环境空气SO_2和NO_2浓度的GA_ANN预测模型研究

Study on GA_ANN based ambient air SO_2 and NO_2 concentration forecasting model

空气中污染物浓度的预测是一个复杂的非线性问题。国内外的研究表明神经网络能够比回归模型更好地预报空气污染物。设计并实现了将用于选择最优预报因子的遗传算法和神经网络算法相结合的GA_ANN空气质量预测模型,利用某市2003～2006年的数据建立神经网络空气质量预测模型,对该市2007年全年SO2和NO2的预测实验表明,GA_ANN模型比单纯的神经网络模型具有更高的预报精度。

The concentration prediction of the air pollutants is a complicated issue with nonlinear feature.Previous researches indicate that compared with the normal step wise regress method,the forecast precision with the ANN（Artificial Neural Networks） method is greatly improved.The GA_ANN air quality forecasting model is developed,which integrates Genetic Algorithm and Neural Network Algorithm.Genetic Algorithm is employed to select the most suitable forecasting factors for better performance. Some city air quality monitoring data and meteorological data from 2003 to 2006 are chosen to establish Neural Network of the GA_ANN model.The forecasting experiments about SO2 and NO2 of some city in 2007 show that the GA_ANN model has a higher forecasting precision than normal Neural Network model.