基于自适应差分进化算法的MT信号激电信息提取

从MT信号中提取激电信息的研究有助于提高大深度探测技术的勘探精度、深度和范围。针对MT信号激电信息提取中存在的非线性和非凸特征,通过改进混沌自适应差分进化算法中进化参数的自适应策略,提出了一种基于非均匀统计分布的自适应差分进化两阶段最小构造反演方法。一方面该方法利用柯西分布和高斯分布的统计特性自适应获取进化参数F和CR,提高算法的全局搜索能力, 通过记忆以往迭代过程中的优秀进化参数提高算法后期的稳定性;另一方面该方法通过引入第二阶段的反演过程强化极化率对观测数据的影响;通过将正则化参数引入差分进化算法的适应度函数解决反演的多解性问题。对含激电效应的MT一维模型的反演结果表明,本文算法能够较好地重构地电结构和提取激电信息并在加噪环境下具有较强的鲁棒性。与其他非线性算法(混沌自适应差分进化算法,标准差分进化算法和粒子群优化算法)的反演结果对比表明,本文算法具有更为优越的全局搜索能力和较高的反演精度,适于微弱激电信息的提取。

IP extraction from magnetotelluric sounding data based on adaptive differential evolution inversion

Induced polarization (IP) information extraction from magnetotelluric (MT) sounding data is of great significance of earth deep structure and hydrocarbon exploration. Taking the nonlinearity and non convexity of MT IP extraction into consideration, a two-stage adaptive differential evolution (DE) inversion based on non-uniform statistical distribution and minimum structure is proposed by improving the adaptive strategy of evolutionary parameter in the chaotic DE algorithm. On the one hand, the statistical properties of Cauchy and Gaussian distribution are used to obtain the evolutionary parameter F and CR adaptively, which improves the global searching ability, and the successful evolutionary parameters in the previous iterative process are recorded to enhance the stability in later stage of the algorithm. On the other hand, the impact of the polarizability on observation data is strengthened by introducing the second stage inversion process, and the regularization parameter is applied in the fitness function of DE algorithm to solve the problem of multi-solutions in inversion. Inversion results of MT 1D model with IP effect show that geoelectric structure can be reconstructed and IP information can be well extracted. And the proposed algorithm is fairly robust in noise environment. Compared with inversion results of other nonlinear methods such as chaotic differential evolution (CDE), DE and particle swarm optimization (PSO), the proposed algorithm has better global searching ability and higher inversion accuracy, which is suitable for the weak IP extraction from MT signal.