Rapid Apparent Resistivity Imaging Of Transient Electromagnetic Using Ann And Application In Grounding Grid Detection

Transient electromagnetic(TEM)is a widely used electromagnetic measurement strategy.In practice,a mass of TEM data is generated due to the large range of detection or dense measurement points.Since transient electromagnetic methods have been proven to be feasible and effective for detecting faults and breakpoints in substation grounding grids,there are few professionals in the power industry that deal with TEM detection data.This has affected the promotion of this method in the field of substation grounding grid testing to a certain extent.In this paper,the technical scheme of neural network for transient electromagnetic fast resistivity imaging is systematically studied.Aiming at the characteristics of transient electromagnetic response and instrumentation,a sub-case neural network rapid imaging method was proposed and experimentally studied.On this basis,the characteristics of resistivity profile of substation grounding grid are systematically analyzed.The neural network clustering algorithm was used to study the difference of apparent resistivity profiles caused by flat steels with different thicknesses,and the concept of relative corrosion of flat steel was proposed.The research of the neural network for transient electromagnetic fast resistivity imaging shows that:(1)The neural network that uses the sampling time point and the induced voltage data together as the input variable and the apparent resistivity as the output variable is rapidly imaged,and the device parameters suitable for transient electromagnetic observation are fixed.This method avoids the problem that one induced voltage corresponds to two resistivity values.(2)The neural network apparent resistivity imaging of the nonlinear equation mode is suitable for the case where the TEM observation data is a vertical magnetic field.This method is applicable to the vertical magnetic field data of all transient electromagnetics,and is not limited by the size of the transmitting coil.(3)Because the limitation of BP neural network may lead to large error in apparent resistivity results,genetic algorithm optimization BP neural network(GABP)improves the accuracy of solution based on fast solution of apparent resistivity,15 measurements The time spent on the point is 0.4084 seconds,which has reached the level of instant imaging.(4)Meme Elite Pareto non-dominated sorting differential evolution algorithm design TEM rapid imaging neural network structure(MEPDEN)can meet the convenience of convenient call and can achieve better accuracy requirements,theoretical model study 62700 sample data consumption 0.405609 seconds The actual measured 143 measurement points consume 9.003 seconds,while the iterative method requires 900.517 seconds.(5)When the transient electromagnetic transmitting device is a small coil structure,the neural network structure for rapid imaging using the induced voltage and the vertical magnetic field data is the same;and the neural network establishing different time windows can solve the large transmitting coil device.There are many problems with induced voltage and apparent resistivity.(6)Using the rapid imaging of the neural network to quickly scan the grounding network for diagnosis,the periodic "physical examination" imaging of the grounding network can be carried out,and the corrosion status tracking and evaluation of the whole life cycle of the grounding grid can be realized through comparison of its own historical images.It is estimated that rapid imaging diagnostic technology and traditional technology can save hundreds of thousands to millions of yuan in economics.Analysis of the transient electromagnetic characteristics of the grounding grid shows that:(1)Transient electromagnetic grounding grid detection is only applicable to the experimental mode using single-shot single reception.(2)The characteristics of the apparent resistivity profile caused by the difference in mesh grid size are different,and the probe size also brings about the difference in apparent resistivity profile.The resistivity imaging obtained from the regional measurement of the grounding grid can more intuitively reflect the structure and fault location of the ground grid.(3)The characteristics of the transient electromagnetic signal and the apparent resistivity profile of the different thickness of the ground flat steel are significantly different.The thin flat steel exhibits a higher value of resistivity and a steeper trend,while the performance of coarse flat steel is reversed,which is the basis for the relative corrosion.(4)SOM neural network clusters the depth-resistivity matrix of the grounding grid to output the categories of intact,corrosion and breakpoints;the relative corrosion of all the measured points obtained by the distance output from the center of the class can be characterized by flat steel.The degree of thickness,indirectly quantitatively evaluate the degree of corrosion of the grounding grid.(5)The field test results of the true testing ground in Wuhan Nari and the Liangting substation in Chongqing validate the effectiveness of the method for judging the corrosion degree of grounding grid using clustering for GG-TEM depth-apparent resistivity data.The field excavation results show that the relative corrosion degree of grounding grid is consistent with the thickness of flat steel.