Application of Artificial Neural Network to Estimate Rate of Penetration for Geothermal Well Drilling in South Sumatera

Abstract

In order to plan an optimum geothermal well drilling scheme, a proper identification of drilling parameters should be well known. Information of the parameters consists of weight on bit (WOB), true vertical depth (TVD), rate of penetration (ROP), foam flowrate (FF), and rotary speed (N). The valuable information can be provided by the drilled geothermal wells. Correlation of the drilling parameters is then obtained based on the information. The application of Artificial Neural (ANN) Network is needed since the relationships among the parameters are very complex and nonlinear. Moreover, the relationships are not easily known. In this paper, Artificial Neural Network was promoted to estimate penetration rate. Data were obtained from three wells at a field in South Sumatera, Indonesia. Three ANN models were generated. Each model includes different input parameters. Based on the comparison results, the ANN-3 model has the best level of accuracy with the average values of the parameters MAE, MARE, MSE, ARMSE, and the correlation coefficients are 0.8883, 9.54%, 1.1878, 1.0825, and 0.9938 respectively. ANN models can play a role in identifying parameters that affect the characteristics of penetration rate. Keywords—Drilling, WOB, Geothermal, ROP, ANN.

Application of Artificial Neural Networks for Predicting Relative Permeability in Talang Akar Formation

“…The sigmoid function is used for training data as an activation function for the neuron in the hidden and output layers. The sigmoid function has a minimum value of 0 and a maximum value of 1 and can be distinguished everywhere by a positive slope [25][26][27]. Therefore, the input-output numeric values need to be normalized in the interval 0 and 1.…”

Optimized artificial neural network application for estimating oil recovery factor of solution gas drive sandstone reservoirs

“…In the process of building a model with ANN, the input parameters involved need to be normalized, while the output needs to be denormalized. The normalization equation used is as follows [ 26 ]. where x j and x j * are the values measured and normalized at data point j; x min and x max are the minimum and maximum data respectively.…”