【摘要】 地震油气预测中的不确定性因素包括地震属性选取、预测算法选择、环境噪声及原始数据观测误差等。为消除这些不确定性因素,本文利用支持向量机与信息融合理论进行地震油气预测,支持向量机首先通过利用内积函数定义的非线性变换将输入空间变换到一个高维空间,在这个空间中求(广义)最优分类面,其分类函数形式上类似于一个神经网络,输出是中间节点的线性组合,每个中间节点对应一个支持向量。支持向量机可以解决分类问题和拟合问题,在解决小样本、非线性及高维模式识别问题中表现出许多特有的优势。信息融合是利用时间、空间的多传感器信息资源,采用数学方法和计算机技术对观测信息在一定准则下加以自动分析、综合和使用,从而比单一传感器观测对象获得更优越的一致性信息和描述,减小环境对决策的影响。将支持向量机与信息融合两者结合应用,能同时减小多种因素引起的不确定性,提高油气预测精度。此方法用于实际数据,得到了较好的预测结果。更多还原

【Abstract】 Undefined factors during the prediction of oil and gas by seismic exploration include selection of seismic attributes, selection of predicting algorithm, environmental noises and surveying errors of raw data. In order to eliminate these undefined factors,the paper utilized supporting vector computer and information-merging theory to predict oil and gas by using seismic exploration,supporting vector computer first transforms the input space into high-dimensional spaces by using non-linear transformation defined in inner product and solve the generalized optimal classified plane, the classified function is formally similar to a neural networks, the output is linear composition of intermediate nodes and each node corresponds to a supporting vector. The supporting vector computer can solve the classified problem and fitting problem,which is characterized by distinctive superiority in solving the small sample, non-linear and high-dimensional pattern recognition problems. Information merging is to use information source of multiple sensors in time and space, adopt mathematical methods and computer technology to automatically analyze,integrate and use the surveyed information under certain criterion,so that can obtain more optimal consistent information and description in comparison with surveyed object with single sensor and reduce the influence of environment on policy decision. Combining application of supporting vector computer with information merging can simultaneously reduce the uncertainty caused by multi-factors and improve oil/gas-predicted precision. Application of the method to real data obtained better prediction results.更多还原