Prediction of air pollutant concentrations based on TCN-BiLSTM-DMAttention with STL decomposition

Abstract

A model with high accuracy and strong generalization performance is conducive to preventing serious pollution incidents and improving the decision-making ability of urban planning. This paper proposes a new neural network structure based on seasonal–trend decomposition using locally weighted scatterplot smoothing (Loess) (STL) and a dependency matrix attention mechanism (DMAttention) based on cosine similarity to predict the concentration of air pollutants. This method uses STL for series decomposition, temporal convolution, a bidirectional long short-term memory network (TCN-BiLSTM) for feature learning of the decomposed series, and DMAttention for interdependent moment feature emphasizing. In this paper, the long short-term memory network (LSTM) and the gated recurrent unit network (GRU) are set as the baseline models to design experiments. At the same time, to test the generalization performance of the model, short-term forecasts in hours were performed using PM2.5, PM10, SO2, NO2, CO, and O3 data. The experimental results show that the model proposed in this paper is superior to the comparison model in terms of root mean square error (RMSE) and mean absolute percentage error (MAPE). The MAPE values of the 6 kinds of pollutants are 6.800%, 10.492%, 9.900%, 6.299%, 4.178%, and 7.304%, respectively. Compared with the baseline LSTM and GRU models, the average reduction is 49.111% and 43.212%, respectively.

Short-Term Forecasting of Ozone Concentration in Metropolitan Lima Using Hybrid Combinations of Time Series Models

“…the ozone concentration model is represented exclusively as a function of time as a relevant factor without considering meteorological factors, the decomposition methods have shown great performance, since in this investigation the significant models (p < 0.05; R 2 max: 0.949) with errors less than 20% (RMSE, RMSPE, MAE, MAPE) showed great performance. These errors have been comparable to other STL decomposition studies that used root mean square error (RMSE: 6.8%) and mean absolute percentage error (MAPE: 10.49%) as benchmarks for forecast reliability for ozone [10]. This evaluation of tropospheric ozone explains its long-term and seasonal behavior with temporary ozone patterns [41], in accordance with what was demonstrated by Carbo-Bustinza [23] for the winter months in these geographic areas.…”

“…For its part, the second component, "seasonality", describes the fluctuations of the periodic seasons (decomposed), and the third, fixed by a short-term component, shows the "rest" of the random data once the first two components have been separated. In addition, other combined decomposition methods or structures have been proposed in series and time convolution and long-term short-memory bidirectional networks [10]. Other models use a non-parametric Theil-Sen estimator as a robust Kendall [11] line-fit method or locally estimated scatterplot models for smoothing to filter the data obtained and subsequently decompose the time series models into trend, seasonal, and residual components of data and then recombine them appropriately [12].…”

Multivariable System Prediction Based on TCN-LSTM Networks with Self-Attention Mechanism and LASSO Variable Selection

“…For complex multivariable systems, it is difficult to establish an accurate dynamic model for simple conventional networks due to the absence of feature extraction or memory function. − The LSTM network and TCN network have unique advantages in modeling time series data with long-term dependence because of their long-term memory and feature extraction capabilities . The LSTM network is a variant of the recurrent neural network (RNNS) to solve the problems of gradient disappearance and gradient explosion when processing long sequences .…”

“… 18 − 20 The LSTM network and TCN network have unique advantages in modeling time series data with long-term dependence because of their long-term memory and feature extraction capabilities. 21 The LSTM network is a variant of the recurrent neural network (RNNS) to solve the problems of gradient disappearance and gradient explosion when processing long sequences. 22 The LSTM network introduces memory cells and a number of gating units to selectively remember or ignore information from input data and transfer information in time.…”

Monthly runoff prediction based on a coupled VMD-SSA-BiLSTM model

“…Chen and Liang 14 combined the empirical modal decomposition method (EMD), attention mechanism, and BiLSTM neural network to predict the daily longitudinal flow at the Qingxi River site in Xuanhan County, obtaining accurate results that met prediction requirements. Regarding the study of bidirectional long short-term memory neural networks, Li and Jiang 15 used STL for sequence decomposition, temporal convolution, a bidirectional long short-term memory network (TCN-BiLSTM) for feature leaning the decomposed series, and interdependent moment feature emphasisation using DMAttention to predict the concentration of air pollutants, achieving accurate prediction accuracy. Sathi et al 16 used a CNN-BiLSTM model to predict the attention to the induced electric field of a transcranial magnetic stimulation coil with similarly excellent results.Current research on monthly runoff prediction models, both domestically and internationally, mainly focuses on coupling optimization algorithms with neural networks to improve prediction accuracy by constructing coupled models.…”