| dc.description.abstract | Recirculating aquaculture systems (RAS) are closed-loop systems that recycle most of the effluent of a fish tank after particle removal and water treatment. This type of process requires good control of the water quality, as the carbon dioxide, ammonia, and nitrate concentrations are toxic for the Atlantic salmon (Salmo salar - fish species considered) when above certain values, and oxygen concentration needs to be in a certain range to the welfare of the fish.
To regulate these quantities into an optimal value considering the process constraints and fish requirements, simplified steady-state and dynamic models were developed, and validated with real data. The developed models include pH modelling, which is an important variable in the system due to its effect on carbon dioxide and ammonia concentrations, so it should be monitored and controlled.
Some of these quantities are not measured in real-time in the RAS facility under study, so it is important to develop soft sensors that can complement the lab measuring procedure that takes time.
Therefore, the main objectives of this thesis are to study the effects of placement of pH and alkalinity adjustments; predict the behavior of the system in the main compartments (fish tank, bio-filter, and stripper); use the model for production optimization and control, including a study of different control structures, such as nonlinear model-based predictive control, economic nonlinear model-based predictive control, and proportional integral control with real-time optimization on top; and study the application of a feedforward neural network as a soft sensor to predict carbon dioxide, ammonia, and ammonium. | en_US |
