Aquaculture 4.0: hybrid neural network multivariate water quality parameters forecasting model

Area of Science:

• Aquaculture
• Environmental Science
• Data Science

Background:

• Effective water quality management is crucial for sustainable aquaculture.
• Forecasting critical water quality parameters aids in proactive problem identification and intervention.
• Existing models may not fully capture the complex dynamics of aquatic environments.

Purpose of the Study:

• To develop and evaluate a novel hybrid deep learning model for multivariate water quality forecasting in aquaculture.
• To assess the model's accuracy in predicting key water quality parameters.
• To demonstrate the model's utility as a decision-support tool for aquaculture management.

Main Methods:

• A hybrid model combining Ensemble Empirical Mode Decomposition (EEMD), Long-Short Term Memory (LSTM) neural networks, and Multivariate Linear Regression (MLR) was developed.
• The model utilized multivariate time-series water quality sensor data from a Salmon offshore aquaculture farm.
• Model performance was validated against measured water quality data and Phytoplankton counts.

Main Results:

• The developed EEMD-MLR-LSTM NN model demonstrated high forecast accuracy for water quality parameters.
• The model effectively captured complex temporal patterns in the multivariate time-series data.
• Validation against real-world data confirmed the model's predictive capabilities.

Conclusions:

• The novel hybrid deep learning model is a valuable tool for enhancing water quality management in aquaculture.
• Accurate forecasting enables timely interventions, leading to improved aquaculture industry practices.
• This approach offers a significant advancement in supporting sustainable offshore fish farming.