Numerical investigation of optimization algorithms for the hydrodynamic model adaptation based on the well test results

This study presents a numerical investigation of optimization algorithms for the adaptation of hydrodynamic models based on the well test results. The considered optimization algorithms are based on the Levenberg-Marquardt method and have differences in the updating method of the Hessian's main diagonal and the calculation of the Jacobian. Two algorithms are presented in the pseudocode form, which allow to configure variations of the Levenberg-Marquardt method. Functions that transform the unconditional Levenberg-Marquardt optimization into a conditional one are proposed. A comparative analysis of the speed and quality of well bottom rates calculation for ten wells with different types by the number of analytical model function calls, calculation time, average relative error showed positive results when using the proposed numerical optimization algorithms.