Analisa Prediksi ROP pada lapangan Gorgon Australia Menggunakan Machine Learning

Abstract

This design / research is about Analysis of ROP prediction in the gorgon field with 7 well datasets using the multilayer perceptron regressor machine learning method (MLP regressor) with the aim to predict the ROP model with 7 well datasets and compare variations in the number of data trains and variations in the number of features. The method used is to formulate a machine learning model architecture consisting of five stages: exploratory data analysis, data preparation, prediction and modelling, hyper-parameter tuning, and model evaluation. The results show that the more data trains used, the accuracy of the model will increase, while the increase in the number of features in this research model will cause the prediction value of blind data to decrease. The best predicted value is in scheme 2 which uses 3 features with 4 wells as trains.