基于PSO优化与非对称信息扩散的海表温度插值算法  

SST Interpolation Algorithm based on PSO Optimization Algorithm and Asymmetric Information Diffusion

作　　者：王敏 石明航 洪梅[3] 谷文杰 黎永顺 郭晓峰 WANG Min;SHI Minghang;HONG Mei(School of Electronics and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044;School of Electronics and Information Engineering,Anhui Jianzhu University,Hefei 230601;Institute of Meteorology and Oceanography,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]南京信息工程大学电子与信息工程学院,江苏南京210044 [2]安徽建筑大学电子与信息工程学院,安徽合肥230601 [3]国防科技大学气象海洋学院,湖南长沙410073

出　　处：《浙江海洋大学学报(自然科学版)》2024年第6期496-503,共8页Journal of Zhejiang Ocean University:Natural Science

基　　金：国家自然科学基金项目(41775165,41775039);安徽省高校杰出青年科研项目(2023AH020022);南京信息工程大学人才启动经费资助项目(2021r034)

摘　　要：针对如何利用稀疏样本填补数据缺失造成的空白问题,根据信息扩散理论,提出了一种基于粒子群优化算法(particle swarm optimization, PSO)与非对称信息扩散相结合的插值算法。在信息扩散插值的基础上,对经验窗宽进行优化并与非对称扩散函数结合,解决了信息扩散插值法对非正态资料插值不精准的问题。以西北太平洋2019年的月均海表温度作为研究对象,选取不同样本容量的海表温度作为已知数据,分别采用克里金插值、正态信息扩散、非对称信息扩散、PSO与非对称信息扩散结合4种算法进行插值试验。结果表明,在已知样本容量为30的情况下,对4个月份的插值误差取平均值,可知所提出算法的均方根误差为0.979,平均绝对误差为0.623,在4种方法中误差最小;在样本容量为100的情况下,所提出算法的均方根误差为0.735,平均绝对误差为0.430,同样为最小误差。故提出的插值算法相较于其他插值方法,在样本稀疏情况下取得了更好的效果,可为海表温度以及其他类似稀疏样本提供切实有效的技术基础。Aiming at the problem of filling gaps caused by missing data with sparse samples,this paper proposes an interpolation algorithm based on particle swarm optimization(PSO)and asymmetric information diffusion according to information diffusion theory.On the basis of information diffusion interpolation,the algorithm optimizes the empirical window width and combines it with an asymmetric diffusion function to solve the problem of inaccurate interpolation for non-normal data in information diffusion interpolation methods.Taking the monthly mean sea surface temperature(SST)of offshore China in 2019 as the research object,SST data with different sample sizes were selected as known data.Interpolation experiments were carried out using four algorithms:Kriging interpolation,normal information diffusion,asymmetric information diffusion,and the combination of PSO and asymmetric information diffusion.The results show that when the sample size is 30,the proposed algorithm has the lowest interpolation error among the four methods,with an average root mean square error(RMSE)of 0.979 and an average absolute error(MAE)of 0.623 across four months.When the sample size is 100,the proposed algorithm again achieves the lowest errors,with an average RMSE of 0.735 and an average MAE of 0.430.Therefore,the proposed interpolation algorithm performs better in the case of sparse samples compared to other interpolation methods,providing a practical and effective technical basis for SST and other similar sparse sample data.

关 键 词：稀疏样本 正态信息扩散 海表温度 克里金插值 PSO优化算法 非对称信息扩散 

分 类 号：P731.1[天文地球—海洋科学]