Skip to content
Conference

IoT-Enabled Deep Learning Framework for Real-Time Water Quality Prediction in Aquaculture

Sep 2026 · International Conference on Communications and Information Technology · pp. 1-6 · 0 citations · 33 references

Abstract

Freshwater Pangasius aquaculture requires reliable real-time forecasting of critical aquatic parameters to mitigate sudden environmental hazards and prevent stock mortality. Traditional manual sampling fails to capture abrupt parameter fluctuations, creating operational vulnerabilities in small and medium-scale farm operations. To solve this problem, an Internet of Things continuous monitoring framework paired with a hybrid deep learning architecture was deployed and evaluated in real field conditions. Hardware built with ESP32 microcontrollers collected continuous stream data from in-situ sensors measuring water temperature, pH, dissolved oxygen, and turbidity, publishing messages through MQTT to cloud storage. Preprocessing comprised linear interpolation, Min-Max normalization, and moving average smoothing across a 24-step sliding window applied to a dataset of 534,754 multivariate time-series records collected from August 7 to November 18, 2025. Models implemented in TensorFlow and Scikit-learn were trained under a 70:15:15 chronological train, validation, and test split. Across experimental evaluations, the proposed CNN-LSTM architecture achieved an overall Mean Absolute Error of 0.0151 and Root Mean Squared Error of 0.0184, outperforming standalone LSTM with Mean Absolute Error of 0.0376 and Root Mean Squared Error of 0.0392, GRU with Mean Absolute Error of 0.0355 and Root Mean Squared Error of 0.0372, and CNN-only models with Mean Absolute Error of 0.0160 and Root Mean Squared Error of 0.0204. Dissolved oxygen reached highest accuracy with Mean Absolute Error of 0.0105 and Root Mean Squared Error of 0.0135. Pond operators can now automate early warning notifications to proactively manage aeration systems.

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.