Research On Fault Classification Of Tower Grounding Grid Based On Deep Adaptation Model

The development of an accurate and efficient fault detecting method is of great significance to the maintenance of the tower.This paper combines electromagnetic induction theory and deep learning methods,and uses deep learning models to realize the fault location and corrosion diagnosis of the pole tower grounding grid.At the same time,according to the actual scene,the domain adaptive method is applied to the feature migration of the model,so that it can be applied to the diagnosis of general faults.The specific content is as follows:(1)The effectiveness of the electromagnetic induction method in the fault diagnosis of the grounding grid was analyzed and an electromagnetic model was established for the fault diagnosis of the pole tower grounding grid.At the same time,the influence of current excitation mode,current excitation frequency and soil layered structure on the distribution of magnetic induction intensity on the surface is discussed;A fault diagnosis experiment for the tower grounding grid was designed,and the data obtained through simulation and experiment were used to compare and analyze the distribution of magnetic induction under different fault conditions,and the general rules for fault location and corrosion degree of the tower grounding grid were obtained.(2)Aiming at the problem of manual classification of grounding grid faults by magnetic induction,a method of applying deep learning theory to diagnosis was proposed;Obtained magnetic field induction intensity data through simulation and model tests and a dataset was produced;The structure and training process of the 1D-CNN-based network model for fault diagnosis was designed,and the network parameters through training was determined;Experiments and comparison with other deep learning methods verify the effectiveness of the proposed diagnostic model.(3)Analyzed the diagnosis process and existing problems of the application of the deep grounding grid fault diagnosis model,and proposed the idea of improving the diagnosis performance of the model through feature migration;Improved the diagnosis model through fine-tune and domain adversarial adaptation(DANN)methods,and improved the migration performance of the model;The simulation diagnosis task verifies the performance improvement of the migration model in fault diagnosis.