Air pollutants concentrations prediction model based on adaptive artificial neural network
This paper proposes an air pollutants concentrations prediction model based on adaptive artificial neural network.Two main approaches are put forward to improve the prediction accuracy.One is sample adaptive optimization,which is based on the criterion of meteorological similarity with threetiered screening mechanism; the other is GA-ANN,combining genetic algorithm to ANN,which is built to make the prediction model more adaptive.These two approaches achieve good results as they mainly focus on predicting day by eliminating the impact of invalid situation.A case study is carried out in Guangzhou and Hong Kong to demonstrate the prediction capabilities and applicability in contrast with the conventional modeling approach.The results show that the mean prediction absolute error of SO2,PM10 and NO2 is 0.015 mg/m3,0.023 mg/m3 and 0.014 mg/m3 respectively in Guangzhou,while 0.011 mg/m3 on average in Hong Kong.
air pollutant concentration prediction sample optimization genetic algorithm BP neural network
L.Li P.Wang M.Cai Y.H.Liu
Cruangzhou Institute of Energy Conversion, Chinese Academy of Sciences, China School of Engineering, Sun Yat-sen University, China
国际会议
香港
英文
181-183
2012-12-01（万方平台首次上网日期，不代表论文的发表时间）