Research On RBF Neural Network Fault Location Method Based On Improved Fruit Fly Optimization Algorithm

In China,the level of economic development varies greatly from region to region.The major bustling cities are concentrated in the southeast coastal areas,which is also the region with the most concentrated power load in China.But the distribution of primary energy is just the opposite,mainly in the western region.This requires us to convert the energy resources in the west into electricity for the eastern region,and realize the country's long-term strategy of transmitting electricity from the west to the east.In order to achieve long-distance transmission,the construction of UHVDC transmission is very necessary.The advantages of UHV DC transmission are obvious,long transmission distance and small energy loss.Due to the long transmission distance,finding a fault location is also complicated when a fault occurs.Therefore,it is necessary to study the positioning technology of UHVDC transmission.When a UHV DC system fault occurs,a natural frequency will be generated on the line.There is a certain non-linear relationship between the fault distance and the natural frequency,but the fine natural frequency is not easy to extract.Therefore,the traveling wave spectrum energy is used instead of the natural frequency for distance measurement.The RBF neural network has high accuracy and is not easy to fall into a local optimum.Therefore,the RBF neural network is used to fit the relationship between the spectral energy and the fault distance.However,since the extended width ? will affect the performance of the RBF network,it will directly affect the accuracy of the final training result.Therefore,the fruit fly algorithm is used to optimize the selection and improve the model performance.According to the shortcomings of the fruit fly optimization algorithm,the fruit fly optimization algorithm was improved by changing the step size of the fruit fly optimization algorithm.The improved fruit fly optimization algorithm's optimization accuracy was significantly improved.Using the improved fruit fly optimization algorithm,the neural network is optimized to improve the performance of the neural network ranging model.The PSCAD / EMTDC software was used to build a simulation model of the Yun-Guang UHV DC transmission system,and the validity of the model was verified.Set a short-circuit fault on the positive pole of the transmission line,collect fault information in the case of different transition resistances and fault distances,and decouple the fault voltage signal using Karenbauer transformation,extract the line mode component,and obtain the line mode voltage component Bring into the wavelet packet decomposition program,perform three-layer wavelet packet decomposition to obtain the signals of each subband of the third layer,and calculate the energy ratio of each subband signal as the feature of the UHV DC line fault signal and input it into the fault location model After learning the model,the trained neural network can be used for fault location.Finally,the simulation verification shows that the errors of the simulation results are all within 1km,and the accuracy is better.Prove the effectiveness of the improved fruit fly optimization algorithm to optimize the fault location algorithm of RBF neural network.