Multi-model adaptive atmospheric PM2.5 concentration prediction method based on small samples

Abstract

The invention discloses a multi-model adaptive atmospheric PM2.5 concentration prediction method based on small samples. The method comprises the steps of: carrying out the real-time online data collection of the atmospheric pollutant concentration of a campus all the year round through a solar power supply system, an outdoor monitoring probe and a cloud network, carrying out the data normalization processing of the collected atmospheric pollutants, using a least square support vector regression method for training collected atmospheric pollutant data in four seasons to obtain atmospheric PM2.5 concentration prediction models in different seasons, taking the collected data as input variables of the prediction models, calculating to obtain atmospheric PM2.5 concentration prediction values in different seasons by using the different prediction models according to algorithms of the prediction models, predicting and comparing a prediction value with actually collected data, wherein the obtained error performance index is the minimum in several models, and adopting the prediction model with the small error to predict the concentration of the atmospheric pollutant PM2.5 in the current state. Therefore, the requirement that a single model cannot deal with the capability of predicting the concentration of the atmospheric pollutant PM2.5 in the current state is met.

Description

technical field [0001] The invention relates to feature extraction, machine learning, multi-model adaptive, PM 2.5 Concentration Prediction. Background technique [0002] At present, the forecasting methods that are relatively mature in technology and widely used are generally time series method, multiple linear regression method, neural network and other methods. Most of these methods establish the linear or nonlinear atmospheric PM in the current state 2.5 Concentration prediction model to achieve PM 2.5 Concentration Prediction. But in the actual atmospheric environment, PM 2.5 Concentrations are easily affected by environmental changes, and the single prediction model established in the current state is often difficult to meet the requirements of PM in the atmospheric environment. 2.5 Concentration changes, which will lead to the decline of the prediction accuracy of the established model or even invalidation. [0003] At present, the forecasting methods that are r...

AI Technical Summary

Problems solved by technology

And for different environments to establish the corresponding prediction model for adaptive switching to achieve the PM 2.5 Concentration prediction methods are not widely used

Images