Effect of Rock Properties on ROP Modeling Using Statistical and Intelligent Methods: A Case Study of an Oil Well in Southwest of Iran

Abstract

Rate of penetration (ROP) is one of the key indicators of drilling operation performance. The estimation of ROP in drilling engineering is very important in terms of more accurate assessment of drilling time which affects operation costs. Hence, estimation of a ROP model using operational and environmental parameters is crucial. For this purpose, firstly physical and mechanical properties of rock were derived from well logs. Correlation between the pair data were determined to find influential parameters on ROP. A new ROP model has been developed in one of the Azadegan oil field wells in southwest of Iran. The model has been simulated using Multiple Nonlinear Regression (MNR) and Artificial Neural Network (ANN). By adding the rock properties, the estimation of the models were precisely improved. The results of simulation using MNR and ANN methods showed correlation coefficients of 0.62 and 0.87, respectively. It was concluded that the performance of ANN model in ROP prediction is fairly better than MNR method. Keywords: ROP, rock properties, MNR, ANNPrędkość wiercenia jest jednym z podstawowych parametrów charakteryzujących tempo prac wiertniczych. Oszacowanie prędkości wiercenia jest zagadnieniem kluczowym dla inżynierów wiertnictwa, gdyż pozwala na dokładne określenie czasu trwania prac, a co za tym idzie także kosztów operacyjnych. Szacowanie prędkości wiercenia odbywa się na podstawie modelu uwzględniającego parametry pracy oraz parametry środowiskowe. Pierwszy krok obejmuje pozyskanie danych o fizycznych i mechanicznych właściwościach skał na podstawie profilowania geofizycznego otworu. Zastosowano korelację odpowiednich par danych dla pokreślenie wpływu głównych czynników warunkujących prędkość wiercenia. Nowy model obliczania prędkości wiercenia opracowany został w okręgu naftowym Azadegan w południo -wo-zachodniej części Iranu. Symulacje prowadzono w oparciu o metodę wielokrotnej regresji nieliniowej a także przy wykorzystaniu sztucznych sieci neuronowych. Poprzez dodanie danych o właściwościach

A review on half a century of experience in rate of penetration management: Application of analytical, semi-analytical and empirical models

“…Using traditional ROP models is a conventional approach in ROP management studies (Bezminabadi et al, 2017). Either as an assistant model for artificial intelligence algorithms or as a basic ROP management tool, they have been used repeatedly (Bahari and Baradaran Seyed, 2007a;Rastegar et al, 2008;Bataee et al, 2010;Sui et al, 2013;Formighieri and Freitas, 2015;Kutas et al, 2015;Soares et al, 2016;Hegde et al, 2019).…”

“…Besides of the studies which have used the analytical and semi-analytical ROP models for the purpose of ROP management, lots of studies used their own ROP models for ROP management to test their applications (Galle and Woods, 1963;Eckel, 1967;Bourgoyne and Young, 1974;Peterson, 1976;Duklet and Bates, 1980;Warren, 1981Warren, , 1987Chia and Smith, 1986;Walker et al, 1986;Winters et al, 1987;Wojtanowicz and Kuru, 1987;Maidla and Ohara, 1991;Hareland and Hoberock, 1993;Hareland and Rampersad, 1994;Caicedo et al, 2005;Hareland and Nygaard, 2007;Osgouei andÖzbayoǧlu, 2007;Shirkavand et al, 2009;Hareland et al, 2010;Motahhari et al, 2010;Seifabad and Ehteshami, 2013;Kerkar et al, 2014;Bezminabadi et al, 2017;Wiktorski et al, 2017;Al-AbdulJabbar et al, 2019;Elkatatny, 2019;Al-AbdulJabbar et al, 2020;Etesami et al, 2021). Table 6 shows a brief review on research studies which have used different ROP models and empirical correlations as the main approach for ROP management.…”

“…Later, Seifabad and Ehteshami (2013) developed an ROP correlation based on geological data. The parameters considered in this correlation are WOB, bit rotational speed, bit torque, mud weight, mud viscosity, bit condition and depth interval from the beginning to the end of drilling Bezminabadi et al (2017). developed two correlations for prediction of ROP in different modes with or without considering rock mechanical properties.…”

A stacked generalization ensemble model for optimization and prediction of the gas well rate of penetration: a case study in Xinjiang

“…Artificial neural networks (ANNs) are widely used in ROP prediction and have good performance (Arabjamaloei and Shadizadeh 2011;Arabjamaloei and Karimi Dehkordi 2012;Kahraman 2016;Bezminabadi et al 2017;Anemangely et al 2018;Elkatatny 2018;Abbas et al 2019b, a;Sabah et al 2019;Ashrafi et al 2019;Diaz et al 2019;Elkatatny et al 2020;Zhao et al 2020;Qian et al 2021). To improve the performance of ANNs, some algorithms are used to optimize ANN models.…”