Deep-structure recurrent neural network-based PM2.5 prediction method

Abstract

The invention discloses a deep-structure recurrent neural network-based PM2.5 prediction method. According to the method, construction a deep-structure PM2.5 prediction model is constructed accordingto deep learning and a recurrent neural network theory by utilizing acquired mass data; and through data feature extraction and training, hazy weather prediction is realized. The method aims at improving the haze prediction efficiency and precision and providing convictive decision basis for haze prevention and treatment. The prediction model does not requirements for a data structure and can carry out self-learning when data is big enough, so that deep learning is suitable for the requirements of the present internet big data applications.

Description

technical field [0001] The invention belongs to the technical field of environmental engineering and detection, and more specifically, relates to a PM2.5 prediction method based on a deep structure recurrent neural network. Background technique [0002] Air quality has always been a major issue related to the future and destiny of mankind. With the progress of society and the sharp increase in car ownership, the content of inhalable particulate matter in the air has risen sharply, and the problem of environmental pollution has become increasingly serious. With the continuous deterioration of air quality, there are more and more haze weather phenomena, and the harm is getting bigger and bigger. Smog is a disaster weather phenomenon. Inhalable particulate matter PM2.5 is the main cause of smog weather, which not only has a serious impact on air quality, but more importantly, poses a huge threat to human health. [0003] There are many ideas and methods for the prediction of ...

AI Technical Summary

Problems solved by technology

[0004] Due to the influence of a large number of uncertain and complex factors such as climate, temperature, and human activities, the time series of various weather data is highly nonlinear and uncertain, and it is difficult for conventional analysis and prediction methods to grasp the changes. Regularity and Variation Characteristics

[0005] The shallow neural network is effective in solving simple or more restricted problems, but due to limited modeling and representation capabilities, it has limited ability to realize some more complex problems involving natural signals in real life.

Images