An artificial neural network to predict bit walk rate for drilling in the Gulf of Thailand

Abstract

Petroleum well drilling requires inclination and direction control of the bit to assure that the reservoir target is met within an acceptable range. General drilling operation uses Adjustable Gauge Stabilizer (AGS) to control inclination while using steerable motor to control direction. The reservoir located at the deep section of the well in the Gulf of Thailand exceptionally contains high temperature where rubber part of the directional control tool is not able to effectively handle. As a result, AGS is solely used in the operation to control inclination without a directional control tool in place. This paper is then attempting to address and model the bit directional behavior with the availably controllable parameters by using an Artificial Neural Network (ANN). The modeling is carried out by using the field data from the Gulf of Thailand and inputting into the ANN to create a bit directional deviation, referred in this paper as bit walk, predictive tool. The study is divided into two perspectives of predicting bit walk direction and quantity. The result shows that the ANN can be effectively used to predict bit walk quantity as well as identifying and confirming affecting parameters which are related to well configuration, drilling parameters and formation. The walk rate prediction model with the best configuration making best prediction when testing with the testing dataset yield a result of satisfied hit fraction (correct prediction).