Power Quality Disturbance Detection System Based On CS-CNN

With the growth of the national economy and technology,a large number of new electrical facilities are connected to the grid,making the voltage and current in the power grid more susceptible to disturbances.These disturbances can shorten the life of the equipment,increase the line loss rate,and even cause serious losses and consequences.In order to alleviate the consumption of power quality data acquisition and transmission and improve the intelligent identification of power quality disturbance categories,this paper combines Compressed Sensing(CS)technology with Convolutional Neural Network(CNN)to effectively solve two major problems.And this paper creatively proposes a method of mapping the reconstructed voltage signal into RGB picture,which realizes the classification of single-phase and three-phase power quality disturbance into the convolutional neural network.This paper can be summarized as the following three points:(1)Simulate the normal signal and 8 single-phase power quality disturbances with the function in MATLAB,and use Simulink to simulate the circuit,and simulate 10 kinds of three-phase power quality disturbance data.(2)Compressing and storing the simulated signal using compressed sensing technology to facilitate transmission;then using orthogonal matching tracking method to recover the original signal.Finally,the single-phase and three-phase power quality disturbances are converted into RGB pictures,which are respectively made into training set,verification set and test set.(3)Construct a complete convolutional neural network in Python by using tensorflow learning framework,adjust and optimize the parameters of the whole network by making a good training set and verification set,and finally use the test set to check the accuracy of the system classification.In the end,the paper tests various functions of the whole system.The test results show that the modules in all parts of the system can work stably and accurately.The system can effectively reduce the data transmission consumption and realize the intelligent classification of power quality disturbance.