Advancing surrogate-based optimization of time-expensive environmental problems through adaptive multi-model search

• DOI: 10.1016/j.envsoft.2023.105639
• Publication Date: 2023-01-30T10:49:04Z
• Authors: Spyridon Tsattalios; Ioannis Tsoukalas; Panagiotis Dimas; Panagiotis Kossieris; Andreas Efstratiadis; Christos Makropoulos
• Keywords: Surrogate model; Metamodeling; Robustness; Benchmarking

Complex environmental optimization problems often require computationally expensive simulation models to assess candidate solutions. However, the complexity of response surfaces necessitates multiple such assessments and thus renders search procedure too laborious. Surrogate-based is a powerful approach for accelerating convergence towards promising Here we introduce Adaptive Multi-Surrogate Enhanced Evolutionary Annealing Simplex (AMSEEAS) algorithm, as an extension its precursor SEEAS, which single-surrogate-based method. AMSEEAS exploits strengths surrogate that are combined via roulette-type mechanism, selecting specific metamodel be activated in every iteration. proves robustness efficiency extensive benchmarking against SEEAS other state-of-the-art surrogate-based global methods both theoretical mathematical demanding real-world hydraulic design application. The latter seeks cost-effective sizing levees along drainage channel minimize flood inundation, calculated by time-expensive hydrodynamic model HEC-RAS.