基于WRF-CMQ-BP神经网络下空气质量的二次建模研究

A Secondary Modeling Study of Air Quality under WRF-CMQ-BP Neural Network Based

摘要: 随着工业化城市经济的飞速发展，大气污染严重危害着生态环境和人体健康。污染防治实践表明，建立空气质量预报模型，提前获知可能发生的大气污染过程并采取相应控制措施，是减少大气污染对人体健康和环境等造成的危害并提高环境空气质量的有效方法之一。本文针对污染物浓度对其进行预报，基于空气质量预报模式系统(WRF-CMAQ)等一次预报模型模拟结果，结合更多的实测数据源进行再建模，以提高预报的准确性，优化预报模型。在建立WRF-CMAQ-BP模型中，采用的数据是六种常规污染物的单日浓度，用模拟预测的值来逼近真实的值，从而达到较为精准的预报值。同时采用皮尔逊相关系数方法来找到影响空气质量的主要因素。

Abstract: With the rapid development of industrialized urban economy, air pollution is seriously endangering ecological environment and human health. The practice of pollution prevention and control shows that establishing air quality forecasting models to know the possible air pollution process in advance and take corresponding control measures is one of the effective methods to reduce the harm caused by air pollution to human health and environment and improve the ambient air quality. In this paper, we forecast pollutant concentrations based on the simulation results of primary forecasting models such as the Weather Research and Forecasting-Community Multiscale Air Quality (WRF-CMAQ), and remodel them with more actual measurement data sources to improve the accuracy of forecasting and optimize the forecasting models. In the WRF-CMAQ-BP model, the data used are the single-day concentrations of six conventional pollutants, and the predicted values are used to approximate the real values, so as to achieve more accurate forecast values. The Pearson correlation coefficient method is also used to find the main factors affecting air quality.