A hierarchical ASM1 calibration framework under limited data conditions: application to a municipal WWTP in Algeria

Abstract

ABSTRACT Flowchart of the calibration workflow, from plant data and ASM1/BSM1 modelling with 19 parameters, through PRCC screening to 13 and Sobol analysis to 8, to GA and PSO calibration and model validation. This paper presents a hierarchical ASM1 calibration framework applied to the Tlemcen municipal wastewater treatment plant, Algeria, under limited data conditions, using the Benchmark Simulation Model No. 1 (BSM1) implemented in MATLAB/Simulink. This study combines Partial Rank Correlation Coefficient (PRCC) screening under steady-state conditions and Sobol variance decomposition under dynamic conditions within a sequential sensitivity analysis framework, progressively reducing the parameter space from 19 to 8 influential parameters, prior to dynamic calibration using Genetic Algorithms (GA) and Particle Swarm Optimization (PSO), simultaneously optimizing five output variables: COD, TSS, BOD5, SNO and SNH. Based on 10 independent runs, PSO significantly outperformed GA in solution quality (mean J = 0.3150 ± 0.0001 vs 0.3167 ± 0.0009) and stability, confirmed by a Wilcoxon signed-rank test (p = 0.0020). The calibrated model showed good agreement with 2024 measured data (NRMSE <0.40, NMAE <0.41, TIC <0.22). Validation on an independent 2023 dataset confirmed temporal robustness (TIC <0.35). The proposed framework provides a practical decision support tool for aeration management and energy optimization, with potential extensions toward automatic control strategies and digital twin development for similar Algerian WWTPs operating under constrained monitoring conditions.