Study On Feature Extraction And Recognition Methods Of Residual Current Transient Process For Low-voltage Power Grids

Residual Current Operated Protective Devices(RCDs)have a wide range of application in low-voltage power grids.RCDs play an important role in preventing electric shock hazard and avoiding fire disaster caused by ground fault.Currently,the stocking current of animal and human being has no relationship with the setting value of action current from the protection devices.In general,the root mean square(RMS)value of residual current detected is considered as the unique criterion to determine if the protector acts or not.Theoretical analysis and operation experiences indicate that such criterion is unavailable in identifying the shocking current signals from animals and human beings.Consequently,it makes human beings unsafe due to the electric shocking,and the disadvantages of the operation principle inherently exist.It results in the malfunction and tripping phenomenon,and greatly decreases the reliability and the rate of proper commissioning for RCDs.To resolve the problems,methods of feature-extracting and identifying residual transient current in low-voltage power grids is developed.In this dissertation,the real-time electric-shocking experiment platform is established.Then,the original data of electric-shocking some animals are detected and obtained.Finally,the feature extraction and identification methods of processing the residual transient current in low-voltage power grids is developed and validated.This work is financially supported by the Projects 1)"the spectral characteristics and rapid identification methods about shock current component of the residual current" from National Natural Science Foundation of China;2)"study on safe use of electricity of key technologies based on residual current protection in rural areas" from State Grid Technology Projects.This dissertation makes use of intelligent information processing techniques such as wavelet transform,genetic computing,artificial neural network,et al.Through the establishment of an electric shock experiment platform,a lot of original data of current shocking wave from human beings and animals are obtained,and the database is built up.The Fourier spectra and waveform characteristics of residual transient current are analyzed,and then,the multi-dimensional characteristics of the amplitude and the energy about the residual transient current component are extracted.Thus,the detection and location of the electric shock fault moment,the identification of the current shocking fault type and the extraction of electric shock branch current component are determined.The main contents of this paper are summarized as follows:(1)The electric shock experiment platform was designed and built up.Through the experiments by simulating various scenarios of electric shocking,a lot of animal/plant electric shock data was obtained.Therefore,organisms shock waveform database for the rural low-voltage power grids was established.(2)The Fourier spectrum of residual current is analyzed by using fast Fourier transform when the organism electric shock fault occurs.The individual component of residual transient current and its variation were studied.An algorithm of combining mathematics gradient and morphological gradient was proposed.In analyzing the residual current waveform characteristics,the characteristic change of the original signal waveform in electric shock moment can be highlighted.(3)A variety of signal processing methods including mathematical statistics,fast Fourier transform and empirical mode decomposition are used.The amplitude calculation and energy feature extraction method of the multi-dimensional residual current transient component is developed.This method highlights the feature vector of the organism electric shock signal,and can be successfully applied to identify the types of electric shock.(4)By using discrete Hilbert transform,the accumulation of the residual current instantaneous phase difference is considered as the criterion,leading to the extraction of the phase mutation feature.By using the mutation calculation method,the accumulation of the sudden increment of the residual current signal amplitude is considered as another criterion,leading to the extraction of the amplitude mutation characteristic.Therefore,the electric shock fault moment detection method based on the phase and amplitude multi-mutation variety criterion was proposed.(5)Through the combined use of wavelet transform and feed-forward neural networks,a method of identifying electric shock fault type was established.By analyzing the residual transient current component amplitude and energy characteristics,the structure of neural networks is optimized by using quantum genetic algorithm.This leads to the improvement of the learning methods of neural network so that the training effect is significantly increased.(6)A novel method of identifying and calculating the electric shock branch current component is proposed,which is based on finite impulse response filtering and radial basis function neural network.The problem that the organism shock branch current is not detected in the engineering is resolved using this method,thus a reliable theoretical basis for developing a new generation of adaptive residual current protection devices is provided,which is based on electric current component of the human body caused its action.