| When we will identify lithology, divided reservoir and recognition well phase of oil, gas and water by well information, the first work is to divide layers by well logs. So divideding well logs accurately is very important to exploration and development of oil and gas. Long-standing, well workers divided layers by hand by the differences and changes between well logs and surrounding layers. This original and manual method may cause some differences of results because of using different well logs and standard by different well workers. And it doesn't match the request of development of process technology by computer. So we raise the technology of automatic segmentation of well logs by computer.The Hailar Basin, in northern part of the Inner Mongolia Autonomous Region and to the west side of the Da Hinggan mountains, was a Late Mesozoic-Paleogene basin, with the basement being Hercynides. The boundary faults to control the basin developing of NE- and NNE-strikings created. Only Urxun depression displays SN distributing, it created in Early Cretaceous, and overprinted by later deformations. Urxun depression is controlled by the fault of the west of Urxun, it is a large unsymmetrical fault-depression sag with strike-slip characteristics in Mesozoic-Cenozoic in Hailaer Basin, the area is 2350 km 2, the depression's pose and develop is mainly controlled by NE and NW fault. The north of the basin is mainly controlled by NNE Urxun west fault, in north the east boundary fault is a little small, but its reverset is distinct, and growth time is early, the middle of sediment leans to a side of Urxun west fault. The geological structure in the south of the Urxun is different from the north of the Urxun, NW fault is its main boundary fault in the west side, is also named Urxun west fault south section, and its reverse is distinct. The fault of the east fault is not grown, and is also a large unsymmetrical fault-depression sag with strike-slip characteristic, we could divide Urxun depression into Urxun north fault, Urxun middle fault, Urxun sorth fault.The difference of diagenesis of Urxun sag is big. The category of rock is complex. There are a lot of thin layers. Some borehole wall collapses. These make vertical resolution of most of the logging methods are distorted. So there are some difficulties in doing automatic segmentation of well logs. Firstly, this paper use intraformation discrepance method and clustering analysis method to do automatic segmentation of well logs of Damoguaihe formation, Nantun formation and Tongbomiao formation in Hailaer Basin Urxun sag. First, we normalize some well logs. We choose GR active curve to stratify fine. We suppose well average value is stable in the same layer. The change of well numerical value in the same layer dose not exceed the allowed error. This allowed error is decided by experiment. If the difference between the mumerical value of adjacent point and the average value less than the allowed error, we make this point belong to this layer. Otherwise, the point is belong to the next layer. Then we will apply the result on SP, LLD and DT curve.Above-mentioned intraformation discrepance method may stratify too fine. We may make the same layer into some different tiny layer. For Overcomeing above-mentioned shortcomings, we can do automatic segmentation reasonably by uniting clustering analysis method. Then we calculate Mahalanobis-distance between every adjacent tiny layer. We decide the smallest critical distance d min.If Mahalanobis-distance d ( k , k+1 )≤dmin, we will merge layers. If Mahalanobis-distance d ( k , k+1 ) > dmin, we will think that the result is true. We program by MATLAB. Then we can get the result. Last we divide the points into three parts. The average value of the middle part is well numerical value of this layer.In addition, we do automatic segmentation of well logs of Damoguaihe formation, Nantun formation and Tongbomiao formation in Hailaer Basin Urxun sag by wavelet transform. This paper choose Guas basic wavelet. We do wavelet transform to SP, GR and LLD alone. Then we look for extreme value of every scale. The boundary must satisfy following: (1)they must be the corresponding points to extreme value of the minimum scale. (2)they must approach the limen what is offered by us. (3)the extreme value must be bigger than a limen. (4) the biggest value of extreme value of all scales must be bigger than a limen. (5) they must not be adjacent maximum or minimum. In that case, we can choose the point with the bigger atsolute value. We program by MATLAB. Then we can get the result. Last we divide the points into three parts. The average value of the middle part is well numerical value of this layer.We use kaben software to display the result. It could contrast with the lithology data of geology. At last we can summarize according rate. The according rate of automatic segmentation of well logs of Hailaer Basin Urxun sag is got by layer in statistical method. The result is: the according rate of intraformation discrepance method and clustering analysis method is 90%. The according rate of SP with wavelet transform is 72%; the according rate of GR is 87%; the according rate of LLD is 81%.Intraformation discrepance method and clustering analysis method can divide layer exactly. This method can sum up the characteristic of many Curves. It is overall. But the calculation of this method is very large. So its speed is slower; When we use wavelet transform for automatic segmentation, we can find that there are a lot of extreme value in the curves with wavelet transform. But generally there is only one boundary in the rising part of one curve. So after getting extreme value, we must make policy. This is reasonable. But this method increase more manual interfere than others. Moreover this method is fit for alone curve. It can not summarize the characteristic of many curves. So it lowers according rate.We can find that the according rate of intraformation discrepance method and clustering analysis is higher. We advice using this method to divide layer in Hailaer Basin Urxun sag.
|
PDF Full Text Request