ANN modelling and GA optimisation of microalgal removal of cortisone acetate, organic matter and nutrients

Microalgal treatment can couple pharmaceutical attenuation with organic-matter and nutrient removal, but the operating windows for saline and freshwater species remain insufficiently defined. This laboratory study combined a central composite design, dual-detector confirmation of cortisone acetate (CA) removal, biomass mass balance, and artificial neural network (ANN) modelling and genetic algorithm (GA) optimisation to evaluate Tetraselmis chui ( T. chui ) and Scenedesmus obtusus ( S. obtusus ) under salinity, pH, and treatment-time variation. Cortisone acetate was evaluated with T. chui , whereas biochemical oxygen demand (BOD), chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP) were measured for both species. At 30.00 g/L salinity, pH 8.00, and 10.00 days, T. chui achieved 96.62% CA removal and 95.69–96.91% removal of the four water-quality parameters. At 0.00 g/L salinity, pH 7.00, and 10.00 days, S. obtusus achieved 96.25–98.41% removal of BOD, COD, TN, and TP. Species-specific ANN models accurately reproduced the experimental response surfaces (R² = 0.997 for T. chui and 0.978 for S. obtusus ), and the GA identified species-specific optima within the experimental ranges. The integrated experimental–analytical–modelling framework identifies salinity as the principal operational control while distinguishing the saline niche of T. chui from the freshwater niche of S. obtusus .