Research In Multi-scale Wavelet Analysis On The Application Of Well Logging Curves Automatic Stratification

Stratigraphic division is indispensable for oil-gas exploration, which makesresearch on automatic division technology is extremely meaningful. So far, there areabout six kinds of automatic segmentation methods of well logs: activity-variancemethod, clustering-sediment curve method, maximum entropy spectral analysis,Walsh transform automatic layering, optimum partitioning method, and wavelettransform method for automatic division of log strata. Most of those methods can onlybe used in division of a certain level stratigraphic units instead of a whole vision level.The formation of stratum in the deposition process is influenced by a variety ofcomplex factors, thus we need to employ various methods to figure out the sequencestratigraghic unit interfaces. Multi-scale wavelet analysis can make thetime-frequency localization of signals more clear, can better simulate the artificialinterpretation of “from rough to accurate, layering step by step”, and can reveal theinternal geological phenomena during the geological process in a microscopicperspective. The focus of this thesis is to do research on multi-scale wavelet analysisfor automatic division of well logging curves which makes a good match to the stratacharacteristics in order to identify different interface accurately.After studying and analyzing the theory of wavelet analysis and automaticlayering methods of well logging curves, this thesis made the following tasks areaimed in this thesis:(1) Summarizing various automatic layering methods, combiningedge detection and optimum partitioning, constructing the improved optimumpartitioning method.(2) Based on the features of logging data and theory of waveletanalysis, discussing the feasibility for the application of wavelet transform on well logging analysis, then constructing a technology of multi-scale wavelet analysis onautomatic layering of logging curves.(3) Combined with the characteristics oflogging signals, figuring out the suitable wavelet functions and waveletdecomposition levels for different logging curves.In this thesis, logging data of X5and X11well from Xinchang gas field ofDeyang in Sichuan province are chosen, and analysis on AC, GR, SP curves toidentify strata interfaces is done by using wavelet. Based on core data, we draw theidentification results on phase diagram of these two wells by carbon software,compare the results by using improved optimum partitioning method. Thecomparative result shows that multi-scale wavelet analysis can be more accurate onstrata layering than optimum partitioning, as it does not only reflect the lithology ofsmall differences in interface, but also be more conducive to identify the details ofstrata. Thus, it is shown by the results of this thesis, that the technology of multi-scalewavelet analysis on automatic layering of logging curves can better reflect theinterfaces when has lithology change, and that the test results shows a higher rate onmatching. These mean that this technology can support prominent theory and ispractically significant for the future research.