Research on PM2.5 Concentration Prediction in the Air Based on the TCN-BiGRU-AT Model

In recent years, the frequent occurrence of urban haze weather has posed a serious threat to people’s health. Accurately and efficiently predicting the concentration of PM2.5 in the air for a certain period in the future and taking corresponding measures based on the concentration is of great significance for reducing the occurrence of haze weather. Traditional CNN have limited capabilities in extracting temporal features, while TCN have better capabilities in this regard. Traditional RNN and their variants, LSTM, are prone to problems such as gradient vanishing, gradient explosion, and overfitting, while BiGRU can effectively alleviate these issues and improve the training efficiency of the model. Therefore, this paper proposes a combined model TCN-BiGRU-AT integrating TCN, BiGRU, and attention mechanisms, taking Shenyang City in Liaoning Province as the research object to predict the concentration of PM2.5 in the air. The experimental results show that the model’s predictive performance indicators, MAE and RMSE, are both lower than those of the comparative models, and the fitting ability indicator $R^{2}$ is closer to 1 compared to the comparative models.