Source Inversion Of Microbial Self-potential Based On Global Optimization Combined With Gradient Method

The self-potential(SP)method is an economical and efficient geophysical method,which has a very broad application prospect in pollutant tracking,microbial activity detection and ecological restoration.However,due to the complex formation mechanism of geomicrobial SP anomalies,there are few two-dimensional and three-dimensional inversion work at present,and the existing algorithms are all based on strict derivative calculation.Although these algorithms perform well in efficiency and accuracy,they depend on the selection of the initial solution.To this end,a hybrid optimization algorithm based on the combination of particle swarm optimization(PSO)and the gradient method is proposed to carry out the two dimensional source inversion of microbial self-potential,for the purpose of accurately locating the spatial distribution of the SP sources closely related to the activities of typical geomicrobes.This hybrid method takes the random search results of PSO as the initial solution of the proposed gradient method,and then takes the iterative results of the gradient method as the final inversion results.On the basis of the classical PSO,the damping factor and the resistivity constraint matrix are introduced to form two variants,and these two variants are then combined in the inversion to meet the exploration and exploitation requirements in different situations.While in the gradient method,the resistivity-based depth weighting matrix is added to promote the migration of the SP sources from the earth surface to their original depth.And to conform to the actual distribution of the SP sources,the minimum support stabilizing function is introduced to impose additional compact constraint.The validity of PSO,the gradient method,and the hybrid method is firstly verified in the synthetic experiments.Firstly,the finite element method is adopted to simulate the surface SP response of the given SP sources.Then the simulated measured data are introduced into the three proposed algorithms respectively for model parameters inversion.Numerical experiment results show that PSO is not affected by the selection of the initial solution and does not need gradient calculation,but the convergence speed is slow and the accuracy of the inversion results is limited.The gradient method has fast convergence speed and high precision,but it is highly dependent on the initial solution.In contrast,the hybrid optimization method can overcome the defects of PSO and the gradient method simultaneously,which can not only give high precision inversion results quickly,but also not affected by the selection of the initial solution.On the basis of synthetic experiments,The sandbox monitoring experiment of biological degradation is carried out,during which the SP data are obtained by continuously recording the SP signals generated by the typical geomicrobes Shiwanella Oneida MR-1.Then the measured SP data is introduced into the hybrid method for inverting the source location.The sandbox experiment further verifies the effectiveness of the hybrid method,and shows its great potential in environmental protection.