Application of Mathematical Optimization Techniques in Reservoir Design and Management Studies

Abstract

Optimal monthly release policies are derived for Hoover Reservoir, Columbus, Ohio, using chance-constrained linear programming and dynamic programming-regression methodologies. Simulation procedures are used to examine and compare the overall performance of the optimal policies derived by the two methods. Results suggest that for a two-sided quadratic loss function, linear release policies are more optimal. It is also established that the maximum R2 criterion, generally used for model selection, does not exactly produce the best form of a release policy, particularly for nonlinear forms. At target releases at or below the safe yield of the case study reservoir, and for a one-sided quadratic loss function, the standard policy is optimal. At higher targets, nonlinear policies give better performance than the standard policy. Other observations are made concerning the performance of the two optimization approaches in a real case study.