Deep Learning Air Quality Forecast with Divided Area Based on K-means

Abstract

Aiming at determining the monitoring stations with strong spatial correlation, a method based on the K-means clustering algorithm for dividing the air quality monitoring stations is proposed. With Nantong as an example,the historical pollutant data in the target area and the meteorological data were gathered. With them, the hybrid CNNLSTM model, which is composed of the convolutional neural network(CNN) and the long short-termmemory(LSTM)neural network, was used to predict the pollutants, and finally extracted the temporal and spatial evolution characteristics of the pollutant concentration to complete the high accuracy of air quality forecast. The experiments show that, with the historical pollutant concentration data of other stations in the area divided by K-means, the CNN-LSTM model can forecast PM2.5 concentrations more accurately.

Keywords

• air quality
• forecast
• deep learning
• k-means clustering
• cnn-lstm model