基于人工神经网络模型的臭氧生成灵敏度预测

Artificial neural network models based prediction for ozone-NO_x-HC sensitivity

本文将臭氧生成灵敏度指标Φ=kHC+OH[HC]/kNOx+OH[NOx]及其影响参数作为输入参数,建立了人工神经网络模型,预测了日本海利尻岛海洋大气边界层臭氧生成的控制因素及其灵敏度特征。预测结果与化学模式计算结果基本一致,证明文中所建立的人工神经网络模型具有较好的预测性能。研究结果表明,人工神经网络模型可以作为一个有效的工具用于对流层臭氧生成灵敏度特征的确定及相关的臭氧消除策略制定。

We constructed a back-propagation neural network （BPNN） model with a P（O3） sensitivity indicator, Ф=KHC＋OH[HC]/KNOx＋OH[NOx], and influencial factors as its input variables. We employed the model to predict the control factors of ozone formation and its sensitivity character at ocean atmospheric boundary layer of Japan Hailikao island. Prediction results basically conformed with chemical computation results. It indicates that this model has better prediction performance. Our research demonstrates that artificial neural network （ANN） can serve as an effective tool to determine the sensitivity of P（O3） and to formulate ozone abatement strategy.