Power quality (PQ) monitoring has been an important issue in smart grid (SG). PQ study involves the detection of the disturbances/events in different power system signals of the SG communication. Detecting disturbances in power system signals itself is a very difficult task and it becomes even more difficult when there are noises in the signal. The detection efficiency of the disturbances is highly downgraded by the noise, which corrupts the detection capability of the system, and hinders the localization and classification of the disturbances. In this work, a PQ monitoring system based on wavelet packet decomposition (WPD) is proposed to detect five different types of disturbances in the presence of noise. The proposed system utilizes the interscale and intrascale relationship of the WPD coefficients for the detection of the PQ disturbances upon noise removal. The correlation among the WPD coefficients exits both within its own scale and across its adjoining scales. For the evaluation, the presented method is applied both on the simulated and real SG disturbance signals obtained from frequency disturbance recorder (FDR), phasor measurement unit (PMU), and power system load. Comparative results are presented with wavelet decomposition.