An emergency multi-objective compromise framework for reservoir operation under suddenly injected pollution

Sadegh Vanda, Mohammad Reza Nikoo*, Narges Taravatrooy, Mojtaba Sadegh, Malik Al-Wardy, Jan Franklin Adamowski

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

An emergency multi-objective framework was developed to achieve an optimal reservoir operating strategy under sudden pollution injection. To assess a wide range of responses to potential future pollution injection events, a large number of reservoir release and pollution injection scenarios (1000 release modes and three injection location scenarios) were considered. The CE-QUAL-W2 model served to simulate pollutant concentration and reservoir cleanup time (RCT) under each scenario. To minimize the reservoir water quality modeling's computational burden, a multilayer perceptron (MLP) neural network was trained and validated against simulated responses to various scenarios. To reduce the dimensions of the problem, forcings of this surrogate model were made orthogonal by principal component analysis (PCA). This surrogate model helps estimate response variables at any intermediate scenario and can be readily coupled with a non-dominated sorting genetic algorithm-II (NSGA-II) optimization model to underpin Pareto optimal solutions. Four objective functions were considered: (i) non-supplied water demand, (ii) weighted combination of frequency and magnitude of pollution violation, (iii) ratio of pollution rate released from reservoir outlets to the total rate of injection, and (iv) difference between the reservoir volume and its normal water storage. Finally, a multi-method decision-making procedure was applied to select the best compromise solution for all stakeholders (i.e., Ministry of Energy, Ministry of Agriculture, Center for Environmental Health and Regional Water Authority). This study is the first to propose a reservoir optimal operation model using an MLP-PCA model coupled with several conflict resolution models. Findings showed that the closer the pollution injection point was to the reservoir outlet, the shorter the RCT. Following a 40-day simulation, the solution selected for the injection site closest to the reservoir outlet resulted in a 4-day RCT. Under identified compromise solutions for scenarios of pollutant (E. coli) injections 470 m, 3290 m, and 6580 m upstream from the dam, the RCT decreased by 1, 2 and 2 days, respectively.

Original languageEnglish
Article number126242
JournalJournal of Hydrology
Volume598
DOIs
Publication statusPublished - Jul 2021

Keywords

  • Fallback bargaining
  • Non-dominated sorting genetic algorithm-II
  • Principal component analysis
  • Social choice rules
  • Sudden pollution injection

ASJC Scopus subject areas

  • Water Science and Technology

Cite this