Research On Detection And Classification Algorithm Of Composite Disturbance Of Power Quality

In modern power systems,with the input and use of a large number of non-linear,impact loads and the wide access of distributed generation,the power quality of power network is facing a severe test.The complex and variable disturbances of power quality constantly threaten the safe operation of power system.Therefore,accurate and effective detection and classification of the characteristic information of power quality disturbance is the premise and key to solve such problems.The existing detection and recognition methods have problems such as large detection errors and low recognition accuracy when dealing with perturbation problems.For this reason,this topic focuses on three aspects:noise reduction,detection and classification of disturbances.WT-SSA fusion noise reduction algorithm,adaptive VMD-HT detection algorithm and improved 1D-CNN classification model are proposed respectively.The accuracy and validity of the algorithm presented in this paper are verified by experiments.In order to approximate the form of disturbance in the actual power network,eight kinds of composite disturbance signals are simulated according to the disturbance standard formulated by IEEE.For the actual power network signal susceptible to noise pollution,this paper first improves the traditional Wavelet Threshold Method(WT),then determines the selection principle of window functions and eigenvalues in Singular Spectrum Analysis(SSA).Combining the two methods,a WT-SSA fusion noise reduction algorithm is proposed.By comparing with the traditional noise reduction algorithm,the performance of the proposed algorithm is significantly better than that of the same kind of algorithm.In the stage of disturbance detection,a detection algorithm based on adaptive variational mode decomposition and Hilbert transformation(VMD-HT)is presented.The optimal number of decomposition layers and penalty factor of VMD are determined by energy difference method and whale algorithm,respectively.The exact detection of instantaneous amplitude,instantaneous frequency and start-end time of disturbance is completed by combining HT and maximum method,and the same detection methods are compared.The algorithm proposed in this paper has higher detection accuracy for both simulated and measured signals.In the stage of disturbance classification,a classification model based on an improved one-dimensional convolution neural network(1D-CNN)is designed.The 1D-CNN is improved by introducing residual structure,and the dynamic attenuation learning rate and Dropout layer are added to optimize the network training process to achieve accurate recognition of multiple complex disturbance signals.Compared with traditional 1D-CNN and cyclic neural network,the proposed classification model is superior to the other two models in recognition accuracy,convergence speed and curve stability after convergence.In the aspect of disturbance and noise reduction,the proposed algorithm can effectively solve the problem that disturbance signals are susceptible to noise interference during the acquisition process.In the aspect of disturbance detection,the proposed adaptive algorithm can detect the characteristic information when multiple composite disturbances occur,and the detection result is accurate and the positioning accuracy is high.In the aspect of disturbance classification,the designed classification model has good classification performance and robust noise.It has laid a solid foundation for the subsequent establishment of a perfect power quality monitoring system.Figure [53] table [12] reference [80]...