Development of a new hybrid model to enhance streamflow estimation using artificial neural network and reptile search...

Area of Science:

• Hydrology and Water Resources Engineering
• Artificial Intelligence in Environmental Science
• Computational Fluid Dynamics

Background:

• Accurate streamflow prediction is crucial for water resource management, especially in data-scarce regions.
• Artificial Neural Networks (ANN) are effective for time-series forecasting but require optimization for enhanced performance.
• Metaheuristic algorithms offer potential for improving ANN model accuracy in hydrological applications.

Purpose of the Study:

• To forecast monthly streamflow in Urmia, Iran, using Artificial Neural Network (ANN) models.
• To evaluate the performance of a novel metaheuristic optimizer, the Reptile Search Algorithm (RSA), when combined with ANN for streamflow prediction.
• To compare the efficacy of ANN-RSA against ANN combined with Particle Swarm Optimization (PSO) and Whale Optimization Algorithm (WOA).

Main Methods:

• Developed five input variable patterns using temperature, precipitation, and streamflow data.
• Trained and tested ANN models using 70% and 30% of the data, respectively.
• Implemented and compared hybrid models: ANN-RSA, ANN-PSO, and ANN-WOA for streamflow simulation.

Main Results:

• The ANN-RSA hybrid model demonstrated promising results across various stations and data patterns, with notable performance metrics at Band, Babaroud, Nazlo, and Tapik stations.
• The ANN-PSO hybrid model consistently outperformed the ANN-RSA model in streamflow simulation accuracy.
• Lagged monthly streamflow was identified as a significant input parameter influencing prediction accuracy, though model performance varied across different hydrological conditions and locations.

Conclusions:

• The novel Reptile Search Algorithm (RSA) shows potential for enhancing ANN performance in streamflow prediction, particularly in specific hydrological contexts.
• Hybrid models, especially ANN-PSO, offer superior accuracy for monthly streamflow forecasting compared to ANN-RSA.
• The study underscores the importance of parameter selection and highlights the potential of hybrid AI approaches for addressing hydrological challenges.