Prediction model of air pollutant concentration based on deep neural network: A case study of Fushun, Liaoning Province

Abstract

With the rapid development of the economy, the sources of air pollution are increasing, and the problem of air pollution is becoming more and more serious. Air quality prediction is a very effective means of predicting air pollution in the future, which helps the government regulatory authorities to provide early warning and protect people’s physical and mental health. In this paper, a prediction model of air pollutant concentration based on deep neural network is proposed. With the concentration of PM2.5 as the prediction target, the neural network chooses bidirectional Long Short-Term Memory (LSTM) and fully connected neural network. First, historical meteorological data and the PM2.5 density from 2016 to 2017 were used as training data in this paper, which were obtained from website. Then, after pre-processing the input data, the data is transmitted to the network and trained multiple times to obtain network parameters that make the prediction effect better. Next, the network model is applied to the test set, and the test results are compared with the actual values to measure the prediction effect. Finally, by comparing with other prediction models, the results show that the proposed model performs better and has higher accuracy.