Magnetotelluric Inversion Method Based On Particle Swarm Optimization

In order to carry out the General Secretary Xi JinPing’s important instruction of “march to the deep earth”,the Ministry of Land and Resources,the Ministry of Science and Technology,the Ministry of Education and other relevant departments jointly launched the national key research and development plan “Deep Resources Exploration and Mining”.Because of the large depth of exploration of the magnetotelluric sounding method,it has a relatively good prospect in the exploration of deep earth resources.Magnetotelluric sounding has been widely used in the exploration of petroleum,metal,groundwater and other resources.However,the key part of the method: inversion,because under certain conditions,the effect is still unsatisfactory and needs further improvement.This paper explores the technical method using particle swarm algorithm to invert the magnetotelluric sounding data.Particle Swarm Optimization is a method for finding the optimal solution and has been widely and effectively applied in many fields.However,there is almost no application in the field of electromagnetic exploration.Firstly,based on the in-depth analysis of the basic idea and detailed algorithm of particle swarm optimization,combined with the basic principles and technical requirements of magnetotelluric sounding inversion,we designed a new inversion method: magnetotelluric inversion method based on particle swarm optimization.Secondly,according to the design inversion method,we have written the forward modeling and inversion procedures of the magnetotelluric sounding on the matlab software development platform,and realized the inversion software of the magnetotelluric sounding based on the particle swarm optimization.Then,we used the theoretical data to test the correctness of the inversion method and proved the correctness of the inversion procedure.Five different geoelectric models were designed.The theoretical response curves were calculated using the forward program.The inversion software developed by us was used to invert the theoretical data,and the inversion results were compared with the known models.Finally,the measured magnetotelluric sounding data of the Wulungu Depression in the Junggar Basin was used to calculate the inversion,and the comparison with the seismic interpretation results shows the effectiveness of this method.The core of the "magnetotelluric inversion method based on particle swarm optimization" proposed in this paper is three points: reasonable design of objective function,design of geoelectric model and high-quality random number.The objective function of the conventional magnetotelluric sounding inversion method basically adopts the L2-norm of apparent resistivity,while the L1-norm of wave impedance is adopted in this paper.In the design of geoelectric model,according to the nature of the problem,the uncertainty of the inversion can be reduced and the effect of the inversion can be improved by limiting the type and number of parameters of the inversion in an appropriate way.In the algorithm implementation,the quality of the random number affects the inversion effect.In this study,a new and higher quality random number generator is adopted: Mersenne Twister.In general,the magnetotelluric sounding inversion requires the inversi on staff to give an initial geoelectric model,and the inversion result is related to the initial model.It is necessary to give a reasonable initial geoelectric model to obtain a more objective inversion result.However,the particle swarm inversion method proposed in this paper does not need to artificially give the initial model to obtain the global optimal geoelectric model,which makes the inversion result more accurate and reliable.