A Heuristic dynamically dimensioned search with sensitivity information (HDDS-S) and application to river basin management

Abstract

River basin simulation and multi-reservoir optimal operation have been critical for river basin management. Due to the intense interaction between human activities and river basin systems, the river basin model and multi-reservoir operation model are complicated with a large number of parameters. Therefore, fast and stable optimization algorithms are required for river basin management under the changing conditions of climate and current human activities. This study presents a new global optimization algorithm, named as heuristic dynamically dimensioned search with sensitivity information (HDDS-S), to effectively perform river basin simulation and multi-reservoir optimal operation during river basin management. The HDDS-S algorithm is built on the dynamically dimensioned search (DDS) algorithm; and has an improved computational efficiency while maintaining its search capacity compared to the original DDS algorithm. This is mainly due to the non-uniform probability assigned to each decision variable on the basis of its changing sensitivity to the optimization objectives during the adaptive change from global to local search with dimensionality reduced. This study evaluates the new algorithm by comparing its performance with the DDS algorithm on a river basin model calibration problem and a multi-reservoir optimal operation problem. The results obtained indicate that the HDDS-S algorithm outperforms the DDS algorithm in terms of search ability and computational efficiency in the two specific problems. In addition; similar to the DDS algorithm; the HDDS-S algorithm is easy to use as it does not require any parameter OPEN ACCESStuning and automatically adjusts its search to find good solutions given an available computational budget.