The Application Research Of Deep Learning In Power Quality Disturbance Identification

With the progress and development of China's power industry technology,the high sensitivity power load is increasing.As a result,the demand for power quality of power users is higher and higher,power quality has been more and more get people's attention.In order to ensure the normal operation of all kinds of electrical equipment,proper measures must be taken to improve the power quality.Firstly,it is necessary to effectively identify various types of power quality disturbance signals.In this thesis,the research status at home and abroad are analyzed and summarized,then from the perspective of artificial intelligence,deep learning algorithms are used to study the signal identification of power quality disturbance by combining the theory of compression perception and phase space reconstruction.In this thesis,the generation mechanism,specific characteristics and characterization parameters of disturbance in power system are introduced.The mathematical model of power quality disturbance is established,and 14 common single and compound disturbance signals are obtained by MATLAB simulation.Then,Deep Belief Network and Convolutional Neural Network are used to realize the recognition of power quality disturbance signals respectively.The specific research contents are as follows:(1)Firstly,this thesis analyzes the characteristics of power quality disturbance signals,the sparse transform base,observation matrix and reconstruction algorithm for Compressed Sensing are determined,use the theory of compressed perception realizes the disturbance signal compression and reconstruction.Then the accuracy of the disturbance signal reconstruction is verified by MATLAB simulation,and the sparse vector can express the original disturbance signal.Through the MATLAB platform to build a Deep Belief Network and input data and conduct training,obtain the disturbance classification model.Finally,the network performance test is conducted to analyze the classification effect of the disturbance signal.Simulation shows that this method can effectively identify the types of disturbance and single compound disturbance,the use of the sparse vector greatly reduced the original disturbance data storage and the number of iterations and training times of Deep Belief Network,improve the efficiency of the power quality disturbance identification.(2)The phase space reconstruction of power quality disturbance signals are realized by coordinate delay method.The reconstructed trajectory picture of disturbance signals iare obtained by MATLAB simulation,the one-dimensional signals transform two-dimensional images,The images are processed by binarization,and the reconstructed trajectory picture of disturbance signals are generated in bulk.Then the Convolutional Neural Network model is constructed under the Caffe framework of deep learning.The trajectory picture sample of the disturbance signals are input,and the Convolution Neural Network is trained from the angleof image processing to extract the graphic features and realize the disturbance classification.Simulation results show that this method has high classification accuracy and good anti-noise capability.