基于扩展场强模型的稀疏AQI空间插值新算法

发布时间：2018-02-04 08:16

本文关键词： 空气质量指数空间插值稀疏数据场强模型　出处：《武汉大学学报(信息科学版)》2017年07期 　论文类型：期刊论文

【摘要】：针对空气质量指数(AQI)监测点分布稀疏,现有空间插值算法精度不高问题,提出了新的扩展场强模型与算法。扩展场强单参数模型引入参数c控制场强衰减程度,通过c与误差关系图并借助二分查找法计算最优c值。扩展场强双参数模型加入参数k调整场强影响范围,通过c、k与误差关系图并借助迭代双线性插值法求解最优c、k组合。以北京、天津、武汉、郑州四个城市2014-08~2015-04的50组AQI监测值为实验数据,采用交叉验证法并以RMSE、AME、PAEE为评价指标,实现了单参与双参模型及参数选取,然后与克里金法及类似的反距离加权法进行对比。实验证明,扩展场强模型能够得到针对稀疏AQI的更高插值精度,且双参数模型精度高于单参数模型。本文算法适用于采样点数目与位置均固定的稀疏数据插值,并可推广至其他类型与维度的空间数据。
[Abstract]:Because of the sparse distribution of air quality index (AQI) monitoring points, the accuracy of existing spatial interpolation algorithms is not high. A new extended field intensity model and algorithm is proposed. The single parameter model of extended field strength introduces parameter c to control the attenuation of field strength. The optimal value of c is calculated by using the graph of c and error and the optimum value of c is calculated by means of binary search method. The parameter k is added to the model of extended field strength to adjust the range of influence of field strength, and the influence range of field strength is adjusted by c. K and the error relation diagram and the iterative bilinear interpolation method are used to solve the optimal combination of cnk. Beijing, Tianjin, Wuhan. The 50 groups of AQI monitoring data from 2014-08 to 2015-04 in four cities of Zhengzhou were used as experimental data. Cross validation method and RMSE AME-PAEE were used as the evaluation index. The single-participating double-parameter model and parameter selection are realized, and then compared with the Kriging method and the similar inverse distance weighting method. The experimental results show that the extended field strength model can obtain higher interpolation accuracy for sparse AQI. The accuracy of the two-parameter model is higher than that of the single-parameter model. This algorithm is suitable for the sparse data interpolation where the number and position of the sampling points are fixed and can be extended to other spatial data types and dimensions.
【作者单位】：武汉大学计算机学院;中国空间技术研究院;武汉大学资源与环境科学学院;
【基金】：中国空间技术研究院创新基金(2014) 装备预研基金(9140A27040414JB11078) 湖北省科技支撑计划(2014BAA149)~~
【分类号】：X51;X831
【正文快照】： 近年来空气污染增多且危害加重,因此空气污染监测与预报已成为关系国计民生的大事。我国已用空气质量指数(AQI)替代原有的空气污染指数(API),且针对单项污染物还规定了分指数,参与AQI评价的主要污染物为PM2.5、PM10、SO2、NO2、O3、CO。目前,AQI值只能通过分布稀疏的气象站点

【相似文献】