The Sdudy On Recognition Algorithms Of Mixed Power Quality Disturbances

The studies on the various factors affecting power quality (PQ), timely detection of leading to the decline of PQ phenomena and effective identification of PQ problems are necessary and meaningful to achieve efficient management and control of PQ. The recognition of PQ disturbances is a very complex scientific issue, since actual power quality disturbances are various, and the signal differences are not very clear with disturbances mixed phenomenon. For recognition problems of mixed PQ disturbance signal, the three aspects, the signal characteristics, the detection algorithm, and the recognition algorithm are analyzed and discussed in the paper.In the paper, the related characteristics of mixed PQ disturbance signal are analyzed, and the mathematical expressions, physical characteristics and classification methods of the single disturbances are introduced., Subsquently, the signals of time-domain and analyzed their spectral are contrasted, the interactions of a single disturbance are explored after being compounded Finally, the recognition problems of mixed PQ disturbance signals were analyzed qualitatively, and the evaluation of identification methods is discussed.In order to find an effective feature extraction method, the paper researches the disturbances detection algorithm, and presents the PQ disturbances detection algorithm based on EEMD and wavelet package Tsallis singular entropy respectively. The HHT detection algorithm based on EEMD can effectively avoid the mode mixing problem. The disturbed signals of intrinsic mode functions (IMF) component decomposed by the EEMD method not only have the physical meaning of the IMFs clearly, but reduce the noise interference. The start-stop moment, amplitude and frequency of the power quality disturbances can be detected according to the transient characteristics. The wavelet package tsallis singular entropy is the combination of wavelet packet technology with non-extensive entropy theory, which can compute singular Tsallis entropy of the coefficient matrix through the wavelet packet decomposition, and detect the signal point-in-time of the signal frequency mutations. Simulation results show that the wavelet packet Tsallis singular entropy can effectively detect signal mutation, have better noise immunity, less affected by the disturbance amplitude changes, and can effectively locate the starting and ending times for a variety of power quality signals. In order to solve the problem of mixed PQ disturbance recognition effectively, this paper proposed two research schemes:①Signal processing method combination with the automatic identification system;②Multi-label classification method.For the first scheme, the paper used modern signal processing methods such as the ensemble empirical mode decomposition and modified incomplete S-transform, created nine time-frequency domain characteristics, and presented a new method for of the mixed PQ disturbance classification based on time-frequency domain multiple features for the usual disturbance signals. Then, the characteristics were input into the sub-block of the automatic classification system for the disturbance identification. This method has fully considered the interferences between the single disturbances, and effectively suppressed them with the complementary time-frequency domain characteristic features. The simulation results show that the method can effectively recognize the power quality mixed disturbances including voltage sag, voltage swell, voltage interruption, impulsive transient, oscillation transient, harmonics, flicker and their mixed ones under noise conditions effectively.For the second scheme, the multi-label classification idea is introduced into the mixed PQ disturbances recognition in this paper. Based on the conventional Rank-SVM, the paper presented a multi-label ranking learning method of Rank-Wavelet Support Vector Machine using Wavelet Support Vector Machine instead of conventional Rank-SVM, and used the method for the mixed power quality disturbances identification. The experimental results show that the method can improve the classification performance of the Rank-SVM effectively, and prove its superiority and effectiveness by the comparisons between the different noise backgrounds and other multi-label methods.