Classification of Power Signal Disturbances Using Wavelet Based Neural Network

The power signal disturbances are classified as impulse, notches, glitches, momentary interruption, voltage sag/swell, harmonic distortion and flicker. These disturbances may cause malfunctioning of the equipments. To improve the quality of the power supply detection of the disturbance must be done accurately. In this paper DWT is employed to capture the time of transient occurrence and extract frequency features of power disturbances. These DWT coefficients when applied as inputs to the neural networks require large memory space and much learning time. Hence along with the multi resolution analysis (MRA) technique the statistical methods are used to extract the disturbance features of the distorted signal at different resolution levels. For neural network structure probabilistic neural network (PNN) and feed forward back propagation network (FFBPN) are used to classify the disturbance type and are compared. The learning efficiency of PNN is very fast when compared to FFBPN, and it is suitable for signal classification problems. Distorted signals were generated by the power system block set in MATLAB. The accuracy rate is improved using wavelets along with the statistical differentiation of the various power signal disturbances.