Technical Method And Application Of Stratigraphic Correlation Intelligently

This paper is based on well-log information to try to set up a proper and scientificautomatic process of stratigraphic correlation by the method of data processing andmathematical calculation, and it is hoped that this process could be high efficiency andaccuracy.Accordding to the summary of well-log information, SP, GR and AC logging data arechosed to carry out automatic stratigraphic correlation. First, by using the method ofanticotangent normalization, the different kinds of logging data are transformed intonondimensional data, then verify the result. Next, the logging data is transformed by themethod of wavelet analysis and walsh filtering, the filtering result of wavelet analysis isless harmful to the original data, but it can not show the lithologic interfaces clearly, andthe result of walsh filtering is harmful, but it can show the lithologic interfaces clearly, so,walsh filtering is chosen. Then, area integral, curvilinear integral and Gaussian model areseparately used to implement automatic strata division, because the logging data is similarand obeys a gaussian distribution, the method of Gaussian model can get the better ruslt.At last, the logging data, thickness of single layer and rhythm are extracted as parametersof reservoir, next, by using PNN neural network, a reliable map could be set up fromlogging data to the result of automatic stratigraphic correlation. For the reason of lost andrepeat stratum, the criticality similarity is set to85percent, when the similarity of theoutput vector and the layer is equal or greater than85percent, the output vector can beincluded in this layer and the final result is1, otherwise, the output vector is missing andthe final result is0.During the correlation process, data filtering、mathematical and fuzzy method andPNN neural network are fully integrated into it, finally, the output numeric results fromPNN neural network must be transformed into imaging result. All of the steps areperformed in the MATLAB software to ensure the high run speed and precision, in fact, ifthere are lots of samples sdutied by the PNN neural network, the result could have a highaccuracy of diagnosis, and the map could be used in many cases.