Analysis Of Power Quality Disturbance Based On Wavelet And Neural Network

With the rapid development of global economy,the frequency and severity of power quality disturbance in power system are aggravated by the use of a large number of non-linear and impact load equipment.For example,capacitive heavy loads and power switches may cause voltage swell.The use of solid state switchgear and nonlinear power switching loads,such as rectifiers,can cause voltage sag or voltage interruption.And lightning strikes may cause flicker.The emergence of these power quality problems will lead to instability or fault operation of electrical equipment,and shorten the service life of electrical equipment,which is contrary to the high requirement of power quality in current power equipment.In view of the common problems of power quality disturbance detection and classification in power system,previous studies have shown that wavelet transform is used to detect disturbance signals with good accuracy,but wavelet transform is less intuitive in disturbance classification,which affects the accuracy of the classification.The neural network method has good identification ability for nonlinear systems because of its unique learning ability.In view of this,in order to analyze the disturbance of power quality accurately,this paper considers the research and analysis method of combining wavelet transform and neural network,uses the existing research for reference,explores the improvement method,and formulates the whole research idea of this paper.Firstly,the mathematical model of power quality disturbance is established according to the actual data of power system.Based on the analysis and comparison of different wavelet basis functions,according to the power quality disturbance,the wavelet basis suitable for power quality disturbance detection is selected,and the optimal wavelet decomposition layer number is analyzed and determined in order to detect the power quality disturbance accurately.It also provides the basis of decomposition scale for the subsequent wavelet denoising and feature extraction in wavelet domain.Secondly,in view of the good time-frequency local analysis ability of wavelet transform,a new power quality disturbance detection algorithm based on improved lifting wavelet is proposed in this paper.This algorithm combines weighted average filtering algorithm to improve the prediction of wavelet decomposition so as to improve the detection accuracy of power quality disturbance signal.Then the principle of modulus maximum is used to detect the starting and stopping points of the disturbance.The feasibility and effectiveness of the algorithm are verified by MATLAB simulation experiment and IEEE test waveform analysis,and the performance of the algorithm is compared with other detection algorithms.The results show that the integrated detection accuracy of the proposed algorithm is higher than that of other detection algorithms.And it has strong anti-noise property.Thirdly,in order to classify power quality disturbances more accurately,we consider extracting disturbance features in wavelet domain and time domain,and optimize them by simulation experiments,and the wavelet energy values and standard deviations of layers 4,5,6 are selected.These six features can best distinguish the types of disturbances.They are made up of feature vectors as the input of the classifier.Finally,a power quality disturbance classification algorithm based on improved BP neural network is designed.The learning rules are improved by reasonably designing the structure of BP neural network,adding momentum term and adaptive learning rate.Through the simulation analysis and experimental verification of the algorithm realized by MATLAB,the accuracy of the algorithm is obtained by comparing the output result with the actual value.The experimental results show that the improved BP neural network classification algorithm can achieve higher classification accuracy for single and composite power quality disturbances.