基于改进粒子群算法的大地电磁反演

粒子群算法是一种粒子群在全空间随机搜索的非线性反演方法，具有易于实现的优点，已在大地电磁（MT）反演得到了广泛应用，但其存在容易陷入局部最优解的缺点，在二维反演中应用较少且效果不好。本文提出了一种改进的优化粒子群算法，整个进化过程引入了局部进化，并且添加收缩因子和惯性权重参数，来改善该算法容易陷入局部最优解的缺点。最后将改进算法应用于二维MT反演，反演时加入核函数，结果表明改进粒子群算法在过早收敛问题上有明显改善，反演异常体位置也与实际模型吻合较好。因此，本文改进的粒子群优化算法提高了MT反演精度，为矿产资源勘探开发提供了理论基础。

MT inversion based on improved particle swarm optimization algorithm

The particle swarm optimization (PSO) algorithm is a nonlinear inversion method based on random search of group cooperation. It has the advantages of easy implementation and has been widely used in magnetotelluric (MT) inversion. However, it is easy to fall into the local optimal solution, which is less applied in two-dimensional inversion and the effect is not good. An improved optimized PSO is proposed in this paper. Local evolution is introduced in the whole evolution process. Shrinkage factor and incremental inertia weight factor are added to the velocity update formula, which improves the fault that the algorithm is easy to get into local extremal. Finally, the improved algorithm is applied to two-dimensional MT inversion, and the kernel function is added to the inversion. The results show that the improved PSO algorithm has a obvious improvement in premature convergence, and the location of the inversion anomaly is also in good match with the actual model. Therefore, the improved PSO algorithm improves the accuracy of MT inversion and provides a theory basis for the exploration and production of mineral resources.