Research Of Location And Recognition With Power Quality Disturbances For Distributed Power Connection With Grid

With the rapid development of the smart grid in China, A large number ofdistributed powers such as wind energy, solar energy, micro-turbines and othersare connected with grid. More and more power quality problems have beenattention increasingly and are becoming a key research direction in the powersystem. In this paper, the research of location detection and classificationrecognition with power quality disturbances for distributed power connectionwith grid is the main content.According to the definition and related standards of power quality, as wellas the ways of distributed powers connection with grid and its characteristics,the article describes several common power quality disturbances.11kinds ofpower quality disturbances,that are voltage deviation, voltage deviation,voltage sag, voltage swell, voltage interruption, voltage flicker, harmonics andthe composite disturbances of harmonic and voltage deviation, voltagedeviation, voltage sag, voltage swell, voltage interruption, voltage flicker withGaussian white noise,are used to analyze and study.Wavelet transform and S transform are respectively used to locate thestarting and ending time of five kinds of composite power quality disturbanceswhich have been denoised. The test results of the two methods show thatwavelet transform with better time-frequency resolution property has moreaccurate location detection than S transform.This paper presents methods for the classification and recognition ofpower quality disturbances based on wavelet transform, principal componentanalysis (PCA) or independent component analysis (ICA), and BP neuralnetwork or support vector machine (SVM). Firstly, Db4wavelet is used for11kinds of disturbances signal with multi-resolution analysis (MRA), and thewavelet energy difference coefficients between disturbances signal andstandard signal are extracted as feature vectors. Then, PCA and ICA areadopted to reduce dimensions and remove redundant information of featurevectors, which gets a new feature vector matrix. Finally, BP neural network andSVM are respectively used for the classification and recognition of powerquality disturbances according to the new feature vectors. The classification results of four methods show that According to the cumulative contribution rateof the feature vector of covariance matrix, it could choose suitable dimensionsof feature vectors and remove redundant information, in order to reduce thetime of classification and identification; With respect to BP neural network,SVM has higher classification accuracy rate and less classification time.In view of the analysis of power quality disturbances for distributed powerconnection with grid, software of power quality analysis platform is createdbased on hybrid programming between Matlab and VS2008. The systemrealizes window of detection and identification, and provides the basis foronline monitoring.