The Design And Implementation Of Lightning Overvoltage Recognition System Based On Wavelet Transform And Principal-Component Analysis

Lightning over-voltage is generated in the external of power system. It has a large impact of high impulse voltage currently, this would do more harm to power system than internal overvoltage. In China, the electric power system have a very wide range and need high degree of stability because it is very complicated. In order to ensure the economy of power grid construction, lightning protection measures different types for different lightning overvoltage. Lightning overvoltage is more difficult and more limited than internal over-voltage. Therefore, the recognition of lightning overvoltage provides an important reference value for the surveillance, detection, treatment and maintenance of ransmission lines fault.This thesis based on the theory of wavelet analysis, principal component analysis and neural network. In view of the transmission line lightning over-voltage classification recognition problem,the thesis has made a deep rearch in Data analysis and feature extraction method based on Wavelet and Principal Components Analysis. Furthermore, this thesis introduced the software design and implementation of lightning overvoltage identification system based on Wavelet principal component analysis and the Improved neural network classification recognition method.This thesis is mainly reflected in the following three aspects:1. data processing for lightning overvoltage:Lightning over-voltage has characteristics of High frequency and transiention. In addition, Wavelet transform has good performance in both time domain and frequency domain. With the thought of reducing the data dimension by using Principal Component Analysis (PCA), this thesis proposed a method to analyze lightning over-voltage based on the wavelet transform and principal component analysis method.2. Improvement based on the classification model of neural network:In traditional BP Neural Network, the speed of data classfication is very slow, and the efficiency of model training is very low. This thesis improved the convergence rate of training and improved the accuracy of classification.3. Design and implementation of the system:Using Java-web techniques, combined with MVC framework, this thesis achieved a software system for the processing, analysis and identification of the lightning overvoltage providing data.This thesis has verified the validity of the classification model, and has achieved the software system based on the classification model. The system will be improved in the fulture research.