Sensitivity-driven calibration of Activated Sludge Model No. 1: application to a municipal wastewater treatment plant in north-western Algeria

ABSTRACT

Three-step graphic showing the ASM1 calibration workflow for the Ain Elhoutz wastewater plant: routine effluent COD data are fractionated and four parameters are optimised at 18 °C, then the same parameters are validated at 22 °C without re-tuning.

Mechanistic simulation of municipal activated sludge plants relies predominantly on Activated Sludge Model No. 1 (ASM1), yet routine application is hindered by over-parameterisation: calibrating the full kinetic set to dynamic plant data invites equifinality and loss of predictive power. This study develops a sensitivity-Driven ASM1 calibration workflow for a low-load plant in north-western Algeria, extending our earlier steady-state and dynamic simulation work in the region. ASM1 was implemented in GPS-X, and influent chemical oxygen demand (COD) was fractionated using the CEMAGREF protocol supplemented by a post-calibration mass-balance adjustment. Four parameters (XBH, μmax, H, YH, and KS), the subset repeatedly identified as dominant for effluent COD variance and practically identifiable in recent sensitivity and collinearity studies, were optimised against a targeted dynamic calibration window at 18 °C using the Nelder–Mead simplex algorithm and then tested on an independent seasonal dataset at 22 °C. Calibration achieved strong agreement (ℛ2 = 0.79), validation confirmed acceptable predictive performance (ℛ2 = 0.68) with no systematic bias, while the uncalibrated nitrogen variables showed a consistent over-prediction, quantified and attributed in Section 3. Requiring only routine COD data, the workflow provides a transferable, operator-grade template for dynamic ASM1 calibration at municipal plants with comparable monitoring programmes.