流域最佳管理措施情景优化算法的并行化
发布时间：2018-11-11 17:05
【摘要】：流域最佳管理措施(beneficial management practices,BMPs)情景优化问题是一个典型的复杂地理计算问题,目前所常用的BMPs情景优化算法需要结合流域模型进行大量的迭代运算,因而花费大量计算时间,难以满足实际应用的要求。本文针对目前代表性的BMPs情景优化算法——ε支配多目标遗传算法(ε-NSGA-II),采用主从式并行策略,利用MPI并行编程库实现了该优化算法的并行化。在江西省赣江上游的梅川江流域(面积为6 366km2)进行BMPs情景优化的应用案例表明,并行化的优化算法当运行于集群机时,加速比随着核数(8~512核)的增加而递增,当核数为512时,加速比达到最大值(310);并行效率随着核数的增加逐渐下降,最高值0.91,最低值0.61,取得了明显的加速效果。
[Abstract]:Optimal watershed management (beneficial management practices,BMPs) scenario optimization problem is a typical complex geographical calculation problem. The commonly used BMPs scenario optimization algorithm needs a large number of iterations combined with watershed model. Therefore, it takes a lot of computing time and is difficult to meet the requirements of practical application. In this paper, the BMPs scenario optimization algorithm 蔚 -dominated multi-objective genetic algorithm (蔚 -NSGA-II) is used to realize the parallelization of the optimization algorithm by using the MPI parallel programming library and the master-slave parallel strategy. The application of BMPs scenario optimization in the Meichuan River basin (area 6 366km2) in the upper reaches of Ganjiang River in Jiangxi Province shows that the speedup ratio of the parallel optimization algorithm increases with the increase of the number of kernels (8512cores) when it runs on the cluster computer. When the kernel number is 512, the speedup ratio reaches the maximum (310). The parallel efficiency decreases gradually with the increase of the number of kernels, the highest value is 0.91and the lowest value is 0.61.The acceleration effect is obvious.
【作者单位】： 中国科学院地理科学与资源研究所;中国科学院大学;加拿大圭尔夫大学地理系;南京师范大学地理科学学院;美国威斯康星大学麦迪逊分校地理系;
【基金】：国家863计划(2011AA120305) 国家科技支撑计划(2013BAC08B03-4) 国家水专项计划(2013ZX07103006-005)~~
【分类号】：TV213.4
，
本文编号：2325555
[Abstract]:Optimal watershed management (beneficial management practices,BMPs) scenario optimization problem is a typical complex geographical calculation problem. The commonly used BMPs scenario optimization algorithm needs a large number of iterations combined with watershed model. Therefore, it takes a lot of computing time and is difficult to meet the requirements of practical application. In this paper, the BMPs scenario optimization algorithm 蔚 -dominated multi-objective genetic algorithm (蔚 -NSGA-II) is used to realize the parallelization of the optimization algorithm by using the MPI parallel programming library and the master-slave parallel strategy. The application of BMPs scenario optimization in the Meichuan River basin (area 6 366km2) in the upper reaches of Ganjiang River in Jiangxi Province shows that the speedup ratio of the parallel optimization algorithm increases with the increase of the number of kernels (8512cores) when it runs on the cluster computer. When the kernel number is 512, the speedup ratio reaches the maximum (310). The parallel efficiency decreases gradually with the increase of the number of kernels, the highest value is 0.91and the lowest value is 0.61.The acceleration effect is obvious.
【作者单位】： 中国科学院地理科学与资源研究所;中国科学院大学;加拿大圭尔夫大学地理系;南京师范大学地理科学学院;美国威斯康星大学麦迪逊分校地理系;
【基金】：国家863计划(2011AA120305) 国家科技支撑计划(2013BAC08B03-4) 国家水专项计划(2013ZX07103006-005)~~
【分类号】：TV213.4
，
本文编号：2325555
本文链接：https://www.wllwen.com/kejilunwen/shuiwenshuili/2325555.html
