PM2.5 Forecasting Based on Artificial Neural Network and Genetic Algorithm

In order to predict PM2.5 concentrations value more accurately, we propose a new prediction technique based on BP artificial neural network model and the multi-population quantum genetic algorithm (this new technique is shorted as MP-QGABP) to improve the PM2.5 concentrations value forecasting system. The MP-QGA-BP model is compared with the traditional BP artificial neural network model for daily maximum of PM2.5 concentrations value forecasting. When we use the real-time monitoring time-series data sample including PM10, CO, NO2 and SO2 pollutant data which are closely related to the value of PM2.5 concentrations value and Mat lab language to program, the result of simulation has shown that the model established by MP-QGABP artificial neural network has a smaller training and predicting error and a better general ability than the model established by traditional BP artificial neural network.