Recognition Methods Of Multiple Power Quality Disturbances

With the rapid development of science and technology, world energy crisis become more serious, scientific concept of sustainable development welcomed by more and more people. States are competing research and development of new and clean energy efficient. Power grid, load structure and power characteristics are undergoing tremendous changes. Various new energy, large-scale power electronic devices and a series of non-linear unbalanced load put into use caused a serious challenge on high quality of electricity and causing a series of power quality problems. But the development of technology makes equipments more stringent to power quality requirements, therefore, power quality problems has been widespread concern. To governance and improve the power quality is urgent, accurately monitor and scientific analysis power quality is the premise of this task. Monitoring and analysis is the key to extract the parameters of a variety of power quality disturbance signal accurately and identify the types. This paper presents a method of single power quality disturbance recognition and composite power quality disturbances identification.In single power quality disturbances recognition using wavelet transform to analyze of power quality disturbances, using Principal component analysis method to reduce the dimension of feature vectors. Then use the support vector machine to classify the disturbance signal. In the process of power quality disturbances identification, decomposition level is determined based on experience, little decomposition level lead to incomprehensive features characterize and poor identification result, more types to be recognized more prominent the phenomenon is. Although overmuch decomposition level can get more comprehensive feature information, but not every dimension can provide useful information that the information extracted by wavelet decomposition is redundant, these redundant scales will result in classifier training and recognition process takes too long, is not conducive to meet real-time requirements. To solve this problem, firstly using wavelet decompose power quality disturbance signal into10layers, using a standard voltage signal as a reference signal to extract the energy difference disturbance signal wavelet features. And then using principal component analysis method allows the energy difference of wavelet feature vector dimension reduction, Finally, the reduced feature vector as support vector machines to classify the input signal disturbance. The experimental results show that this method can recognition eight kinds of single power quality disturbances, recognition accuracy rate can reach98.81%. Ensuring the recognition rate target and reducing the time of the recognition process greatly, and own excellent anti-noise performance. For comparison with the existing literature, the method shows effectively.In the study of complex power quality disturbance signal recognition method, using a variety of signal analysis methods to comprehensive analysis complex power quality disturbance according to the characteristics of composite power quality disturbance signals, presents a complex power quality disturbances classification method. Firstly, signal is analyzed by S transform to extract five features from the baseband signal amplitude, frequency, high frequency, using dynamic measurement to posit the frequency extreme point of power quality disturbance signal and extracting four features, using Ensemble Empirical Mode Decomposition to extract energy feature and amplitude characteristics of the first IMF. Taking full account of the issue of characters tics failure, combing the features and selecting the threshold, a decision tree classifier is devised to classify the type of disturbance. Experimental results show that the system can accurately identify eight kinds of single power quality disturbances and16kinds of complex types of power quality disturbances with40dB,35dB,30dB SNR conditions, own stable recognition performance and strong anti-noise Performance.This paper work is supported by the program for New Century Excellent Talents in university (NCET-11-0715), the project sponsored by the Scientific Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry and the National Natural Science Foundation of China (61170016,61373047).