Conjunctive Operation Optimization in Arid Regions Using AI

By Samudrapom DamReviewed by Susha Cheriyedath, M.Sc.Jan 24 2024

In an article recently published in the journal Nature Communications, researchers proposed an artificial intelligence (AI)-based model for optimal conjunctive operation of groundwater and surface water resources.

Limitations of conjunctive operation

In semiarid and arid regions, water resource management is a significant challenge for decision-makers and managers. Countries like Iran have witnessed a substantial rise in water scarcity in the past decades owing to its climatic conditions. The expansion of industrial, urban, and agricultural activities and population growth has further aggravated the problem of water scarcity. In such a scenario, excessive groundwater exploitation has led to a sharp decline in the groundwater level in large parts of Iran, which increased the importance of preserving groundwater resources in the country.

Conjunctive operation of groundwater and surface water resources can be a feasible solution to reduce the extreme exploitation of groundwater resources and preserve such valuable natural resources. It can also eliminate the need to make additional investments in surface water resources for dam construction and water transmission system design with an over-optimal capacity. However, properly implementing the conjunctive operation of water resources is a challenging engineering problem.

Classical methods cannot easily solve such complex problems, such as identifying the best operation scenarios, due to several limitations, which necessitated the development of more effective methods. In the last decade, soft computing techniques that search based on the initial population have been developed to identify near-optimal solutions to highly complex problems.

The proposed AI-based approach

In this study, researchers proposed a hybrid simulation-optimization model for the optimal conjunctive operation of surface water and groundwater resources. This second-level model was developed by identifying and merging the best aspects of two metaheuristics, including the symbiotic organization search algorithm and the moth swarm algorithm, and the resulting algorithm was then connected to an artificial neural network (ANN) simulator.

Researchers evaluated the efficiency of the developed model by comparing its results with two first-level simulation-optimization models. The study's objective was to create an efficient simulation-optimization model based on the hybridization of three powerful AI methods for the optimal allocation of groundwater and surface water resources in the Halilrood basin.

Initially, a structural sensitivity analysis was performed on the moth swarm algorithm structure to identify the algorithm’s most efficient operators, including Lévy-mutation, transverse orientation, and celestial navigation. The identified operators were then exported to the symbiotic organism search algorithm’s search process to form a robust hybrid optimizer/hybrid symbiotic organism search-moth swarm algorithm.

Eventually, the ANN as a simulator was linked with the developed hybrid optimizer to create a second-level simulation-optimization model, designated as symbiotic organisms search-moth swarm algorithm-ANN (SOS-MSA-ANN), which can minimize water scarcity in various sectors of the Halilrood basin considering the limitations on groundwater drawdown.

Significance of the study

The radial basis function (RBF)-ANN model was used to estimate groundwater in the studied areas, including Rabor, Jiroft, and Baft. Results showed that the determination coefficient values in all regions were over 0.99, and the error indices values were extremely small/less than 0.05 for mean absolute percentage error (MAPE). These results indicated that the selected ANN simulation model/neural network model was effective in estimating the groundwater level in all studied areas and could be reliably linked with the optimization models.

Subsequently, the second-level simulation-optimization SOS-MSA-ANN model results for the conjunctive operation of groundwater and surface water resources in the Halilrood basin were compared with the results of two first-level simulation-optimization models, including MSA-ANN and SOS-ANN.

The objective function values of the proposed hybrid SOS-MSA-ANN model, MSA-ANN model, and SOS-ANN model were 399.28, 544.68, and 769.27, respectively, which indicated the higher capability of the proposed model compared to the first-level simulation-optimization models.

Additionally, SOS-MSA-ANN supplied 99.73%, 99.50%, and 99.49% of the water demands in Jiroft, Rabor, and Baft, respectively, which was higher than the demands supplied by the two first-level models, which indicated that the SOS-MSA-ANN could better reduce/manage the total water deficits in the study period compared to the other two simulation-optimization models.

The SOS-MSA-ANN model had the lowest deficit values of 1.21, 2.82, and 3.17 in Baft, Rabor, and Jiroft, respectively, among all evaluated models. Moreover, the sustainability index values obtained by the SOS-MSA-ANN model to supply total demands in Jiroft, Rabor, and Baft were 97.06, 81.58, and 89.30, respectively. These values were significantly higher than the sustainability index values obtained by MSA-ANN and SOS-ANN.

Overall, the findings of this study demonstrated that the proposed hybrid SOS-MSA-ANN simulation-optimization model can be effectively utilized in the optimization of the conjunctive operation of groundwater and surface water resources.