Advanced Machine Learning Methods for Accurate Prediction of Loss Circulation in Drilling Well Log
Saeed Beheshtian, Sara Kishan Roodbari, Hamzeh Ghorbani, Mohamadreza Azodinia, Mohamed Mudabbir, Annamária R. Várkonyi-Kóczy
Abstract
Drilling mud (DM) plays an important role for exploration of oil and gas reservoirs. Lost circulation (LC) has highly influenced drilling costs due to the waste and loss of the drilling fluid (DF), which stems from some variables such as mud composition, permeability, and mud pressure. To predict this important parameter, a dataset includes 1003 data points was collected from two oil field wells in southern Iran. For the prediction of LC in this article, three hybrid algorithms were used, combining three optimization algorithms with one network algorithm, including Social Ski-Driver (SSD), Particle Swarm Optimization (PSO), and Genetic Algorithm (GA) with the Multilayer Perceptron Network (MLP). Analysis shows the combination of the MLP with SSD give good and higher accuracy than other algorithms. The statical error meteoric for the test dataset, as indicated by an RMSE of 30.85 and an impressive R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.94, show their results are better and over than other competing algorithms.