Rate of Penetration (ROP) Prediction Using Artificial Neural Network for Nearby Well in a Geothermal Field

Abstract

With heterogeneous formation, it is necessary to find the correlation between varied and complex parameters with the rate of penetration (ROP). More data sets need to be taken in order to predict accurately and properly. On the other hand, procuring newer data sets is economically and technically arduous. Hence, a method called as Data Driven Modelling has been developed in order to utilize previous obtained data for predicting ROP. Lately, the industry has been shifting to the development toward easier way of analyzing data by setting up a machine learning algorithm, a computational system for identifying and classifying data. In this case, supported by assortments of drilling surface measured input data parameters, Artificial Neural Network (ANN) is promoted to predict the ROP. Setting up the input-output mapping with interconnected neural, industry is capable of accurately forecasting the output data. ANN conduct training cycle until error for data validation has been derived. This cycle will generate the relation between input and output. The output of ANN will be more accurate and has been theoretically proven by contemplating more than one parameter to obtain one set of goal. All the surface measured input data parameters including mud rheology and bit data will be expected to predict the ROP. This paper discusses the first stage of the present study conducted by the authors in applying ANN for predicting drilling ROP in a geothermal field in Indonesia, which includes literature review on previous models and correlation analysis between ROP and surface measured input data, such as weight-on-bit (WOB), mud density, lithology, rotation-per minute (RPM), etc.