Exploring a multi-output temporal convolutional network driven encoder-decoder framework for ammonia nitrogen forecasting | JoVE Visualize

Area of Science:

Environmental Science
Artificial Intelligence
Water Resource Management

Background:

Artificial neural networks (ANNs) are increasingly used for water quality prediction due to their learning and generalization capabilities.
Encoder-Decoder (ED) structures effectively capture complex nonlinear relationships by learning compressed data representations, removing noise and redundancy.
Accurate ammonia nitrogen forecasting is crucial for effective water quality management and pollution control.

Purpose of the Study:

To propose and evaluate a novel multi-output Temporal Convolutional Network based Encoder-Decoder (TCN-ED) model for ammonia nitrogen forecasting.
To assess the efficacy of combining ED structures with advanced neural networks for reliable water quality prediction.
To compare the performance of the proposed TCN-ED model against established models like LSTM-ED, LSTM, and TCN.

Main Methods:

A TCN-ED model was developed using hourly water quality and meteorological data from a Shanghai case study.
Input data included one hourly water quality factor and aggregated hourly meteorological factors from 32 stations over 24 hours.
Comparative analysis was performed using Long Short-Term Memory based ED (LSTM-ED), LSTM, and TCN models with training and testing datasets.

Main Results:

The TCN-ED model successfully mimicked the intricate dependencies between ammonia nitrogen and influencing factors.
TCN-ED provided more accurate ammonia nitrogen forecasts (1- to 6-h-ahead) compared to LSTM-ED, LSTM, and TCN models.
The TCN-ED model demonstrated superior accuracy, stability, and reliability in water quality prediction.

Conclusions:

The developed TCN-ED model offers a significant advancement in ammonia nitrogen forecasting accuracy and reliability.
This improved forecasting capability can enhance river water quality monitoring and early warning systems.
The findings support better water pollution prevention strategies, contributing to river environmental restoration and sustainability.