Prediction breakthrough time of water coning in the fractured reservoirs by implementing low parameter support vector machine approach

Abstract

Owing to water coning, water flows into the production wellbore from below the perforated channels and normally causes several technical issues in wellbore and surface production facilities such as separators. Knowledge about the time that water coning happens could help us greatly to plan the production scheme and overcome the addressed hurdles. Due to this fact, this research goes to great lengths of a low parameter approach development to figure out smartly the breakthrough time of water coning in fracture reservoirs (FRs). To reach the goal of this work, least square support vector machine (LSSVM), artificial neural network (ANN) and hybrid of fuzzy logic, Kalman filter and genetic algorithm (HFKGA) were utilized to predict breakthrough time of water coning in FRs. To scrutinize the proposed approaches of estimating the breakthrough time of water coning, a numerous number of real data from the northern Persian Gulf oil fields was implemented. Outputs of LSSVM approach draw parallel with the corresponding experimental values and results of HFKGA and ANN models depicting the giant potential implication of the proposed approach to predict the breakthrough time of water coning in FRs. Moreover, advantages of statistical parameters like the average absolute relative deviation (AARD%) was gained to quantify robustness and accuracy of the proposed model. Thanks to implementation of this cutting edge research, knowing timely the breakthrough time of water coning in oil wells provides this opportunity to plan more accurately and operate responsibly further.