Research And Implementation Of Automatic Layer Identification And Comparison

Using well logging curve to divide strata, is the main layered approach of loggingcompanies or research institutes currently. Layering is a fundamental step, we can notcontinue study the multi-well and reservoir parameters, unless this step has been done.According to the division of the thickness, different Wells can be divided into multiple levels,this paper mainly studies a small sand layer and group level. Layer division is not an easy task,which requires a lot of effort, if the geology is complex will more easily to get wrong. Almostall of the current domestic companies draw small layers manually, first, because therequirements are different for each company, and second, there is little commercial automatedtiered software.With the development of computer technology, more and more domestic andforeign people studied automatic stratification. Although the proposed introduction of adomestic or foreign ways, but still a long way for practical application. Geometry classstratification method can simulate some extent people in a hierarchical manner, so that theresults tend to requirements specification, but it is difficult to adapt to complex curves;Statistics class methods, having precise boundaries, but only adjust the thickness, clustering isnot easy to achieve difficult artificial layered effect. Common drawback is more than one timeon a visual level division, not as artificial as to grasp the relationship between the whole andthe parts.The purpose of this paper is comprehensive advantages of two methods, and researchingon the global and local issue. Based on the current popular technology in-depth study andcontrast, my research is on the automation of stratification of a oil field project. Writesoftware of manual stratification, while adding an automatic hierarchical system usingautomatic division plus manual correction mode. The aim is to further reduce the workload ofmanual operation, but also to ensure quality.Hand-layered system parts, based on personal experience and others layered proposalwriting. Try to make it fast, mass, use machines instead of manpower more, save more time.This part consults some of the current popular layered software. Which is consistent with data formats of laboratory other software, fault tolerance, and can be custom strong, the level ofcode is clear and logical, and can be reusability well, ease of future expansion.Automatic layered system parts, using the reference standard wells mode, featureextraction can be automatic done. Using statistical methods to find and fix the scope of thereservoir. Find the most obvious characteristics of sand group place to determine theboundaries of the layers, and using logic inference method to simulate artificial reservoircorrection and fill the other layers. The whole process starting from the perspective of theglobal for many times, use sequential comparison. To some extent overcome the local andglobal conflicts. Stratification results in experiments with more than ten wells, cross layer rateis low, the boundary discrepancies with the hand, but can satisfy the precision required by thecompany. Influenced by standard well, instead of the similar layered, the better the results, theopposite more mistakes. Manual correction is much less time than all hand layered, can savemanpower.Small layer of partitioning method, are to find each basis in the literature, and the needfor appropriate improvements are according to the actual situation. According to thecharacteristics of the region such as curve data, presented special methods and multipleprocessing. Found most easily defined layer priority on the floor, then deal near can wellavoid stratification. Only with the statistical method of layered to look for reservoir and definethe boundary, not fully layered, not only can use the method of accurate advantage, but alsowill not affect the logic behind, does not destroy the holistic. Because looking for reservoirmain requirements correctly, so use large window activity method, enhanced anti-jammingcapability requirements. After repeated comparison and logical inference, cut and fill andborder correction, and finally reach the results of number and location are reasonable.The combination of the two methods, not only keeps statistics layered precision, but alsohas the simulation of a layered geometry. If you can do well in the logic module, the resultswould be much better than former basic mathematical clustering. Using the former to find oilreservoir, and then reasonable correction, is simpler than using only the latter for complexcurve. Although the generality is weak, there are enough wells in one block. If we keep onin-depth study and improve the versatility, it will be of great practical value.