Identification and Application of Detrital Diagenetic Facies Logging Based on Unsupervised Xi Technology: A Case Study of the Mesozoic in Chengdao-Zhuanghai Area

The Mesozoic in Chengdao-Zhuanghai area is affected by complex tectonic evolution, diverse sedimentary types and lithology, and the reservoir heterogeneity is extremely strong, and the prediction of reservoir quality is difficult, and the accurate identification and division of lithofacies types plays a crucial role in the classification and evaluation of reservoirs. In the well section with relatively few corings, four logging curves sensitive to diagenesis, GR, AC, DEN, and RD were selected as the basis for diagenetic facies division, and the diagenetic facies division was carried out by the method of machine Xi. The traditional machine Xi is divided into two Xi: supervised Xi and unsupervised, in which supervised Xi requires a large number of Xi samples to ensure its accuracy, and unsupervised Xi does not need to learn Xi samples, but the classification results may not be the expected classification type. Combined with the characteristics of strong heterogeneity, relatively few coring sections and limited results of unsupervised Xi in this area, the method of unsupervised Xi with single factor constraint was considered to identify and divide the logging facies of the three formations in the Chengdao-Zhuhai area. Combined with the geological data such as core, cast thin section identification, logging data, etc., the calibration of logging facies and diagenetic facies is realized, so as to complete the identification and division of regional diagenetic facies. Finally, the accuracy of the Xi method is verified by comparing the thin section identification results, which provides a basis for the identification of reservoir diagenetic facies in the lack of coring well sections.

Keywords: clastic rocks; Chengdao-Zhuanghai area; The Mesozoic;Diagenetic facies logging identification; Univariate constrained unsupervised learning