Forecasting Dissolved Oxygen Level in Land-Based Fish Farms using a Context-Aware Recurrent Neural Network

Predicting Dissolved Oxygen (DO) levels in precision fish farming is crucial as it directly impacts the well-being and growth of fishes. In this paper, we propose a sensing method that is suitable to be used in edge-computing and which makes use of deep learning to estimate dissolved oxygen in fish farms based on a context-aware recurrent neural network trained by the relationship between the inlet dissolved oxygen, the estimated biomass, the period and time of measurement, and the food given to the fish. The proposed technique has been applied to a real-world dataset coming from a trout fish farm located in Trentino, a region in Northern Italy.