Air pollution prediction method

Abstract

The invention provides an air pollution prediction method. The method comprises the following steps: (1) acquiring air pollution meteorological data; (2) converting the air pollution meteorological data into a pixel matrix and performing data filling; (3) carrying out time-space domain-oriented spatial-temporal feature unified modeling through a three-dimensional convolutional neural network model; (4) taking the output of the three-dimensional convolutional neural network model as the input part of a convolutional long-short term memory network, and carrying out long-time and short-time dependent modeling; (5) generating an air pollution prediction model based on a space-time dynamic advection method; and (6) optimizing the air pollution prediction model through an orthogonal regularization algorithm, and then carrying out environment prediction. The decoupled three-dimensional convolution is used to carry out spatial-temporal feature unified modeling, the representation capability of the spatial-temporal feature extraction is enhanced, the fusion of the spatial-temporal features is truly realized, and the training cost is reduced and the training speed is improved while the spatial-temporal convolution performance is improved.

Description

technical field [0001] The invention relates to the technical field of environmental protection, in particular to an air pollution prediction method. Background technique [0002] In the field of air pollution prediction, some scholars regard air data as traditional time series data, and use the least square method or neural network to fit the data change curve; In addition, some scholars use deep learning methods combining convolutional neural networks and recurrent neural networks to predict air pollution. [0003] The shortcoming of treating air data as traditional time series data is that it lacks consideration of spatiotemporal factors. Air occupies a certain space and changes with time, so the data has typical spatiotemporal attributes, including high-dimensional spatial correlation. Sexuality and temporal correlation, and dynamic spatiotemporality. Since neural network and machine learning methods are shallow integration methods, they cannot handle high-dimensional ...

AI Technical Summary

Problems solved by technology

[0003] The shortcoming of treating air data as traditional time series data is that it lacks consideration of spatiotemporal factors. Air occupies a certain space and changes with time, so the data has typical spatiotemporal attributes, including high-dimensional spatial correlation. Sexuality and Temporal Correlation and Dynamic Spatiotemporality

Since neural network and machine learning methods are shallow integration methods, they cannot handle high-dimensional information

[0004] The disadvantage of the knowledge-driven mechanism model is that it is limited by factors such as single knowledge, small amount of knowledge, and lack of knowledge representation technology, so it is impossible to obtain all knowledge completely and accurately.

[0005] The disadvantage of the deep learning method that combines the convolutional neural network and the cyclic neural network is that it cannot describe the dynamic interaction between the spatial structure and time, and the deep network is difficult to train.

The convolutional network is almost independent when modeling the spatial structure, and is less affected by the recurrent network; similarly, the recurrent network is relatively independent when modeling the temporal dynamics, because the spatial correlation is difficult to extract in the feature vector of the convolutional network. fully expressed in

However, air data has a tightly coupled spatiotemporal correlation, and the deep network is very easy to fall into local optimum and gradient disappearance. The problem of gradient explosion makes the network difficult to train.

Images