A Dynamically Dimensioned Search Allowing a Flexible Search Range and Its Application to Optimize Discrete Hedging Rule Curves

Abstract

The discrete hedging rule for reservoir operation includes time-varying trigger volumes used for the onset and termination of water rationing, which complicates its optimization problems. A dynamically dimensioned search can be easily applied to complex optimization problems, but the performance is relatively limited in constrained optimization problems such as deriving reservoir operation rules. A dynamically dimensioned search allowing for a flexible search range is proposed in this study to efficiently solve constrained optimization problems. The modified algorithm can recursively update the search ranges of decision variables with limited overlaps. The above two algorithms are applied to derive hedging rule curves for three reservoirs. Objective function values are closely converged to optimum solutions, with fewer evaluations using the modified algorithm than those using the traditional algorithm. The modified algorithm restrains an overlapped search range of decision variables and can reduce redundant computational efforts caused by unreasonable candidate solutions that violate inequality conditions.