Comparison of parallel optimization algorithms on computationally expensive groundwater remediation designs

Area of Science:

• Environmental Science
• Computational Science
• Optimization

Background:

• Contaminated groundwater poses global risks to human health and ecosystems.
• Groundwater remediation is vital but often hindered by high costs and computationally intensive design processes.
• Efficient planning requires evaluating numerous management decisions, demanding significant computational resources.

Purpose of the Study:

• To evaluate the performance of a novel parallel surrogate-based optimization algorithm, parallel stochastic radial basis function (p-SRBF), for groundwater remediation design.
• To assess the efficiency and effectiveness of p-SRBF compared to other parallel optimization algorithms on real-world contaminated aquifer problems.
• To demonstrate the potential of p-SRBF in accelerating cost-effective groundwater remediation planning.

Main Methods:

• Application of the parallel stochastic radial basis function (p-SRBF) algorithm to two superfund site case problems (Umatilla and Blaine Aquifers).
• Comparison of p-SRBF with genetic algorithm, mesh adaptive direct search, and asynchronous parallel pattern search optimization.
• Evaluation based on solution quality, computational cost reduction, and robustness across multiple trials using parallel computing up to 64 cores.

Main Results:

• p-SRBF achieved exceptional (superlinear) speedup with 4-16 cores and excellent speedup up to 64 cores, maintaining 80% efficiency.
• p-SRBF outperformed other algorithms, finding the optimal solution for both Umatilla and Blaine cases and reducing computational budget by at least 50%.
• Statistical analysis confirmed p-SRBF's superior performance over alternative methods at a 5% significance level.

Conclusions:

• p-SRBF is a highly effective and efficient algorithm for computationally demanding groundwater remediation design problems.
• The algorithm offers significant speedup and cost reduction, making it a promising tool for environmental management.
• This study advances theoretical and practical approaches to real-world groundwater cleanup strategies.