Atmospheric ozone concentration prediction method based on mixed CNN-Transformer model

Abstract

The invention relates to an atmospheric ozone concentration prediction method based on a hybrid CNN (convolutional neural network)-Transform model, and designs a hybrid CNN-Transform model, and the model is formed by combining a convolutional neural network and an improved Transform model. The convolutional neural network is composed of two one-dimensional convolutional layers, the Transform model is composed of encoders and decoders, each encoder is provided with three encoding layers, and the decoders are provided with three decoding layers. Meanwhile, on the basis of a traditional Transform encoder decoder architecture, a cross multi-head attention layer from encoder to encoder is added between different encoding layers of the encoder, and association of encoding information between the different encoding layers is further mined. The CNN model can well extract effective information on the feature dimension, and the problem that in a Transform model, the information extraction capacity of an encoder is insufficient is solved. According to the prediction method, the influence of multivariate data on the ozone concentration can be reflected more truly, and the influence mode is learned through the CNN-Transform model, so that a more accurate ozone concentration prediction result is given.

Description

technical field [0001] The present invention relates to atmospheric ozone concentration prediction method, relate in particular to a kind of atmospheric ozone concentration prediction method based on hybrid CNN-Transformer model. Background technique [0002] Currently, accelerated industrialization and urbanization have led to a sharp decline in air quality. The main pollutants in the atmosphere include sulfur dioxide, nitrogen oxides, particulate matter, carbon monoxide and ozone. Long-term exposure to too high or too low concentrations of ozone will cause some physical hazards, such as difficulty breathing. Considering the impact of ozone on people's bodies, there is an urgent need for a method that can accurately predict the concentration of ozone so that relevant agencies can take measures to control it. [0003] With the development of artificial intelligence and deep learning, using deep learning models to predict the concentration of pollutants in the atmosphere ha...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a method for predicting atmospheric ozone concentration based on a hybrid CNN-Transformer model, which can accurately and effectively predict atmospheric ozone concentration, and solve the problem of poor multivariate data extraction capabilities in the traditional Transformer architecture. The method can extract The relationship between different features in multivariate data, making full use of historical data to make accurate predictions of future ozone concentrations

Images