Fast TT transform and optimized probabilistic neural network-based power quality event detection and classification

This paper accomplishes the detection and classification of power quality disturbances (PQDs) using a fast time–time (TT) analysis and differential evolution (DE)-based probabilistic neural network (PNN). Applying a translatable and scalable Gaussian window, the TT transform divides a primary time series to a secondary set of time localized time series. In this secondary time series representation, the higher frequencies are highly concentrated in the midpoint of the Gaussian, in comparison with lower frequencies. In this study, fast TT transform is considered to accommodate arbitrarily scalable windows. With generalized S transform, the TT transform is extended to resolve the times of PQ event initiations. Further, to be computationally less complex and faster, the considered TT transform for feature extraction is accommodated with automatic scaling. The extracted features are used as input to the PNN classifier for the classification of the PQDs. Further, a modified mutation-based DE is used to enhance the PNN performance by optimizing the weights and optimum setting of the spread constant value. The obtained simulation results prove the better performance of the proposed approach with significantly less complexity and higher classification accuracy to detect and classify the power quality events even with noise contamination.