摘要
可控源电磁反演利用人工信号获取地下结构信息,为地质勘探和资源开发提供准确的数据支持。然而,传统电磁反演方法面临低分辨率的挑战,主要是由于简化处理和观测数据的限制,导致模型平滑、细节丢失,从而削弱了反演的准确性,影响了电磁勘探的效率。为了解决这一问题,本文提出将传统反演结果与响应数据同时作为深度网络的输入数据,为深度网络反演提供先验物理信息,并结合深度学习算法提升可控源电磁反演的计算效率。通过模型试验,对合成的电阻率模型分别进行了传统反演、智能反演以及融合先验物理信息的智能反演。结果表明,融合先验物理信息的智能反演能够更好地刻画异常体结构特征,有效提升反演效率,并且得到的电阻率参数更符合实际。最后,将该反演技术应用于金川铜镍矿床的可控源数据反演解释,取得了较为可靠的应用效果。
Controllable source electromagnetic inversion utilizes artificial signals to obtain underground structural information,witch can provide accurate data support for geological exploration and resource development.However,traditional electromagnetic inversion methods face the challenge of low resolution,mainly due to the limitations of simplified processing and observation data.As a result,traditional model smoothness and loss of details have weakened the accuracy of inversion and affected the efficiency of electromagnetic exploration.To address this issue,this work proposed to use both traditional inversion results and response data as input data for deep network inversion,providing prior physical information for deep network inversion.Combined with deep learning algorithms,the computational efficiency of controllable source electromagnetic inversion was improved.Through model experiments,traditional inversion,intelligent inversion,and intelligent inversion integrating prior physical information were performed on the synthesized resistivity model.The results indicate that intelligent inversion based on prior physical information can better characterize the structural characteristics of anomalous bodies,effectively improve inversion efficiency,and obtain resistivity parameters that are more in line with reality.Finally,the inversion technique was applied to the controllable source data inversion interpretation of the Jinchuan copper-nickel deposit,and achieved relatively reliable application results.
作者
李雄
罗伟奇
金小燕
傅群和
毛寅
金妮
肖青
贾卓
LI Xiong;LUO Weiqi;JIN Xiaoyan;FU Qunhe;MAO Yan;JIN Ni;XIAO Qing;JIA Zhuo(Hunan Center of Natural Resources Affairs,Changsha,Hunan 410118;School of Civil Engineering,ChangshaUniversity of Science and Technology,Changsha,Hunan 410015)
出处
《地质与勘探》
CAS
CSCD
北大核心
2024年第4期800-808,共9页
Geology and Exploration
基金
国家自然科学基金国家重大科研仪器研制项目(部门推荐)(编号:42327901)
湖南省自然科学基金(编号:2023JJ40223)联合资助
关键词
可控源电磁反演
深度学习
先验物理信息
计算效率
电阻率参数
金川铜镍矿床
controllable source electromagnetic inversion
deep learning
prior physical information
computational efficiency
resistivity parameters
Jinchuan copper-nickel deposit
