Application of Non‐linear Discriminate Function in Oil‐gas Prediction
Yingjun Feng,Xuefeng Wang,Yanlai Li,Shi Chen
DOI: https://doi.org/10.1190/1.1821277
1999-01-01
Abstract:Summary In this paper, the method of non-linear oil-gas-bearing prediction in thin reservoir is presented. This method has the advantag e of easy programming and it can be efficiently incorporated with the known geologic constraints available for thin reservoir. We tested this new approach for oil-gas prediction on several field data cases. To tackle this problem, in this expanded abstract, we propose a new method. This new approach uses the non-linear discriminate decision function based on the seismic attribute parameters. To test this new approach, the non-linear algorithm i s applied to oil-gas prediction in Dongsan Oil Field, HaiNan Oil Field and LiaoHe Oil Field, China. The tests shown reasonable feasibility of the new method. Theory and algorithm Based on the properties of dynamics and kinematics in seismic waves, as well as the reservoir characterizations, more than ten attribute parameters can be acquired from reservoir data and each single seismic trace. It is impossible to apply an oil-gas- bearing prediction for the reservoir by using these parameters directly, because the computation is prohibitively costly, since the space of characteristics is too large. Moreover, some of these parameters may be correlated, and some information may be hypercatalectic. During the information extraction, the following rules must be followed: 1. The entropy must be preserved; 2. Auto-correlation must be eliminated; 3. Energy must be conserved, and 4. Energy must be redistributed and concentrated. Based on these rules, we select seven seismic attribute parameters among nearly forty original parameters. All these seven parameters are calculated in a predefined time window. From now on, we suppose that the time window is the interval, (t i, ti+1), within the duration of the seismic wavelet. Before the computation, we partition the time window into three sub-intervals, (t i, ti 1 ), (ti 1 , ti 2 ) and (ti 2 , ti+1), where ti 1 -ti= 0.25(ti+1-ti), t i 2 -ti 1 = 0.25(ti+1-ti), t i+1-ti 2 = 0.5(ti+1-ti). We calculate the attribute parameters separately in each interval and sum them up with different weights. The seven parameters to be used include: 1. Energy parameter of seismic wave