MODEL-BASED PREDICTION OF DISSOLVED OXYGEN CONTENT IN FISH PRODUCTION

摘要

Dissolved oxygen (DO) plays a vital role in fish production, but DO sensors are expensive and often malfunction as a result of corrosion and deposition of debris. In this work, polynomial regression, back-propagation neural network (BPNN), and long short-term memory (LSTM) were applied to develop models for estimation of DO from non-DO measurements in fish production. The root mean square error (RMSE), mean absolute percentage error (MAPE), coefficient of determination (R-2), index of agreement (d), and Nash-Sutcliffe efficiency coefficient (NSE) were used to measure the performance of the three models. The results showed that the LSTM model achieved the best performance. A trained LSTM model can be used to estimate DO from water pH, temperature, and turbidity without using DO sensors. The estimated DO values can be used as a warning signal for DO sensor maintenance and as estimates of compromised DO measurements in fish production.