Research Of Power Quality Disturbances' Detection And Classification Method

In recent years along with the expanding scale of power system,the Power electronics technology continues to improve,a large number of nonlinear shock wave load were introduced,wind power and photovoltaic power generation have brought a series of power quality decline.However,the continuous development of science and technology,Rapid increase of computer level,a variety of high sensitive precision instruments in the proportion of industrial production growth,the demand for power quality also will increase.At the same time,the quality of power will have a huge impact on industrial development,the improvement of living conditions of the residents and the rapid development of the national economy.As a result,analysis all kinds of power of disturbance signal,extract the characteristic parameters and identify the disturbance category for the improvement of the subsequent electricity,add compensation equipment,power system and control power quality falling pollution has the important meaning.In this paper,based on the existing research and combined with the tutor's project,the detection and classification methods of the disturbance signal in the power quality are studied.This paper begins by analyzing the power quality and has carried on the thorough research to the existing detection and classification methods.And finding a certain rule in the process of analysis and research,the mathematic model is used to establish the mathematical model of normal voltage,voltage drop,voltage rise,voltage interruption,harmonics,harmonic surge and harmonic sag.Simulation is carried out by using matlab simulation platform.In this paper,the power quality related issues are described in detail.On the use of the transformation theory such as:S transformation of the formula derivation process and the wavelet transform to be described in detail.At the same time,the structure,principle and the advantages and disadvantages in the classification problem of BP and PNN artificial neural network are described.This paper mainly uses two methods to study the detection and classification of power quality disturbance signals:(1)The perturbation matrix is obtained by using the S transform principle.And 11 kinds of eigenvector sequences of each disturbance category are extracted as the classification basis in the time-frequency domain.And in the time-frequency domain to extract each category of disturbance,11 kinds of characteristic vector sequence is taken as the basis of classification.Due to PNN probabilistic neural network good pattern classification ability,two PNN networks are used to classify the perturbations(PNN-I solves the presence of harmonic components and PNN-? to solve transient categories).(2)A method of detecting and classifying perturbation based on wavelet transform and synthetic neural network classifier is proposed.Using the multiresolution analysis capability of wavelet transform,the simulation perturbation is decomposed into nine different scales.The energy values of each scale are obtained by wavelet reconstruction as the basis of synthetic neural network classifier.In this paper,the single BP neural network classification is easy to fall into the local optimal,the convergence rate is slow,Particle swarm optimization algorithm(PSO)easy to "gather" appear "precocious" phenomenon and other issues to explain.Using the simulated annealing algorithm(SA)to optimize the three parameters(??C1?C2)of the Particle Swarm Optimization(PSO)algorithm.And then using SAPSO to optimize the weight threshold in BP network to get the integrated neural network classifier(SAPSO-BP).This classification of network compared to a single BP network in computing speed and accuracy is obviously improved.At the same time,greatly improve the PSO global search capabilities,and effectively reduce the probability of particles into the local extreme.Lastly,this paper compares the simulation results of the two detection methods,and find that the wavelet transform and the integrated neural network are better than the power quality disturbance detection and classification method.