PERBANDINGAN METODE MACHINE LEARNING DAN STUCK PIPE RISK (SPR) UNTUK MEMPREDIKSI PIPA TERJEPIT PADA LAPANGAN PANAS BUMI

Abstract

This research is about "Comparison of Machine Learning and Stuck Pipe Risk (SPR) Method to Predict Stuck Pipe in Geothermal Field" with the aim is to determine the best machine learning classification based on the calculation of the precision, recall, and F1 Score on testing with Y3 well, determining which have the best alert time interval between SPR method and machine learning in predicting the stuck pipe of Y3 well, determining the most affecting parameters in predicting and modeling the stuck pipe incident. The first method used is SPR analysis, with quantitative data processing using Microsoft Excel software. The second method is machine learning modeling using Python software, where data from 9 geothermal wells is trained and produces a model that can predict Y3 well data. The results show that the best model based on the F1 Score is logistic regression with a score of nonstuck 0.99 and stuck 0.89 (maximum 1). Alert time prediction results with machine learning (logistic regression) is better than the SPR method with the nearest alert having an interval of +- 2 hours before the stuck pipe incident.