改进的二元蚁群算法结合分形理论预测雾霾天气形成的关键因子

发布时间：2018-03-28 12:20

  本文选题：雾霾　切入点：分形理论　出处：《系统科学与数学》2017年02期

【摘要】：随着工业化进程的加剧,雾霾已严重影响到人类的日常生活,分析天气因素进而得出影响雾霾天气的关键因子尤为重要.预测雾霾天气形成的关键因子是一个不断剔除冗余因素保留关键要素的过程,每一个天气因素都有两种状态,被选中为关键因子与否,文章根据该特点,从一维细胞自动机入手,提出了一种以二元蚁群算法作为搜索策略,分形理论作为子集评估度量准则的混合方法.因二元蚁群算法前期信息素匮乏需要较长搜索时间,引入二元粒子群算法对其进行优化,将粒子经过多次迭代之后得到的最优位置通过模糊函数映射成蚂蚁所需的信息素,在较短的时间内形成一条信息素落差明显的路径,缩短算法前期运行时间.最后将所用方法应用于北京,广州和上海三地雾霾天气关键影响因子的预测中,并结合10-交叉验证和SVM算法对预测结果分类准确率进行分析,通过与其它算法进行对比,结果表明文章算法预测结果具有较高可信度,为后期的雾霾治理工作提供了重要的参考依据.
[Abstract]:With the aggravation of industrialization, haze has seriously affected the daily life of human beings. It is very important to analyze the weather factors and find out the key factors that affect the haze weather. The key factor to predict the formation of haze weather is a process in which redundant factors are continuously removed and the key elements are retained. Each weather factor has two states. According to this feature, a binary ant colony algorithm is proposed as a search strategy based on one-dimensional cellular automata, which is selected as the key factor or not. Fractal theory is used as a mixed method for subset evaluation metric. Because the lack of pheromone in the early stage of binary ant colony algorithm requires a long search time, the binary particle swarm optimization algorithm is introduced to optimize it. The optimal position of particles after several iterations is mapped to pheromone needed by ants through fuzzy function, and a path with obvious pheromone drop is formed in a short time. Finally, the method is applied to predict the key weather factors of haze in Beijing, Guangzhou and Shanghai, and the classification accuracy of the forecast results is analyzed by combining 10-cross validation and SVM algorithm. Compared with other algorithms, the results show that the prediction results of this paper have high reliability, which provides an important reference for the later work of haze governance.
【作者单位】： 合肥工业大学管理学院;教育部过程优化与智能决策重点实验室;新加坡南洋理工大学计算智能中心实验室计算机工程学院;
【基金】：国家自然科学基金(71271071,71301041,71490725);国家自然科学基金重大培育项目(91546108) 国家云制造主题项目(2015AA042101) 安徽省教育厅自然科学研究项目(KJ2013Z089)资助课题
【分类号】：TP18;X513
，

本文编号：1676299