PM2.5 high-precision space-time prediction method based on deep learning

Abstract

The invention discloses a PM2.5 concentration space-time prediction method based on deep learning, and the method comprises the steps: firstly carrying out data preprocessing and space-time matching of PM2.5 data and multi-source data of a ground monitoring station; carrying out influence factor analysis on multi-source data such as meteorological data, spatial related data and physical characteristic data, and selecting related characteristic factors as auxiliary input data of the model; on the basis of the space-time correlation analysis of the PM2.5 data of the ground monitoring stations, obtaining appropriate time lag variables, and clustering the monitoring stations; on the basis of the historical sequence and the PM2.5 concentration value of the adjacent grid, in combination with related auxiliary data, using an improved LSTM model to output the PM2.5 concentration value of the grid at the future moment and carrying out fine PM2.5 space-time distribution mapping. The method can improve the precision of long-term space-time prediction and can also be used for predicting the PM2.5 concentration at the continuous large-range future moment in space. And the requirements in practical application are met.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence and information application, and specifically relates to a PM based on deep learning 2.5 A method for spatiotemporal prediction of concentrations. Background technique [0002] PM 2.5 High-precision spatio-temporal prediction of concentration has important practical significance for air pollution management and prevention, public environmental information services, etc. With the development of artificial intelligence, deep learning-based models, especially long-short-term memory models, are widely used in PM due to their excellent long-term sequence dependence performance. 2.5 Concentration Prediction. Many scholars have established a model based on the normal short-term memory network, also known as the LSTM model, for PMs in different regions. 2.5 Concentration prediction research was carried out, and the model was evaluated, and the LSTM-based model was verified in PM 2.5 A...

AI Technical Summary

Problems solved by technology

[0003] However, the current research based on the LSTM model has not effectively integrated the time-dependent features and the spatial correlation features of the pollutant particle concentration, and there is no high-precision forecast result for long-term missions, and the current research predicts PM 2.5 Concentration rarely has national-scale high-precision spatio-temporal forecast results

At present, related technologies only use the prediction of sites, but the number of existing monitoring sites is small and the distribution is sparse and uneven. Only the prediction of sites cannot meet the needs of applications. PM 2.5 Space-time continuous prediction plays an important role in various applications

Most current models ignore PM 2.5 Geographical distance between monitoring stations and the effect of spatial correlation on Tobler's First Law of Geography

Images