An adaptive moment estimation framework for well placement optimization

Abstract In this study, we propose the use of a first-order gradient framework, adaptive moment estimation (Adam), in conjunction with stochastic approximation, to well location and trajectory optimization problems. The Adam framework allows incorporation additional information from previous gradients calculate variable-specific progression steps. As result, assists search be adjusted further for each variable convergence speed-up problems where need approximated. We argue that under computational budget constraints, local algorithms provide suitable solutions heuristic initial guess. Nonlinear constraints are taken into account ensure proposed not violation practical field considerations. performance algorithm is compared against steepest descent generalized pattern search, using two case studies — placement four vertical wells 20 nonconventional (deviated, horizontal and/or slanted) wells. results indicate consistently outperforms tested methods terms efficiency final optimum value. Additional discussions regarding parameterization insights simultaneous perturbation approximations.