FNN-BiLSTM-Attention-DA: A hybrid fuzzy neural network and BiLSTM with multi-sensor information fusion for water quality monitoring and warning

Abstract

To conduct water quality anomaly alerts and set water quality alarm thresholds, a difference analysis (DA) model dependent on the FNN-BiLSTM-Attention mechanism is proposed in this study. The model efficiently lessens the impact of outliers and values that are missing in the statistical sample data on the predicted values, increasing the preciseness of the water condition alarms while accounting for the effects of seasonal and hydrological cycles on data changes. Five water quality indicators were used to describe the input data, which FNN first analyzed to extract the data's geographical properties. The time series features were then obtained by feeding the prior outputs into the forward and backward LSTM layers, respectively, via the BiLSTM layer. The FNN-BiLSTM-Attention model has the best MAE and MAPE on all water quality measures, according to the experimental data, and it has the lowest average MAE and MAPE on the water quality indicator dataset (YRB dataset), which is 0.174 and 6.32 %, respectively. Also, it has the highest average correlation coefficient of 0.936. In addition, the performance of the model was further validated on another proposed wastewater treatment plant dataset (WTPD dataset) in order to verify the generalization performance of the model.