Deep-learning based approach for forecast of water quality in intensive shrimp ponds

With the enormous development of aquaculture, reducing the impacts of effluent discharge and improving water quality has become a critical global environmental concern. It is important to assess and predict water quality in the environmental management process of shrimp mariculture. Meanwhile, accurate forecast of water quality in mariculture systems is still in the initial stages at present. In this study, deep belief networks (DBN) models were used to forecast water quality in intensive shrimp culture system. This method based on deep learning includes a five-layered structure to extract relationships between the quantitative characteristic of water bodies and water quality variables. The water quality was forecasted using the Canadian Water Quality Index (WQI) obtained from the output layer of simulated model. The results show that the DBN model has a great potential to predict water quality and the ability of generalisation and accuracy of model is satisfied.