Appling Gaussian Process Regression to Predict
Drilling Rate of Penetration with
Hyperparameter Optimization using Bayes
algorithm: A case study of FORGE site
⋆
Kanghui Zeng
1[00009−0002−2449−8379]
, Second Author
2,3[1111−2222−3333−4444]
,
and Third Author
3[2222−−3333−4444−5555]
1
China University of Geoscience
2
Springer Heidelberg, Tiergartenstr. 17, 69121 Heidelberg, Germany
zkh@cug.edu.cn
http://www.springer.com/gp/computer-science/lncs
3
ABC Institute, Rupert-Karls-University Heidelberg, Heidelberg, Germany
{abc,lncs}@uni-heidelberg.de
Abstract. The prediction of rate of penetration (ROP) is a key tech-
nology to optimize the drilling process and improve drilling efficiency.
The existing calculation models are mainly based on physical experi-
ments and theoretical analysis. The lack of application of measured data
in drilling engineering makes the calculation accuracy difficult to meet
the complexity on-site demand. In this study, we have used a public data
set from Utah FORGE geothermal wells project. After cleaning and pre-
processing the data, we identified 7 relevant characteristic parameters
based on Spearman correlation coefficients to streamline the model and
enhance efficiency. Subsequently, we selected 4 models and optimized
their hyperparameters. Our results indicate that the Gaussian process
regression (GPR) model, optimized using the Bayesian algorithm, out-
performed the other models, achieving an R
2
value of 0.969, RMSE of
3.572, and MAE of 1.094 in predicting the ROP behavior in our field.
Keywords: rate of penetration (ROP) · predictive modeling · machine
learning · Gaussian process regression (GPR) · Random Forest
1 Introduction
With the research and development of integrated drilling technology in geological
engineering, the drilling rate equations obtained based on single-factor analysis
in the early stage are difficult to meet the application requirements of massive
actual drilling data on-site, severely affecting the adjustment and optimization
of safe and efficient integrated geological drilling schemes
[1]
. There is an urgent
need to establish new drilling rate equations. In recent years, with the application
⋆
Supported by organization x.
