Abstract
The concentration of ammonia nitrogen plays a crucial role in aquaculture, as an excessive level can hinder the growth of aquatic animals, or even cause death. Commercially available ammonia nitrogen sensors are expensive, have limited lifespans, require frequent maintenance, and may experience drift over time. In this study, we designed a recirculating aquaculture system in the lab and conducted 52 days experiment. Based on this system, a new mathematical model was developed to predict ammonia nitrogen, combining fish bioenergetics with mass balance of ammonia nitrogen. A sensitivity analysis of the model parameters was conducted to identify the key ones, and then a Bayesian optimisation method was used to calibrate these key parameters with data collected from the recirculating aquaculture system in the lab. Validation on unseen data demonstrated that the model can give a reasonable prediction of ammonia nitrogen.