Transient Power Quality Identification Based On ELMD Multi-scale Fuzzy Entropy And Probabilistic Neural Networks

Electricity as an indispensable part in people’s life, the whole community are paying more and more attentions to the power quality problems. Improve and enhance the power quality of important premise is to effective detection and accurate classification. In this paper, starting from the definition of power quality and power quality research present situation. A new method of power quality disturbances identification and classification is proposed based on the common recognition method. The main works of this paper are as follows:(1) The basic theory of transient power quality is elaborated introduction, including the concept of power quality and classification of power quality disturbance. It bring the necessary to de-noise of the power quality signal before detection and recognition, using wavelet threshold method for signal noise reduction. Research on the basic principle of local mean decomposition(LMD) and its application in signal decomposition, the LMD method appear to the model aliasing, so So the introduction of noise auxiliary ELMD decomposition method.(2) On the basis of master Fuzzy Entropy algorithm, Fuzzy Entropy combined with multi-scale analysis, multi-scale fuzzy entropy(MFE) as a measure of the time series complexity under different scale factors is troduced. Then, ELMD and MFE method of combining the charateristics of disturbance signals are extracted. Feature vectors are able to reflect complexity information of signal disturbance under different frequencies. Therefore, MFE to build into a feature vector for the disturbance signal recognition classification provides the basis.(3) The common neural networks structure model and performance indicators and their advantages and disadvantages are analyzed, probabilistic neural network(PNN) is the main researched. PNN as disturbance identification method is determined. Using simulation experiments on the performance of several neural networks were compared and analyzed. Further evidence that the effectiveness of PNN is validated.(4) Using MATLAB to establish simulation of transient power quality disturbance signals model. Combination of EMD and Hilbert Transform method for the disturbance signal location. For recognition of disturbance signal, using wavelet thresholding method for de-noise of disturbance signal, and ELMD and MFE method of combining the charateristics of disturbance signals are extracted. PNN is completed for identification. Simulation results show: This paper proposes method can be effectively used for transient power quality disturbance detection and recognition and to achieve good results.