RNN-Driven Smart Water Quality Management for Vannamei Shrimp Aquaculture

The food crisis that may occur is due to the increasing population and the dwindling number of food sources. So the Indonesian government in a structured manner carries out a program to increase vannamei shrimp (Litopenaesus vannamei) cultivation intensively, to ensure its survival in the future. Intensive vannamei shrimp cultivation requires intensive management of water quality control, which of course requires high-tech equipment. This study proposes modeling of control parameters, namely shrimp feed, CaCO3 to maintain acidity levels, extra aerators to maintain dissolved oxygen conditions, and alarm flags to provide notification of abnormal conditions in shrimp ponds. These control parameters are set based on the results of multimodal sensor readings, namely pH sensors, delta pH, dissolved oxygen sensors, and water surface temperature. Since shrimp pond conditions are dynamic and non-linear, Recurrent Neural Network-based modeling was implemented. Because of the machine learning base applied, we named this system smart water quality management (SWQM). The results show that RNN-based modeling has an accurate RMSE value and can be applied in real time according to expert rules obtained from shrimp farmers. Levenberg Marquardt-based RNN can produce optimal RMSE for feed scheduling modeling with RMSE of 0.0007 and 0.0057 for CaCO3 scheduling.