The Early Warning Model of Dust Concentration in Smart Construction Sites Based on Long Short Term Memory Network

The air quality issue at the construction site of smart construction sites is a widely concerned issue among stakeholders. Regulating the emission of dust particles such as PM10 from construction sites is an effective measure to improve environmental quality. This paper proposes a smart construction dust particle concentration PM10 excess warning model based on LSTM neural network and time series processing methods, which achieves multi-step prediction by using predicted data as input for the next prediction. In the experiment, the dust data is first preprocessed as time series, then the LSTM model structure is adjusted to predict the PM10 concentration, and the predicted values are processed using a sliding window method. Depending on the warning threshold and conditions, the system warns 30 minutes before the exceedance of the standard.. Finally, the N-city smart construction PM10 dataset is used to implement online construction project supervision based on dust predictions. The RMSE, MAE and MAPE of the prediction model in this research were 3.60,1.61 and 1.87, respectively, with high accuracy of prediction.