Intelligent feeding technique based on predicting shrimp growth in recirculating aquaculture system

Abstract

Precise feeding in the recirculating aquaculture mode is a critical scientific problem that urgently needs a solution. This study aimed to develop an intelligent feeding technique in a recirculating aquaculture system for rearing Litopenaeus vannamei. The core of the intelligent feeding technique is the shrimp biomass prediction model. Accurate prediction of shrimp biomass could determine the appropriate feeding amount and ensure stable water quality. The data-driven prediction model was developed based on water quality indicators and aquaculture management data collected during shrimp rearing. Multiple linear regression, artificial neural networks and a support vector machine (SVM) were introduced to develop the shrimp biomass predicting model. Results showed that the SVM model gave the lowest root mean square error (0.6500), mean absolute error (0.4368) and mean absolute percentage error (3.70%), as well as the highest accuracy (90.91%). By analysing the predictive ability of the machine learning models, it was determined that the SVM model was the optimal model for predicting biomass. The intelligent feeding machine can apply the optimal model to calculate the shrimp biomass and determine the appropriate feeding amount by reading the sensors in real time.