Research On Fault Diagnosis Of Three-level Inverter Support Vector Machine

With the continuous changes in the world energy field,the new energy technology has gradually become a hot field of research today.More and more attention is also paid to the inverter as the core of the new power grid connection.The safety and reliability of the inverter directly affects the stability of the whole new power generation system.The neutral point clamped three-level inverter is a relatively common high-power grid-tied inverter.This paper takes the NPC threelevel inverter as a research object and conducts the following main studies:From the analysis of the topology structure and the working principle of the NPC three-level inverter,the commutation path of the inverter and the possible wire breakage errors are first discussed,which are divided into two according to different wire breakage error types.The Simulink module in the simulation experiment is used to improve the sample data for later error diagnosis of the inverter.According to the sample data reflected by the error of the NPC inverter,through the study of wavelet packetization and entropy theory,the error signal combining wavelet packetization and Shannon entropy is used to solve the problem that the wavelet transform may have bad high-level processing effect-frequency signals feature extraction method.According to the characteristics of the support vector machine parameters,the Gray Wolf algorithm is selected to optimize the parameters of the support vector machine,and according to the characteristics and drawbacks of the Gray Wolf algorithm,dynamic weights and improved convergence coefficients are introduced and optimized.The performance of the improved algorithm has been tested.Finally,we will establish an NPC three level inverter failure diagnostic model that optimizes the parameters of the support vector machine using the improved Gray Wolf algorithm and optimizes the support vector machine based on the improved Gray Wolf algorithm.The standard Gray Wolf algorithm and particle swarm optimization support vector machine are compared,and the improved gray wolf algorithm optimization support vector machine greatly improves the efficiency of failure diagnosis.