Ozone concentrations prediction in Lanzhou, China, using chaotic artificial neural network

摘要

Among the important industrial cities in western China, Lanzhou has made great progress during recent years with its economic development policy. However, it has been accompanied by air pollution, particularly ozone (O-3) concentrations, which have adverse effects on the environment and human beings. Using Lanzhou as a research sample, chaotic artificial neural network (CANN) was used to establish an ozone concentrations prediction model. Meanwhile, several parameters were considered, such as the ozone concentrations by day at four monitoring stations from 2016 to 2018; sulfur dioxide (SO2), nitrogen dioxide (NO2), O-3, fine particulate matter with an aerodynamic diameter of 2.5 mu m or less (PM2.5), and fine particulate matter with an aerodynamic diameter of 10 mu m or less (PM10) concentrations over the same hours; temperature, humidity, instantaneous wind speed and direction. Seventy percent of the data were used to train CANN, 30% for testing, and back propagation (BP) network, artificial neural network (ANN), multiple linear regression (MLR) model were used for comparison. The results show that CANN has the smallest root mean squared error (RMSE) and a coefficient of correlation (R-2) near 1 compared with BP, ANN and MLR, and this means CANN can provide exact prediction result for pollution treatment in Lanzhou.