Machine Learning Approach for Water Control Diagnostics Plots

Abstract

Chan’s water control diagnostic plots are the most common way to investigate the mechanism that causes water production, but the traditional investigation by human has a degree of uncertainty and requires long time. The purpose of this project is to build a supervised machine learning model using ridge classifier to detect water production mechanism WPM (coning or channeling) accurately and time effectively. Firstly, we performed a conventional identification of Chan’s plots pattern from Heglig’s production data, then the data is divided into training set and test set, also the training set is split into two trends to train two different Models and create the ensemble classifier, then we evaluate the model ability on the test set, and then a hyper parameter (alpha) was tested many times to improve each model accuracy. We find out that Model1 demonstrates high accuracy to detect the WPM, and it is able to overfit the training set by 100% also it showed high degree of accuracy to classify unseen data (generalization) by 100% and for Model2 the overfitting is 100% and generalization is 89%, The accuracy of ensemble classifier is 100% on the test set. The project showed successful application of Machine Learning by using ensemble classifier and Ridge Classification algorithm to classify the WPM based on Chan’s water control diagnostic plots in efficient way that would make the WPM detection much easier. Some other mechanisms may be included in the future work that would be done to develop this model.