Hybrid Methods for Fast Detection and Characterization of Power Quality Disturbances
- PDF / 1,452,628 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 75 Downloads / 194 Views
Hybrid Methods for Fast Detection and Characterization of Power Quality Disturbances S. Upadhyaya1 · S. Mohanty1 · C. N. Bhende2
Received: 10 December 2014 / Revised: 11 June 2015 / Accepted: 21 July 2015 © Brazilian Society for Automatics–SBA 2015
Abstract In this paper, recently developed variants of wavelet transform, namely the maximum overlapping discrete wavelet transform and the second-generation wavelet transform, are used for detection of ten types of the power quality (PQ) disturbance signals. Further, the features of PQ signal disturbances are extracted using these wavelet transforms. Those extracted features are then used to classify various PQ disturbances. Random forest (RF) classifier is presented in this paper. The RF is constructed with multiple trees for classification of large number of classes simultaneously. In order to represent realistic situation, the proposed technique is tested with noisy data. Keywords Power quality disturbances (PQD ) · Maximal overlap discrete wavelet transform (MODWT ) · Secondgeneration wavelet transform (SGWT ) · Random forest (RF ) · Power quality (PQ ) · Classification accuracy (CA )
1 Introduction The power quality is an important factor to be considered in the power system which attributes toward the performance of the system with disturbances generated in the power signals.
B
S. Upadhyaya [email protected]; [email protected] S. Mohanty [email protected] C. N. Bhende [email protected]
1
Department of EE, NIT Rourkela, Rourkela 769008, India
2
School of Electrical Science, IIT Bhubaneswar, Bhubaneswar, India
In order to improve the quality of power, different power quality (PQ) disturbances are need to identified first and remedies are taken to mitigate them. Therefore, detection and the classification of PQ signals are the important aspects in the power quality analysis. In order to identify the disturbances, the different techniques such as the Fourier transform (FT), the short-time Fourier transform (STFT), wavelet transform (WT), neural network, fuzzy logic and S-transform have been used (Gaouda et al. 1999; Angrisani et al. 1998; Douglas 1993; Gu and Bollen 2000). The FT is a fast technique which gives the information about the frequency component but does not give any information regarding the time of occurrence and the duration of the disturbance lasted for. So, FT is not suitable technique for the analysis of non-stationary signal. On the other hand, the time frequency information related to the disturbance waveform can be obtained in STFT (Gabor 1946). However, STFT is not suitable to track the transient signals perfectly due to its fixed window property (Biswal et al. 2014). Similarly, the S-transform suffers from computational burden which limits its applications (Brown and Frayne 2008). The wavelet transform affords the timescale analysis of the non-stationary signal due to multi-resolution analysis (MRA) property. The property of MRA of WT represents the signals into different timescales rather than the time frequency like
Data Loading...
