Multivariable air quality time sequence prediction method based on multi-task multi-channel wavelet transform nested long and short term memory model

Abstract

The invention discloses an air quality time sequence prediction method based on a multi-task multi-channel wavelet transform nested long and short term memory model (MTMC-WNLSTM). The method comprises the following steps: collecting six pieces of air quality index data (PM2.5, PM10, SO2, NO2, CO and O3), recording the concentration value of each index at each collection point in sequence according to a time sequence, obtaining one-dimensional time sequence data, and carrying out the standardization; decomposing single-dimensional time sequence data of each variable in original data into four data subsequences through wavelet transform; expanding four data subsequences obtained by decomposing the single-dimensional time sequence from single-dimensional time sequence data into multi-dimensional data by using a sliding time window method; dividing the data into a training set and a test set; and establishing a multi-task multi-channel NLSTM deep learning model, inputting the training set to train and optimize the model, inputting the test set to test the performance of the model after the training is completed, and finally predicting six air quality indexes by adopting the obtained deep learning model.

Description

technical field [0001] The invention relates to the technical field of environmental protection, and more specifically, relates to a multi-task multi-channel wavelet transform nested long short-term memory model (MTMC-WNLSTM) multivariate air quality time series prediction method. Background technique [0002] Due to industrialization and urbanization, more and more areas of the world are threatened by air pollution. Poor air environment not only destroys the ecological balance of nature, but also affects our living environment and restricts economic and social development. Common air pollutants include 6 main components: fine particulate matter (PM2.5), respirable particulate matter (PM10), sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO) and ozone (O3). PM2.5 and PM10 are mainly suspended particles from human production and domestic waste, with very complex physical and chemical components, which seriously affect health after inhalation and can cause res...

AI Technical Summary

Problems solved by technology

Changes in a single variable can lead to multi-chain reactions and subsequent lasting effects

Therefore, the current air quality prediction model still has problems such as low prediction accuracy, poor generalization, and prediction lag.

Images