Design of an Intelligent Variable-Flow Recirculating Aquaculture System Based on Machine Learning Methods

A recirculating aquaculture system (RAS) can reduce water and land requirements for intensive production. However, a traditional RAS uses fixed circulation flow rate treatment. In general, the in an is highly turbid only when animals are fed they excrete. Therefore, quality regulation technology based on process control proposed this paper. The intelligent variable-flow was designed circulating pump-drum filter linkage working model. Machine learning methods were introduced to develop model maintain clean stable environment. Results showed that long short-term memory network performed with highest accuracy (training set 100%, test 96.84%) F1-score 93.83%) among artificial neural networks. Optimization including grid search, cuckoo linear squares, gene algorithm improve classification ability of support vector machine models. all models passed cross-validation could meet standards. summary, (accuracy: training 98.95%; F1-score: 99.17%) suitable as optimal RAS.