Transformer Fault Diagnosis Based On Kernel Principal Component Analysis And Quantum Genetic Neural Network

Transformer is one of the key equipment of power system,and it is also one of the factors that cause frequent failure of power system.Its operating state is directly related to the stable operation of the entire power system.There were relatively complex nonlinear relationships between the signs and types of transformer failures.The traditional detection methods obviously couldn't meet the requirements of real-time,rapidity and accuracy of the system.In recent years,intelligent diagnostic algorithms had shown strong advantages.In order to further met the requirements of real-time,rapid and accurate on-site,a method for transformer fault diagnosis using KPCA and QGA to optimize neural network was proposed.BPNN was used as the final link of transformer fault diagnosis.It had the advantages of parallel distributed processing,self-organization,self-adaptation,self-learning,etc.The activation function of BPNN could adopt any function,and its approximation ability was very suitable for dealing with transformer fault diagnosis.However,there were also disadvantages of slow processing speed and easy to fall into local extreme values.The simulation results showed that BPNN had not met the specified error standard in 2000,and the fault recognition rate was only 83.117%.the genetic algorithm could improve the BP network.The powerful parallel processing capability and superposition state of quantum computing could realize the optimization of genetic algorithm.Combined with the two,QGA could improve the BP network.According to the network topology structure,QGA was used to encode the structural parameters of the BP network,and the quantum revolving gate was used to update the population and search for the global optimal value of the BP network parameters.Transformer fault identification used the optimized network.The results showed that BPNN optimized by QGA achieves accuracy requirements in 634 generation,and the fault recognition rate reached 98.017%,which improved the diagnostic rate and diagnostic accuracy.The huge data repository in transformer fault detection systems was often redundant and nonlinear.In order to better apply QGA and BPNN to make accurate and rapid diagnosis of faults,KPCA first selected data information,eliminates irrelevant information and interference information.KPCA could effectively simplify the network structure,reduced the network operation dimension,and achieved the purpose of improving accuracy and speeding up the operation.The application of KPCA accelerates the convergence of QGA and neural networks.The simulation results showed that the quantum genetic neural network added to KPCA could achieve the accuracy requirement in 323 generation,and the optimal solution was obtained.The fault diagnosis recognition rate reached 98.836%,and the speed and recognition rate were improved.Figure26;Table26;Reference 55...