The Sdudy On Detection And Recognition Algorithms Of Mixed Power Quality Disturbances

With the increasing of smart grid and power automation, the electric power plays a more and more important role to the daily life and production. High quality power supply can improve the utilization of power and guarantee the reliability. Power quality detection and identification can provide the basis for fault diagnosis and evaluation. There is variety of power quality (PQ) disturbances in real power system, and the detection and recognition algorithms are lack of uniform standards. Based on the study of PQ disturbance characteristics and existed detection and identification algorithms, some new algorithm for single and complex PQ parameter detection and classification method based on spectral kurtosis (SK) and multi-label classifier are proposed in this paper.The existed methods of frequency detection of transient oscillation signal in power system are complex, susceptible to noise and poor in versatility. To overcome these disadvantages a new method based on Morlet Wavelet-based-Spectral Kurtosis (SK) is proposed in this paper. In the proposed method in the paper, the Morlet wavelet transform is adopted to obtain the wavelet coefficients firstly. Then the wavelet-based-SK is calculated according to the definition of SK. The frequency of transient oscillation is detected by the maximum of the SK at last. The similarity function is proposed and defined to analyze the influence of noise for the detection results in the paper. The simulation results show that this method is applicable and effective for the most of common oscillation signals. In addition, the calculation process is simple, and the accuracy can meet the requirement of actual system.The computation of SK is based on the time-frequency analysis. The different time-frequency analysis method has different characteristics, so the computation methods of SK decide the characteristics of transient disturbances. The characteristics of different SK methods are discussed through theoretical and experimental analysis in this paper. And a new computation method based on Choi-Williams distribution (CWD) is proposed in the paper. The comparisons and analysis of different computation methods for five transient disturbances show that the proposed method can better reflect the difference of transient PQ disturbances. In order to ensure the validity and accuracy of the recognition algorithm, the relationships between SK and transient disturbance parameters (amplitude, phase, duration, noise and so on) are analyzed and discussed in detail, and a recognition plan is proposed at last. The simulation signals with random disturbance parameters and different noise, RTDS and real-life signals are adopted to prove the recognition performance. The experimental results show that the method is feasible and simple, and the recognition rate of five transient disturbances is high and satisfied. In addition, the disturbances are classified with thresholds, which avoid the large amount of calculation caused by complex classifier.Due to the complexity of the actual power system, the actual PQ disturbances are likely to be complex disturbances composed by different single ones. Therefore, it is necessary to study complex disturbances classification method. The existed feature extraction methods of complex disturbance are complex and difficult, and the dimension of the characteristic quantity is high. In order to solve these problems, a new feature extraction method based on SK and root mean square (RMS) is proposed in this paper. The new method is able to reduce the dimension of feature quantity and improve the efficiency of the classifier without affecting the classification accuracy. Finally, multi-label ranking wavelet support vector machine is used to classify complex disturbances feature quantities. The simulation and comparison results show that the algorithm is effective.