MPA-RNN: A Novel Attention-Based Recurrent Neural Networks for Total Nitrogen Prediction

Accurately predicting the short- and long-term variations of total nitrogen (TN) is vital for operating the wastewater treatment plants (WWTPs), considering the critical role TN plays in reflecting the eutrophication of wastewater. However, only a few relevant water quality parameters with limited samples can be obtained in WWTPs, which tremendously increases the difficulty in precisely predicting TN concentration. In this study, a multiphase attention-based recurrent neural network (MPA-RNN) is proposed. Benefited from its unique decomposition-summary attention structure, MPA-RNN first learns the temporal correlations and effectively excavates the useful information hidden in the historical data. Then, by designing a two-channel structure to transmit attention information, summary attention can integrate the decomposed information and learn the spatial relationships without information loss. Experimental results demonstrate that MPA-RNN achieves the best performance on both the SML2010 and practical TN datasets with the smallest root-mean-squared error, mean absolute error, and mean absolute percentage error when compared with the other state-of-the-art methods.

I. Introduction

As one of the major water quality parameters (WQPs), it is of significance for wastewater treatment plants (WWTPs) to accurately predict the short- and long-term variations of total nitrogen (TN). Short-term TN prediction can solve the monitoring lag problem that the online sensors usually have and real-time reflect the working situation of WWTPs [1]. Long-term TN prediction, on the other side, can forecast the possible over-discharging of TN happening in the future and help the WWTPs to take necessary actions beforehand, avoiding serious accidents and environmental pollution (such as aquatic eutrophication and ecosystem dysfunction) [2]. Therefore, no matter for the normal operation of WWTPs, or for avoiding accidents and emergencies, there is an urgent need to accurately predict the future variations of TN concentration.