基于GA-BP神经网络的空气质量指数预测研究

Research on the Prediction of Air Quality Index Based on GA-BP Neural Network

近年来,随着经济的发展及城市规模的不断扩张,城市空气质量问题越来越严重.本文基于许昌市2015年1月—2017年4月AQI日均值数据以及同期气象数据,利用遗传算法改进的神经网络建立许昌市空气质量指数AQI预测模型.仿真结果表明:利用遗传算法改进的神经网络对许昌市AQI进行预测,平均相对误差在22%左右,预测准确率为80.44%,根据预测结果预测出的空气质量等级,预测准确率为82.5%,通过与BP神经网络进行对比分析,遗传算法改进的BP神经网络预测结果具有更高的准确率.该模型的建立,为空气质量的监测、预警与调控提供了科学依据.

In recent years,with the development of economy and the expansion of city scale,the air quality in cities is becoming more and more serious.This paper is based on the mean value of AQI in 2015-2017 and the meteorological data of Xuchang.The prediction model of Xuchang air quality index is established by using the improved neural network of genetic algorithm.The simulation results show that the average relative error is about 22% by the improved neural network using genetic algorithm to predict AQI in Xuchang City.The prediction accuracy is 80.44%.According to the predicted air quality grade,the prediction accuracy is 82.5%.BP neural network prediction results of the improved genetic algorithm has higher accuracy and precision than the BP neural network.The model provides a scientific basis for air quality monitoring,early warning and regulation.