Abstract <p>The problem of structural-parametric identification of permeability in a fluid-saturated porous medium developed by a well network system is considered. A new effective approach to solving this problem is proposed, using discrete flow equations, gradient methods and cluster analysis techniques. This approach enables the determination of the permeability field that provides the best forecast of fluid flow rates at the wells. Using the proposed method, the existence of an optimal number of parameters defining the permeability field, which yields the best prediction of well flow rates, is computationally established. It is demonstrated that a good match of the model to the development history does not guarantee the best forecast.</p>