Multiple reservoir optimization using constraint programming for water scarcity resilience in the Chao Phraya River Basin
Issued Date
2026-05-01
Resource Type
ISSN
21905487
eISSN
21905495
Scopus ID
2-s2.0-105035816369
Journal Title
Applied Water Science
Volume
16
Issue
5
Rights Holder(s)
SCOPUS
Bibliographic Citation
Applied Water Science Vol.16 No.5 (2026)
Suggested Citation
Sawangphol W., Rittima A., Phankamolsil Y., Kraisangka J., Tabucanon A.S., Talaluxmana Y., Vudhivanich V. Multiple reservoir optimization using constraint programming for water scarcity resilience in the Chao Phraya River Basin. Applied Water Science Vol.16 No.5 (2026). doi:10.1007/s13201-026-02820-2 Retrieved from: https://repository.li.mahidol.ac.th/handle/123456789/116383
Title
Multiple reservoir optimization using constraint programming for water scarcity resilience in the Chao Phraya River Basin
Corresponding Author(s)
Other Contributor(s)
Abstract
Reservoir optimization has gained increasing attention as a cornerstone of sustainable development and water resource resilience, particularly in regions prone to climate extremes. To address these challenges, this study introduces a novel Constraint Programming (CP) for multiple reservoir optimization to enhance water scarcity resilience in the Chao Phraya River Basin (CPYRB). CP models with and without incorporating travel time taken from dams to water demand nodes, CPM1 and CPM2, were accordingly formulated to find optimum daily release schemes of major dams in the basin. The development of the CP models in CPYRB utilized Python's GEKKO library with the IPOPT solver, which offers a user–friendly modeling environment for large–scale, nonlinear, and constrained optimization problems. The simulated results from 2000 to 2020 demonstrated that CPM1 and CPM2 models could generate different daily release schemes to recommend a release guideline trajectory, while average annual releases accomplished by CP models could closely replicate current operation. Moreover, both models could increase the end–of–wet–season storage in the reservoir system by 2,712 and 1,265 MCM/yr, respectively, to potentially supply overplanting during the dry season, revealing improved water scarcity resilience in the basin. By incorporating travel times between releases from different dams and the various demand nodes, CPM2 offered more realistic and effective operations to timely and spatially distribute water in the irrigation farm area than CPM1. This ensured that water was supplied at various water distribution zones at the right time of use while satisfying system–wide objectives and time–dependent constraints. Importantly, the CPM2 model, which exhibited the smallest discrepancy in average total releases from the Bhumibol and Sirikit Dams, demonstrated a strong long–term quantitative agreement with the current operation. Moreover, the CPM1 model demonstrated a significant advantage over other optimization models including deep reinforcement learning, non–linear programming, and adaptive neuro fuzzy inference system in view of increasing the long–term end–of–wet season storage levels of two main storage dams, achieving + 15.73% and + 16.36% for Bhumibol and Sirikit Dams, respectively. This proposed CP framework provides a robust operational tool for dam operators to enhance water scarcity resilience in complex, highly constrained multi–reservoir systems.