Research On One Dimensional CSAMT Inversion Based On Improved Genetic Algorithm

With the rapid development of the national economy,the social demand for mineral resources,groundwater,geothermal,coal and other resources is becoming more and more urgent.CSAMT is widely used in resource exploration because of its advantages such as controllable emission source,large depth of detection,high precision and so on.As an important part of the controlled source audio frequency magnetotelluric method,the inversion results directly affect the accuracy of geological survey.Some of the traditional inversion methods such as Bostick method,inverse matrix,Marquardt method etc.mostly belong to the linear or the local linear method.Although these methods are classic and well known,but because of the inversion process prone to multiple solutions,will make the inversion effect with the initial establishment of model.However,because the inversion process is prone to multiple solutions,the inversion effect is related to the initial model.Because genetic algorithm is simple,easy to operate,do not rely on the initial model and other advantages,it is favored by many scholars.However,the ability of the standard genetic algorithm to find the optimal solution in the large search range is not prominent,and it will also appear "premature" and non convergence.Therefore,it is necessary to improve the standard genetic algorithm.In this paper,according to the characteristics of the problem to be solved,the appropriate objective function and fitness function are set up.Then,the standard genetic algorithm is improved,and the effectiveness of the algorithm and the anti noise performance of the algorithm are verified by using the layered model.Finally,the improved genetic algorithm is applied to the actual geological data to verify its practicability.The main work of this paper is as follows:1.The basic theory of CSAMT forward calculation is studied.This paper studies the horizontal layered geoelectric section,through the forward formula of CSAMT,calculates the electric field and magnetic field,and obtains the Cagniard resistivity.Because the Hankel integral appears in the forward formula,this paper uses the 120/140 linear filter coefficients of Guptasarma and Singh to solve the calculation problem.In this paper,the forward calculation of various geoelectric models is carried out,including D section,H section,A section and HK section.By using the measured data of geological data,the results are compared and analyzed.The implementation of forward calculation is the basis for the establishment of the objective function and fitness function.2.The paper studies basic theory of genetic algorithm,masters the operation mechanism of genetic algorithm.The paper understands the operation of standard genetic algorithm,and finds out the problem of standard genetic algorithm.3.The genetic algorithm is improved to solve the problems of premature and local convergence.In order to improve the selection operator,the selection operator is combined with the ranking method and the optimal individual reservation strategy.The improved selection operator can not only guarantee the convergence of the algorithm,but also maintain the diversity of the population.In this paper,the crossover operator is improved.The crossover probability is based on the adaptive probability method,and the father and son competition strategy to enhance the adaptability of the operator and retain the good parent.4.The improved genetic algorithm is applied to one-dimensional inversion of CSAMT.Firstly,the simulation results of two layer,the three layer and the four layer model are carried out to verify the effectiveness of the improved algorithm.Then,the random noise is added into the model to verify its anti noise performance.Finally,based on the actual data of a project in a certain area,this paper proves the practicability of the improved genetic algorithm.