Research On Analysis Method Of Cascading Tripping In Power Network Based On Pattern Recognition Technology

Technological developments in the past half-century have enhanced the demand for electricity throughout the world.As a result,modern power grid systems are progressively large-scale,resulting in an increase in the complexity and vulnerability of the entire system.Due to the greater complexity of the power grid system,unpredicted initial faults events may cause cascading trip events even complex cascading failure in the power grid.Various researchers have introduced different techniques for predicting and controlling cascading trip or cascading failure in the power grids.This paper proposes an algorithm based on pattern recognition method for predicting the cascading trip according to the node injection power.The main content of this study includes:(1)According to the basic principle of power flow distribution in the power grid,the relationship between the electrical quantity of each branch in the power grid and the power injected by the node is analyzed in detail.Further,according to the chain trip occurred in the grid,the action behavior of the electric protection is summarized.When the power flow is transferred after the initial fault rectification of the power grid.The equations of the current type backup protection and the distance type backup protection are used for this purpose.The detailed discussion is made according to the action equations of the two types of backup protection,and the electric quantities such as current and impedance in the action equation.The relationship between the chain trip and the node injection power is generated.(2)The model identification method is used to predict the cascading trip of the grid which is based on the relationship between grid trip and node injection power.Sample data,as well as input and output features of the system,are presented in this work.The input feature quantity is the node injection power in the power grid,and the output quantity is the result of the cascading trip of the power grid.(3)The analysis program is compiled in the MATLAB environment.The grid-oriented cascading trip prediction algorithm based on pattern recognition method is given.The detailed algorithm flow is given for the related theory of BP neural network combined with the pattern recognition method.The IEEE-14 node system and the IEEE-39 node system are used to analyze the examples.The effectiveness of the proposed algorithm is further proved by the analysis of these examples.In this thesis,when training and testing BP neural network,the total number of samples is 400,and the hidden layer of neural network contains up to 60 neurons.The maximum number of iterations for training is 1500,the training requirement accuracy is0.00000002,and the learning rate is 0.1.The results show that the greater the number of neurons,the greater the accuracy.It has been observed that for the IEEE-39 node system,When using 60 neurons,the prediction accuracy is 100%,with a simulation time of 42 seconds.On the other hand,for the IEEE-14 node system,When using 60 neurons,the prediction accuracy is 98.79%,and the simulation time is only 6 seconds.The observed difference in time is due to the complexity of the system.In summary,this study has verified through the research that it is effective to use BP neural network to predict grid trips,which can provide reference for further research and actual operation.