Groundwater parameters estimation: A hybrid approach of convolutional layers with asynchronous and distributed bio-inspired algorithms
No Thumbnail Available
Date
2025
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This research focuses on aquifer hydraulic parameters estimation using bio-inspired algorithms since they can tackle groundwater model non-linearities. We propose two novel hybrid frameworks that combine the advantages of convolutional layers (CL) to enhance pattern recognition with heuristic search of Particle Swarm Optimization (PSO) and Differential Evolution (DE) algorithms. These integrations are implemented using an asynchronous and distributed approach to address efficiency issues in large-scale problems, resulting in ADPSO-CL (Asynchronous and Distributed Particle Swarm Optimization with Convolutional Layers) and ADDE-CL (Asynchronous and Distributed Differential Evolution with Convolutional Layers). The distributed method employs virtual machines, where a server generates and assigns particles to workers, which run in parallel with asynchronous iterative solution exchanges. We assess different algorithm configurations in an integrated water management model by coupling two software: Water Evaluation and Planning (WEAP) and MODFLOW. Results indicate that ADPSO-CL outperforms ADDE-CL by demonstrating more stable asynchronous communication, with fewer incomplete experiments (more than one worker was disconnected before completing all iterations), 33% in contrast to 71%. Additionally, produces results closer to the expected values, with mean absolute percentage error (MAPE) values of 78.25% for hydraulic conductivity and 55.56% for specific yield, compared to 299% and 209% in ADDE-CL. Moreover, ADPSO-CL has the fastest convergence rate, achieving efficient results in about half of the total iterations. This study introduces a novel and scalable architecture for intricate simulation–optimization problems, demonstrating its potential for future applications in real-world water resources planning and management
Description
Keywords
Aquifer parameters estimation, Inverse problem, Particle Swarm Optimization, Differential Evolution, Simulation–optimization model