A LSSVM approach for determining well placement and conning phenomena in horizontal wells

Understanding the time of water/gas breakthrough has a prominent role in cost effective oil production, improve oil recovery and extension the reservoir production time. The importance lies in the fact that once water or gas has broken through, the fluid distribution and the fluid relative permeabilities in the system will change. Accordingly, applying robust predictive models in this area to arrive at a proper estimation of breakthrough times as well as optimal horizontal well placement in heterogeneous and homogeneous reservoirs as a function of density difference ratio and rate is of great interest in oil and gas production system. The current study plays emphasis on applying the predictive model with the aim of the LSSVM (least square support vector machine) to estimate breakthrough time and optimum fractional well placement. Genetic algorithm (GA) was utilized to choose and optimize hyper parameters (γ and σ²) which are embedded in LSSVM model. Utilization of this model showed high competence of the applied model in terms of correlation coefficient (R²) of 0.9999 and 0.9999, mean squared error (MSE) of 0.000000142 and 0.000000622 from actual values for estimated dimensionless breakthrough time and optimum fractional well placement, respectively.