Comparative evaluation of PID controller performance for dissolved oxygen control in flow-through aquaculture systems

Flow-through aquaculture systems are highly sensitive to continuous water inflow and outflow, causing dissolved oxygen (DO) to fluctuate with influent water quality, fish oxygen consumption, and oxygen supply conditions. Consequently, maintaining adequate oxygen remains challenging. Most farms still rely on fixed oxygen input or simple on/off control, often resulting in oxygen oversupply and poor maintenance of the target DO range. This study compared the DO control performances of P, PI, PD, and PID controllers in a flow-through aquaculture system to assess their relative control effectiveness and practical applicability. Using water-quality and oxygen flow-rate data collected from an operating farm, a measured-data-based dynamic DO model was developed using a mass-balance equation based on the continuous stirred-tank reactor framework. The internal model control method was then employed to derive controller parameters using the identification dataset, and the selected controller gains were subsequently evaluated on a separate held-out validation period. Under held-out validation, PI and PID controllers maintained 100% of simulated DO within ±5% of the 10 mg/L target at controller update intervals of 1, 10, and 20 min. At 20 min, the corresponding within-band percentages of P and PD decreased to 25.86% and 26.13%, respectively. However, all controller structures deteriorated markedly at 30- and 60-min update intervals, indicating that sparse controller updating can become a dominant limitation. These results demonstrate that integral action is important for correcting persistent DO deviation under short-to-moderate update intervals, while the controller update interval itself must also be appropriately selected.