Air quality prediction method based on convolution full-connection bidirectional gating circulation unit

Abstract

The invention discloses an air quality prediction method based on a convolutional full-connection bidirectional gating circulation unit, and the method comprises the steps: extracting the depth features contained in meteorological data through a convolutional neural network, learning the time sequence and coherence of the data through a cyclic neural network, and finally predicting the needed dataat the next moment through a full-connection layer. In the patent, because meteorological data has no high requirement for anti-rotation performance, a special pooling-layer-free convolutional network is selected to extract related features. According to the method, the air pollution index at the next moment can be accurately predicted in a large number of learning samples. On the whole, the method not only can predict a single index, but also can predict all indexes of input data characteristics, reasonable prediction can be carried out in combination with global characteristics to a certainextent, and the prediction function in the patent is realized.

Description

technical field [0001] The invention belongs to the field of deep learning, and in particular relates to an air quality prediction method based on a convolution fully connected bidirectional gated cyclic unit. Background technique [0002] With the rapid development of science and technology, artificial intelligence and deep learning have been widely used. At present, the methods of smog prediction can be roughly divided into three categories: the first category is statistical methods, the second category is traditional forecasting methods, and the third category is deep learning algorithms. Statistical methods include linear regression, gray system forecasting, Markov forecasting, etc. Most statistical models have certain requirements on data, and these models also have a relatively clear mathematical form, it is difficult to describe data with complex components with limited mathematical formulas. The traditional method is simple, the theory is mature, and it is easy to ...

AI Technical Summary

Problems solved by technology

Due to the high complexity and strong timing of meteorological data, it is difficult for a single CNN network or RNN network to adapt to the current forecasting needs

As for the proposed combination of CNN and RNN, without exception, these models cannot extract the required deep features for prediction.

Images