Dynamic modeling of recirculating aquaculture systems with an integrated application of nonlinear model predictive control and moving horizon estimation

Abstract

Growing concerns regarding the sustainability of the aquaculture industry has led to the development of recirculating aquaculture systems (RAS) where the addition of wastewater treatment units is accompanied by a reduction in water consumption and waste release. In this study, a mechanistic dynamic model of RAS was proposed and validated using experimental data available in the literature. Fish health is crucial to the profitability of an aquaculture facility; thus, fish performance and welfare measured in terms of growth and mortality were also incorporated within the proposed model. The model was then used to provide insights regarding the operation and management of RAS. According to the results of this analysis, continuous feeding was found to result in smaller fluctuations of waste product concentrations which is more desirable for the stability of wastewater treatment. Furthermore, under low rates of water exchange, addition of a denitrification unit to the RAS system would be necessary to avoid accumulation of nitrate. The environment used for fish growth (i.e., rearing environment) plays a significant role in feed utilization as well as fish performance. Thus, water quality parameters such as concentration of oxygen and waste components should be constantly controlled in a way to meet fish requirements. To achieve this goal, a nonlinear model predictive control (NMPC) integrated with moving horizon estimation (MHE) was implemented in this study to control RAS environment. The performance of the proposed control scheme was evaluated under full and partial accessibility to the states in the presence of process uncertainty and measurement noises. Assessment of the proposed controller was conducted by simulating the failure of unit or malfunction of a measurement device. In all scenarios, the proposed framework demonstrated the ability to maintain the water quality parameters close to target, indicating the promise of this control strategy in the closed-loop operation of RAS.