WRF-CMAQ-BP神经网络空气质量预报模型研究

Study on Air Quality Forecast Model of WRF-CMAQ-BP Neural Network

为减轻大气污染给人类带来的危害,国内外的学者在上个世纪就开始研究空气预报模型,高效且准确性高的预测模型可以预测未来若干天的大气污染情况,人们可以据此做出有效的应对措施以减少大气污染带来的危害。目前常用WRF-CMAQ模拟体系对空气质量进行预报,但由于受到各种不确定因素的影响,WRF-CMAQ预报模型的结果并不理想。本文主要工作是在WRF-CMAQ预报模型的基础上二次建模,提出了两个模型分别为WRF-CMAQ-BP和WRF-CMAQ-XGBoost模型,在构建网络后进行测试,在调参过程中发现WRF-CMAQ-BP模型的效果是更优的,因此本文用WRF-CMAQ-BP模型预测六种常规污染物的单日浓度值。

In order to alleviate the harm brought about by atmospheric pollution, scholars at home and abroad began researching the air forecast model, high-efficiency and highly high prediction models can predict the future of atmospheric pollution in the next few days, people can make effective response measures to reduce the harm caused by atmospheric pollution. At present, the WRF-CMAQ simulation system is used to predict air quality, but the results of WRF-CMAQ forecasting model are not ideal due to the influence of various uncertain factors. This paper mainly works two models on the basis of the WRF-CMAQ forecast model, and two models are proposed to be WRF-CMAQ-BP and WRF-CMAQ-XGBOOST models, and test after building networks. During the adjustment process, it is found that the WRF-CMAQ-BP model is better, so the WRF-CMAQ-BP model is used to predict the single concentration value of six conventional pollutants.