基于差分演化算法的大地电磁测深数据反演方法研究

Magnetotelluric sounding method, as one of today's important geophysical methodsin detecting electrical structure of the deep Earth and continental dynamics research, is playing an increasingly important role in the deep energy exploration, natural disasterreduction, engineering and other fields. The conventional method of inversion has the difficulties of computational complexity and non-uniqueness, however, more and more difficult to meet the underground situation is more complex to deal with higher standards and requirements. For this method to solve difficulties in the inversion process, the project will introduce the differential evolution algorithm to the magnetotelluric sounding and research the magnetotelluric sounding data inversion method： (1) Building constraints in the inverse problem model, the differential evolution algorithm can automatically take advantage of the process of solving a variety of geological information and prior knowledge, thereby improving the inversion of multiple solutions. (2) Differential evolution algorithm is a real number coding scheme can effectively deal with the scope of the parameters, so that the inversion is not sensitive to the number of parameters and dimensions, adapt to the growing number of developing trends of magnetotelluric inversion parameters. (3) In order to solve the inversion computational complexity, a differential evolution algorithm is proposed and implemented under the cloud platform by modifying the adaptability of Map-Reduce model. Such intelligence magnetotelluric sounding inversion will lay the foundation for magnetotelluric sounding method which will be large-scale practical application.

大地电磁测深法作为当今地球深部电性结构探测和大陆动力学研究的重要地球物理方法之一，在深部能源探测、减少自然灾害、工程应用等方面发挥着越来越重要的作用。然而，常规的反演方法存在解的非唯一性和计算昂贵性等困难，越来越难以满足现在的地下情况更复杂、处理标准更高的要求。本项目拟将差分演化算法引入到大地电磁测深中，进行大地电磁测深数据反演方法研究：（1）通过在反演问题模型中构建约束条件，差分演化算法能在求解过程中自动利用各种地质信息和先验知识，从而改善反演的多解性问题；（2）差分演化算法的实数编码方案，能有效的处理参数的范围，使反演的结果对于参数个数和维数不敏感，适应大地电磁反演参数越来越多的发展趋势；（3）对Map-Reduce 模型进行适应性修改，提出并实现基于云计算的差分演化算法，解决反演计算的昂贵性问题。通过这种具有智能性的大地电磁测深反演方法，将为大地电磁测深法的大规模实际应用奠定基础。

大地电磁测深法作为当今地球深部电性结构探测和大陆动力学研究的重要地球物理方法之一，在深部能源探测、减少自然灾害、工程应用等方面发挥着越来越重要的作用。然而，常规的反演方法存在解的非唯一性和计算昂贵性等困难，越来越难以满足现在的地下情况更复杂、处理标准更高的要求。本项目将差分演化算法引入到大地电磁测深中，进行大地电磁测深数据反演方法研究：（1）通过在反演问题模型中构建约束条件，差分演化算法能在求解过程中自动利用各种地质信息和先验知识，从而改善反演的多解性问题；（2）差分演化算法的实数编码方案，能有效的处理参数的范围，使反演的结果对于参数个数和维数不敏感，适应大地电磁反演参数越来越多的发展趋势；（3）对Map-Reduce 模型进行适应性修改，提出并实现基于云计算的差分演化算法，解决反演计算的昂贵性问题。通过这种具有智能性的大地电磁测深反演方法，将为大地电磁测深法的大规模实际应用奠定基础。