Quadratic Model of Air Quality Prediction

Abstract

It is necessary to predict the air pollutants under the condition of increasing air pollution. In this paper, the quadratic mathematical modeling of air quality prediction is mainly based on the primary forecast data and the measured data, and the daily concentration values of six conventional pollutants in the next three days are predicted. Firstly, based on the KNN regression principle, the optimal k-value quadratic prediction model suitable for A1, A2 and A3 monitoring points is established; Secondly, considering the influence of pollutant concentration between adjacent areas, a collaborative prediction model based on random forest algorithm and BP neural network is established to predict the pollutant concentration of four monitoring points A, A1, A2 and A3. At the same time, considering the different variables in the characteristic set of time and space input, a prediction data is introduced to adjust the parameters of the results of the model, spatiotemporal hybrid prediction model based on regression prediction is established. The final results show that the accuracy of prediction is improved based on the first prediction data and combined with the measured data.