摘要
针对目前挥发性有机物(Volatile Organic Compounds,VOCs)质量浓度插值方法单一且插值精度不高的问题,为了提升区域内VOCs质量浓度空间插值的精度,构建BOA-ELM(Butterfly Optimization Algorithm-Extreme Learning Machine)空间插值模型,首次将神经网络模型应用于VOCs质量浓度空间插值。首先对研究区域进行网格划分,其次利用BOA-ELM进行插值研究,同时讨论气象特征对空间插值的重要性,最后将VOCs质量浓度空间插值结果可视化。以陕西省为例,对VOCs质量浓度进行插值,结果显示:加入气象特征变量能提高模型插值精度,且与原始极限学习机(Extreme Learning Machine,ELM)模型以及传统插值方法反向离权重法(Inverse Distance Weighted,IDW)对比,显示BOA-ELM模型的均方根误差(ZRMSE)、平均绝对误差(ZMAE)、平均绝对百分比误差(ZMAPE)均为最小值,分别为8.69μg/m^(3)、6.72μg/m^(3)、7.10%,优于IDW模型。结果表明BOA-ELM模型能很好地应用VOCs质量浓度空间插值,为大气污染物的空间插值提供了新的思路与方法。
To improve the accuracy of spatial interpolation of VOCs mass concentration in the region under the problems of a single VOCs mass concentration interpolation method and low interpolation accuracy,the neural network model was applied to spatial interpolation of VOCs mass concentration in this paper,and a spatial interpolation model of VOCs mass concentration was established based on Butterfly Optimization Algorithm(BOA)and Extreme Learning Machine(ELM).Firstly,the selected research area was divided into 10 km×10 km grids to meet the needs of VOCs fine management.Secondly,the Mean Square Error(MSE)was used as the fitness value of BOA fitness function to optimize the weights and thresholds randomly set by the ELM model.The BOA ELM model was constructed to study the spatial interpolation of VOCs mass concentration,and the importance of meteorological characteristics for spatial interpolation was discussed by inputting different characteristic variables,to visualize the spatial interpolation results of VOCs mass concentration.The spatial interpolation of VOCs mass concentration was studied,taking Shaanxi Province as an example.The results show that the inclusion of meteorological characteristic variables can significantly improve the interpolation accuracy of the model.Compared with the original ELM model and the Inverse Distance Weight Method(IDW)in the traditional interpolation method,the BOA ELM model has the best interpolation effect.The Root Mean Square Error of the BOA ELM model is 8.69μg/m^(3),which is 1.81μg/m^(3) and 3.32μg/m^(3) lower than the ELM model and IDW method,respectively.The Mean Absolute Error is 6.72μg/m^(3),which is 1.42μg/m^(3) and 2.99μg/m^(3) lower than the two comparison methods,respectively.The Mean Absolute Percentage Error is 7.10%,which is 1.70%and 3.84%lower than the two comparison methods,respectively.The experimental results show that the BOA ELM model is feasible for spatial interpolation of VOCs mass concentration,and the interpolation accuracy is higher than that of the traditional spatial interpolation method,which provides a new idea and method for spatial interpolation of air pollutants.
作者
黄光球
虞欣
陆秋琴
HUANG Guangqiu;YU Xin;LU Qiuqin(School of Management,Xi'an University of Architecture and Technology,Xi'an 710055,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2023年第9期3362-3371,共10页
Journal of Safety and Environment
基金
国家自然科学基金项目(71874134)。
