An adaptive surrogate-based, multi-pollutant, and multi-objective optimization for river-reservoir system management

Abstract

Integrated management of quality and quantity of river-reservoir water can provide comprehensive information to manage river-reservoir water resources. However, high computational bottlenecks have prevented such management from being applicable in real-world systems. Accordingly, in the present study, we proposed a multi-objective optimization algorithm based on modular support vector regression (SVR) in which several small sub-SVR modules trained through an efficient adaptive procedure cooperate to solve a large-scale problem related to integrated management of quality and quantity of river-reservoir. The performance of the proposed approach was evaluated through an adaptive surrogate-based simulation-optimization (ASBSO) framework under reservoir selective withdrawal scheme (SWS) in Ilam integrated River-Reservoir. The ASBSO framework provided a set of non-dominated optimal solutions to alleviate Ilam River water quality standard violations, enhance Ilam Reservoir outflow water quality, and maximize the downstream water supply satisfaction. The analysis of the Pareto-front indicated that the implementation of MPWLA (multi-pollutant waste load allocation) programs at any level could alleviate the water quality problems in Ilam Reservoir. Furthermore, the reservoir water storage in the regular patterns to meet downstream water demands have resulted in water quality deteriorations in Ilam Reservoir. The results obtained from examining the proposed approach, showed that integrated river-reservoir system management improved water quality in the range from 7 to 28% at the checkpoints of the Gol-Gol Branch of Ilam River and in the range from 5 to 21% at Ilam Reservoir outflow.