基于深度学习的空气质量预测方法研究
首发时间:2019-03-04
摘要:随着深度学习的进一步发展,该项技术也日益改善着人类的生活方式。在空气质量预测方面,可观测的数据量呈现几何倍数的大幅增长,传统的时间预测方法已不堪众任,在本文中,根据空气质量数据与气象数据来预测每个监测站未来48小时的空气质量,利用了深度学习的方法对空气质量预测问题进行研究与实现,考虑到空气污染物的空间相关性,将空间稀疏空气质量数据转换为一致输入,以模拟污染源,以此基于LSTM网络与GRU网络提出了空域数据与时域数据相结合的预测方法,消弱了地域因素对空气质量预测带来的误差,提升了预测的准确精度。
For information in English, please click here
Research on Air Quality Prediction Method Based on Deep Learning
Abstract:With the further development of deep learning, this technology is also improving the human lifestyle. In terms of air quality prediction, the amount of observable data shows a large increase in geometric multiples. The traditional time prediction method is unsuccessful. In this paper, based on air quality data, meteorological data predicts the air quality of each monitoring station for the next 48 hours. Using the deep learning method to study and realize the air quality prediction problem, considering the spatial correlation of air pollutants, the spatial sparse air quality data is converted into a consistent input to simulate the pollution source, based on the LSTM network and the GRU network. A prediction method combining spatial data with time domain data is proposed, which reduces the error caused by regional factors on air quality prediction and improves the accuracy of prediction.
Keywords: Deep learning Spatial correlation Time domain data
基金:
引用
No.****
同行评议
勘误表
基于深度学习的空气质量预测方法研究
评论
全部评论0/1000