A De-noising Scheme for Wavelet Based Power Quality Disturbances Detection and Classification system

A Power quality Classification system can easily extract features from the second detail signal obtained after Discrete
Wavelet Transform and using these features to construct a Rule Based Algorithm for identifying types of disturbances that exist in
the captured power signal. Unfortunately, the signal under investigation is often polluted by noises, rendering the extraction of
features a difficult task, especially if the noises have high frequency spectrum which overlaps with the frequency of the
disturbances. The performance of the Classification system would be greatly degraded, due to the difficulty in distinguishing the
noises and the disturbances. To overcome this difficulty and to improve the performance of the system, this paper proposes a
suitable de-noising scheme to be integrated into the system so that the classification system is still workable in a noisy
environment. In the proposed de-nosing scheme, a noise shrinkage threshold used to minimize or eliminate the noise coefficient in
the second compressed detail obtained after discrete wavelet transform is determined adoptively according to the background
noises of the signal. The ability of the Classification system can then be restored. To test the effectiveness of the de-noising
scheme, the system is tested with noise-added disturbance signals generated by MATLAB programming language and some field
data obtained from the PQDIF resource centre. Using the de-noising scheme proposed in this paper, a higher tolerance to noise
can be achieved by the Power Quality Problem Classification system.