大地电磁的人工鱼群最优化约束反演

大地电磁的反演问题是非线性,如果采用线性反演方法容易陷入局部极小,使得反演结果非唯一性严重.本文将人工鱼群算法引入到地球物理反演之中,提出了非线性的大地电磁人工鱼群最优化反演.该方法不需要进行偏导数的求取,可以对反演的范围进行约束,以减小反演结果的非唯一性.同时我们对搜索步长进行了改进,给出适用于大地电磁反演的人工鱼群参数.大量的理论数据试算表明,人工鱼群反演算法能够较好地寻找到全局最优解.实测数据的处理结果表明,该方法可以用来处理实际资料,并且能够取得很好的应用效果.

Constrained inversion of magnetotelluric data with the artificial fish swarm optimization method

The inversion problem of magnetotelluric (MT) is nonlinear. It often falls to local minimum and serious non-uniqueness of inversion results if a linear inversion method is adopted. In recent ten years, with the development of artificial intelligence some new bionic algorithms such as the ant colony algorithm and particle swarm algorithm are gradually introduced to solve geophysical inverse problems. However, the research in MT inversion with the artificial fish swarm algorithm which belongs to bionics algorithm has not been yet reported in the relevant literature. A constrained inversion method of MT data by the artificial fish swarm algorithm is presented in this paper. The basic implementation of the MT data inversion based on the artificial fish swarm algorithm includes the following four steps: model selection, preying, swarming and following. A given artificial fish swarm is the constrained model space of MT data, which consists of resistivity and thickness. The initial model is generated by the uniform probability distribution in the model space to simulate the random behavior of the artificial fish. Preying, swarming and following are the natural behaviors of fish to find food. Fish usually stays in the place with a lot of food, so the behaviors of fish based on this characteristic are simulated to find the global optimum, which is the basic idea of the artificial fish swarm algorithm. Meanwhile, a variable search step method is presented, and the artificial fish swarm parameters suitable for MT inversion are given. We test our inversion algorithm on a three-layer model and a six-layer model to demonstrate the validity of the method. The average results of the inversion are almost the same as the true model parameters, which proves the validity of the algorithm. Real data of Line A is taken as an example for artificial fish swarm inversion and RRI inversion. The calculated apparent resistivity section of artificial fish swarm inversion agrees well with the observed apparent resistivity section. It indicates that the inversion result is reliable. The layer of artificial fish swarm inversion is consistent with the result of RRI inversion. It can provide a reliable basis for the precise interpretation of lithology and lithofacies by seismic data.We have succeeded to develop the MT data inversion with artificial fish swarm algorithm, and the search step is improved for MT data inversion. The calculation of the partial derivative is not needed and the inverted parameters can be constrained, which can reduce non-uniqueness of the inversion result. The synthetic data tests prove that the artificial fish swarm inversion algorithm can be used to find the global optimal solution effectively. The real data test shows that this method can be used to process field data and a good result can be achieved.