基于BP神经网络和模糊隶属度的PM2.5浓度校准
PM2.5 Concentration Calibration Based on BP Neural Network and Fuzzy Membership Degree
周云1
作者信息
- 1. 信阳航空职业学院,河南 信阳 464000
- 折叠
摘要
基于大量数据,采用Pearson相关系数与模糊隶属度分析自建点与国控点的细颗粒物(PM2.5)浓度数据相关性.其间通过建立反向传播(Back Propagation,BP)神经网络模型进行训练,并采用遍历试错法确定神经网络的最优算法与相关参数.经反复调试,校准结果相对于国控点数据的均方误差下降到 0.005,均等系数为 0.95,系统显示出优异的校准性能.研究结果表明,结合模糊隶属度预处理原始数据后,训练算法选用适宜、结构设定合理的BP神经网络能很好地校准自建点PM2.5 浓度数据,提高自建点数据精度.
Abstract
Based on a large amount of data,this paper uses Pearson correlation coefficient and fuzzy membership degree to analyze the correlation between fine particulate matter(PM2.5)concentration data of self built points and national control points.During this process,a Back Propagation(BP)neural network model is established for training,and the optimal algorithm and related parameters of the neural network are determined by using the ergodic trial and error method.After repeated debugging,the mean square error of the calibration results relative to the national control point data decreases to 0.005,and the equalization coefficient is 0.95,and the system shows excellent calibration performance.The research results indicate that after combining fuzzy membership degree preprocessing with raw data,selecting a suitable and structurally structured BP neural network for training can effectively calibrate the PM2.5 concentration data of self built points and improve the accuracy of self built point data.
关键词
PM2.5/浓度/BP神经网络/模糊隶属度/校准Key words
PM2.5 concentration/Back Propagation(BP)neural network/fuzzy membership degree/calibration引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据