Automatic power quality disturbances detection and classification based on discrete wavelet transform and artificial intelligence

In this paper some patterns based on discrete Wavelet transform are studied for detection and identification of both, low frequency disturbances, like flicker and harmonics, and high frequency disturbances, like transient and sags. The Wavelet function Daubichies4 is used as base function in detection and identification because of its frequency response and information time localization properties. Based on these patterns, power quality disturbances are automatically classified by using several artificial intelligent techniques: back propagation neural network (multilayer perceptron), kohonen neural network (self organizing map), Bayesian (linear statistical method) and support vector machines (SVM). Neural networks and SVM exhibit the best performance as classifiers (90 percent of success for the most disturbances) in spite of similitude between some disturbance patterns. The whole strategy was integrated on a MatLab(R) Graphical User Interface and tested by using synthetic signals (according to international standards) which were collected in a disturbance database.