Data Mining Based Power Quality Disturbance Detection Using Wavelet Transform

This paper presents the identification and classification of power quality disturbances using wavelet transform and data mining models. The various power quality events like oscillatory transient, voltage swell and voltage sag are simulated for a small distribution network using MATLAB/SIMULINK. The simulated events are then processed by the wavelet transform with Daubechies Db5 mother wavelet and wavelet coefficients are generated. The energy, standard deviation, mean, skewness, kurtosis and variance are obtained for these wavelet coefficients for the purpose of distinguishing the events. A data mining based Random Forest model and Decision Tree model are then generated using these features of voltage signal and these models are used for final classification into sag, swell and transients. The test result shows that the wavelet algorithm along with decision tree is more sharp in identifying events.