Detection And Analysis Of Power Quality Disturbance Based On Local Characteristic-scale Decomposition Method

With a large number of loads have access to the modern power grids such as volatility loads,impact loads and non-linear loads,the problems of electricity grids are increasingly serious and prominent.In order to improve the power quality effectively,it is necessary to position the starting and ending times of the power quality disturbances quickly and detect the instantaneous parameters of the power quality disturbances accurately,so that various types of power quality disturbances can be identified and some appropriate governance measures can be adopt to improve power quality.This paper focuses on the detection and classification of power quality disturbances.Firstly,aiming at the problems of EMD method such as false components and end effects,the LCD method is adopted to improve it.Furthermore,aiming at the mode mixing phenomenon still exists in LCD,one Ensemble LCD with Complementary Adaptive Noise is proposed.By adding pairs of positive and negative noises,the proposed method can suppress the residual noise in components and improve the decomposition speed and the accuracy of the components even the number of ensemble trials is few.At the same time,the Empirical AM-FM Decomposition method is adopted to solve the energy leakage and large errors in demodulation results of Hilbert Transform.The superiority of the Empirical AM-FM Decomposition method is verified by the demodulation results of single component.And then,one detection method of power quality disturbances based on ELCDCAN and EAM-FMD is proposed.The feasibility and adaptability of the proposed method to the positioning of full and non integer period transient disturbances' time,the extraction of instantaneous amplitude and frequency of steady-state and time-varying harmonics,the extraction of multi-frequency voltage flicker envelope and the detection of combined disturbances are verified by simulation experiments.Aiming at the large residual noise of high-frequency transient disturbance denoised by ELCDCAN,a denoising method combining ELCDCAN and wavelet soft-threshold denoising is proposed.It is verified that the method has good denoising effect through SNR and other parameters.Also,the validity of the proposed method in the application of power quality disturbance detection engineering is verified by the measured current harmonic data of the power grid.Finally,one classification method of power quality disturbances based on Multi-features and optimized SVM by Artificial Bee Colony is proposed.The features are extracted based on the amplitude and frequency of power quality disturbances and the amplitude and energy of the first component decomposed by ELCDCAN.Aiming at the penalty factors and kernel parameters need to be set by human experience in SVM model,this paper utilizes ABC to optimize the SVM model to determine the optimal parameter combination.Through the comparison of GA-SVM,PSO-SVM and ABC-SVM in the running time during parameter optimization and the classification accuracy of the single power quality disturbances,the statistical results show that ABC-SVM is superior to the other two methods.Meanwhile,it is verified that ABC-SVM can realize accurate classification for both the single and dual power quality disturbances under different SNR environment.