Corrosion State Prediction Methods For Grounding Grids

As the large-volume, long-distance high-voltage grid, super grid and ultra-high voltage grid appear successively, higher voltage leads to stronger earth circuit current of the electric power system and higher potential rise in the grounding grids of substations. This sets a higher requirement for the safe and stable running of substations. However, the open-cut inspection is often used to know the running condition of grounding grids because they are buried under the ground. Therefore, it is of great theoretical and practical significance to study the predicting methods for the corrosion status of grounding grids in order to grasp the operation condition of grounding grids in time and make pre-warnings and maintain facilities.Taking the pure resistance model of grounding grids as the object of study, this thesis puts forward a dynamic method to trace the branch resistance values of grounding grids based on the extended kalman filter. It defines a state equation of increments of resistance value in the time (n+1) and time n and an observation equation of voltage increments and resistance increments of accessible testing nodes in the time n, which can be combined with the initial resistance value of branches to predict the average change rate of branch resistance.As for the static corrosion rate prediction of grounding grids, some important environmental factors in the soil that are related to the metal corrosion are made use of to put forward a static prediction method on the base of PCA-BP neural net. PCA method is mainly used in the dimension reduction of the high dimension matrix of soil corrosion factors to weaken the cross correlation among every values of soil factors. After the dimension reduction, the principal component vectors are inputted in and train the BP neural net. Then the prediction process of the BP neural net will involve more environmental information, which will stimulate the function of influential factors of soil where grounding grids are and better fitting between the static prediction of corrosion rate and real environment.In order to integrate the results of dynamic tracing and the static prediction, this thesis poses a classification method for the corrosion condition of grounding grids on the foundation of supporting vector machine. The dynamic tracing results of branch resistance change rates in the grounding grid, the static prediction results of the corrosion rates of the grounding grid are combined with important corrosion factors in the soil, all as characteristic quantities, to be normalized to train the SVM as inputs. Testing statistics are inputted in the trained SVM and then the classification results are compared with the actual corrosion condition of the grounding grid. The accuracy of the prediction method of corrosion condition can be proved.