Optimizing Drilling Fluid Properties Using Deep Learning Algorithms to Reduce Drilling Problems in a Middle Eastern Oil Field

Document Type : Original Article

Authors

1 Pars Drilling Fluids Co.

2 Department of Mining and Petroleum Engineering, Imam Khomeini International University

3 Kavoshgaran Consulting Engineers

10.22077/jgm.2024.7709.1028

Abstract

Drilling fluid is among the most important requirements for drilling oil, gas, and geothermal wells. Many drilling challenges are directly or indirectly related to the drilling fluids; therefore, optimizing the drilling fluid has a significant effect on the quality of drilled wellbore, and the risk of drilling operations. In this paper, it is tried to develop deep learning algorithms to optimize drilling fluid properties to minimize the possibility of occurrence of possible problems in one target field in the Middle East. This paper deals with the method of artificial intelligence for the first time to investigate the possibility of estimating optimum drilling fluid parameters using drilling and geological parameters to minimize problems- without considering the location of the target wells. Two artificial intelligence algorithms ‘’LSSVM’’ and ‘’MLP-FFBP’’ were used to train the machine to optimize drilling fluids (such as mud density, yield point, plastic viscosity, etc.) to minimize drilling fluid challenges such as stuck pipe, tight hole, formation influx, and even loss circulation. Results showed that for optimizing drilling fluid parameters in a newly drilled well, the developed AI networks have good capability to estimate parameters for some drilling fluid parameters such as mud density, plastic viscosity, water percentage, and API filtration properties with the accuracy of more than 95% for train and more than 85% for test data. Moreover, results showed that drilling fluids have a direct effect on the tight holes and stuck incidents.

Keywords

Journal of Geomine
Volume 2, Issue 1
March 2024
Pages 1-12

Files

Share

How to cite

Statistics