, interruption, harmonics, notching, oscillatory, fluctuation, and several combinations of these disturbances. The results demonstrate that the features extracted by MM and TEO are effective and the PNN classifies PQ disturbances with high accuracy rate.