基于支持向量机的时深建模方法

TIME-DEPTH MODELING METHOD BASED ON SVM

为了解决海上勘探初期地震资料解释过程中的时深关系问题,提出基于粒子群和支持向量机的时深建模方法,利用粒子群算法来优化支持向量机的参数。首先由合成地震记录标定得到每口井的时深关系;然后利用支持向量机方法建立时间与深度之间的关系模型;最后以此模型对时间域等T0图进行转换,从而得到深度域的构造图。实际测井数据和三维地震资料试验结果表明,该方法建立的时深关系模型适用于整个研究区,且时间域转换深度与井分层深度、构造深度之间的误差较小,能够满足精细构造解释的精度要求。

In order to solve the problem of time-depth relationship during the seismic data interpretation for the initial offshore exploration,the time-depth modeling method was proposed based on the particle swarm and support vector machines（ SVM）,and moreover the SVM parameters were optimized by Particle Swarm Optimization（ PSO）. Firstly the time-depth relationships for each well were obtained by the synthetic seismogram calibration;and then the relationship model between the time and depth was established with the help of SVM method; finally with the help of the model,the conversion was conducted for the T0 contour in the time domain,then the structural map for the depth domain was obtained. The test results of the actual logging data and 3 D seismic data show that the time-depth relationship model built by the method is suitable to the whole study area,and moreover the errors are pretty small between the inverted depth in the time domain and the well separated depth and the structural depth. In a word,the model can meet the precision requirements of the fine structure interpretation.