基于神经网络的空气质量预测模型构建研究

Research on Construction of Air Quality Forecast Model Based on Neural Network

污染防治作为决胜全面建成小康社会的三大攻坚战之一,相关单位及部门应加以重视。为了打赢这场蓝天保卫战,包头市于2018年制订了《包头市打赢蓝天保卫战三年行动计划实施方案》,方案中明确指出经过3年的攻坚,到2020年颗粒物浓度较2015年下降20%,空气质量优良天数比率达到80%,为了实现预定的目标,进一步了解包头市的大气污染情况,有必要了解空气变化趋势,并根据之前的监测数据及时、准确、全面地预测未来的空气污染。笔者主要利用TensorFlow软件库构建基于时间序列进行预测的LSTM(长短期记忆网络)网络结构,对包头市主要的污染源PM2.5浓度进行精确地预测。研究成果可以为相关大气污染预防及防治工作提供数据支持和新的预测方法。

Pollution prevention and control as one of the three key battles to win the victory of building a well-off society in an all-round way,relevant units and departments should pay attention to it.In order to win the blue sky defense war,Baotou City formulated the three-year action plan implementation plan for Baotou city to win the blue sky defense war in 2018.The plan clearly points out that after three years of hard work,by 2020,the concentration of particulate matter will be 20%lower than that in 2015,and the ratio of days with good air quality will reach 80%.In order to achieve the predetermined goal and further understand the air pollution situation of Baotou City,it is necessary to It is necessary to understand the trend of air change and predict the future air pollution timely,accurately and comprehensively according to the previous monitoring data.The author mainly uses tensorflow software to build the network structure of LSTM(short and long term memory network)based on time series prediction,and accurately predict the PM2.5 concentration of the main pollution source in Baotou city.The research results can provide data support and new prediction methods for air pollution prevention and control.