Half Multi-Resolution Morphological Gradient and Probabilistic Neural Network for Identification and Classification of Disturbances

Abstract

Power quality disturbances occur in the power system. It becomes a good challenge for researches to get better results in the detection and classification of these problem. A new combination method in identification and classification of power quality disturbances is presented in this paper using Mathematical Morphology and Probabilistic Neural Networks (PNN). In this strategy, Half Multi-resolution Morphology gradient (HMMG) is used to process the filtered signal with the sampling frequency at 6.4kHz or 128 samples per cycle. This HMMG is applied to detect the time location of disturbances. and followed by calculating the mean, maximum, minimum, standard deviation, and total harmonic distortion (THD) of the signal. These results are used as inputs of the PNN. There were 5 inputs and 20 hidden neurons are used on this PNN. Using this strategy, the nine types of disturbances classification can be obtained. Matlab is used for simulation of this proposed method and found that PNN with HMMG resulting in the percentage correct of classification at about 99.167%.