{"title":"Hybridized Probabilistic Machine Learning Ranking System for Lithological Identification in Geothermal Resources","authors":"P. Ekeopara, J. Odo, B. Obah, Valerian Nwankwo","doi":"10.2118/212015-ms","DOIUrl":null,"url":null,"abstract":"\n Geothermal resources are characterized by hard rocks with very high temperatures making it difficult to implement conventional tools for petrophysical analysis such as lithological identification. Several computation and artificial intelligence models such as K-means clustering algorithms have been applied, however, these algorithms are limited to certain applications due to the available data utilized and high computation time. It is hence pertinent to consider a robust model that can meet up with these requirements.\n In this study, a proposed hybrid machine learning probabilistic ranking system was developed which considered the integration of several pattern recognition algorithms in the identification of formation lithology. The ranking system leverages on the large volume of drilling and log data collected from conventional oil and gas operation to develop five embedded lithology identification models: K-means clustering, Hierarchical clustering using ward linkage, K-mode clustering, Birch, Mini-batch kmeans. The analysis was carried out using gamma ray logs, density logs, neutron porosity logs and Spontaneous potential as input parameters in building the lithology identification models while rate of penetration, surface RPM, Flow in, surface torque and pump pressure were utilized to predict the different lithologies using the different pattern recognition models as outputs. The output derived from the respective lithology identification models are further ranked based on a probabilistic approach to predict the actual lithology of the encountered formation. The results show that the implementation of the ranking system was effective in identifying the lithology of the drilled formation.","PeriodicalId":399294,"journal":{"name":"Day 2 Tue, August 02, 2022","volume":"47 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Day 2 Tue, August 02, 2022","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.2118/212015-ms","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Geothermal resources are characterized by hard rocks with very high temperatures making it difficult to implement conventional tools for petrophysical analysis such as lithological identification. Several computation and artificial intelligence models such as K-means clustering algorithms have been applied, however, these algorithms are limited to certain applications due to the available data utilized and high computation time. It is hence pertinent to consider a robust model that can meet up with these requirements.
In this study, a proposed hybrid machine learning probabilistic ranking system was developed which considered the integration of several pattern recognition algorithms in the identification of formation lithology. The ranking system leverages on the large volume of drilling and log data collected from conventional oil and gas operation to develop five embedded lithology identification models: K-means clustering, Hierarchical clustering using ward linkage, K-mode clustering, Birch, Mini-batch kmeans. The analysis was carried out using gamma ray logs, density logs, neutron porosity logs and Spontaneous potential as input parameters in building the lithology identification models while rate of penetration, surface RPM, Flow in, surface torque and pump pressure were utilized to predict the different lithologies using the different pattern recognition models as outputs. The output derived from the respective lithology identification models are further ranked based on a probabilistic approach to predict the actual lithology of the encountered formation. The results show that the implementation of the ranking system was effective in identifying the lithology of the drilled formation.