Optimal Control of Water Quality in a Recirculating Aquaculture System

This study compares nonlinear model predictive controls, using setpoint tracking objective function (NMPC), and economic objective function (E-NMPC), with a hierarchical control structure of PI controllers applied to the water treatment of a Recirculating Aquaculture System (RAS), which consists of a tank, a biofilter, a stripper and an oxygen cone. Two of the three control structures used in this work consist of an optimization layer on top of proportional integral (PI) control loops or NMPC. The optimization layer reduces the degrees of freedom with an economic objective function by deactivating some of the manipulated variables when choosing between buffer and base addition and location of addition to the system. The third structure consists of a single layer with E-NMPC. The PI control structure is easy to design once we know the relationship between the variables. The PI controllers' performance were satisfactory, but a small back-off would be needed, if the constraints would not be soft constraints, or use of buffer with base would be necessary to improve speed and avoid constraint violation, as some components took more than 6 days to reach the optimal concentration. The NMPC and the E-NMPC performed better: the NMPC was much faster on conducting the controlled variables to optimal conditions, despite not providing optimal economic cost during the trajectory; and the E-NMPC provided a mixture of smooth trajectory and optimal cost. Copyright (c) 2022 The Authors. This is an open access article under the CC BY-NC-ND license(https://creativecommons.org/licenses/by-nc-nd/4.0/)