The Study On The Application Of Corrosion Rate Prediction Of Grounding Grids Based On The Neural Network

Substation grounding grid is an equipment to guarantee safe and reliable operation of thepower system, and it is important to ensure the safety of personnel. Due to the grounding gridis buried underground, it is under perennial soil corrosion and easy to break.Severe breakagemay directly endanger the stable operation of the power grid and electrical equipment andpower station staff safety，and it also can affect the normal life of the people. So researchingthe prediction method of the corrosion of ground grid, timely troubleshooting, nipping in thebud and preventing the loss of life and unsafety are of great importance in an accident and candirectly serve the electric power industry.This article discusses the domestic and international ground ing grid detection technologyresearch，focuses on three methods of nondestructive testing technology of grounding gridincluding resistance measurement, magnetic field measurements and electrochemical testmethod, and analyzesits features and deficiencies. On this basis, combined with geneticalgorithm and BP neural network advantages, a grounding grid corrosion rate predictionmodel is established, which can provide some technical support for the grounding gridprotection. The main work of this paper are as follows:First the principle of grounding grid corrosion is analyzed, and the various factorsincluding soil porosity, water content, resistivity, pH and ionic elaborated are discussed. Torelated corrosion situation, including main micro-cell corrosion, focus on macro cell corrosioncaused by stray current corrosion and corrosion caused by microorganisms, we determined thephysical and chemical mechanism.Secondly, the corrosion rate forecast models including linear regression model and theBP neural network model are studied, and the advantages and disadvantages are analyzed.Thus we have a good knowledge preparation for the new model.Then the basic principles of genetic algorithms and artificial neural networks are studiedand analyzed, especially the principle of BP neural network learning algorithm, the trainingprocess and its application in the prediction of ground grid. On this basis, combined withprincipal component analysis and genetic algorithm, the traditional BP neural networkprediction model is optimized, effectively reducing the number of variables in the networktraining process and optimizing network initial weights and thresholds. The optimized model can be effectively applied to the prediction of grounding grid corrosion rate.Finally, the numerical software MATLAB is used for simulation, and the optimizedmodel is compared with the traditional models. Numerical result shows the accuracy andvalidity of the optimized model.