“…In that case, the objective function is ultimate recovery or net present value, and the control variables are the well rates, well pressures, or well valve settings. Initially, this was done for the optimization of tertiary recovery processes, see Ramirez [32], later followed by water flooding optimization, see, e.g., [3,5,34,36,45] and [15]. In both the history matching and the recovery optimization problems, the necessary conditions for optimality lead to the gradients, which are now the total derivatives with respect to the controls of the objective function, modified to include the reservoir model constraint.…”