Sample Optimization Method for Air Quality Forecast

Abstract

The invention aims to solve the technical problem of providing an air quality forecast oriented sample optimization method which is high in forecast accuracy and good in adaptability. The method comprises the following steps: (1) determining relevant parameters of sample optimized screening, which is concretely implemented as follows: (1.1) determining a meteorological factor weight matrix by using a method for carrying out comprehensive analysis on pollutant concentrations and meteorological factors; (1.2) determining a threshold matrix by using an orthogonal experimental method; and (1.3) determining a sample size by using an experiment checking method; and (2) carrying out optimized screening on samples by using the determined parameters, which is concretely implemented as follows: (2.1) screening out samples with a meteorological similarity reaching a threshold range; (2.2) screening out samples with an overall meteorological similarity reaching the threshold range; and (2.3) screening out samples with a meteorological background which has the highest similarity with the meteorological background at the forecast date.

Description

technical field [0001] The invention belongs to the field of environment prediction, especially the field of air quality prediction. Background technique [0002] Since the 1990s, artificial neural networks have been applied in the field of air quality prediction. Artificial neural network plays a data-driven role in the process of air quality forecasting. The input layer neurons are calculated through the neural network structure and the output layer neurons are obtained. The neural network structure is adjusted and finalized through the learning process of historical data. [0003] With the continuous improvement of forecast accuracy requirements, research on the use and improvement of neural networks is also gradually developing. BP neural network is an artificial neural network that uses the error backpropagation algorithm. By continuously repeating the forward propagation and backward propagation processes until each training mode meets the end condition, the learnin...

AI Technical Summary

Problems solved by technology

However, when the improvement of the network structure and learning algorithm reaches a certain level, the improvement of the prediction accuracy of the air quality forecasting model will be limited.

Images