摘要
分布式水文模型往往考虑了研究流域的气候变化及降雨径流等水文特征要素,水文过程模拟的精度和范围不断提高,进而导致模型的计算量急剧增加,模型的计算速度慢、计算效率低等问题逐渐显露。GPU并行运算技术的快速发展使得普通计算机能够高效快速运行,文章基于GPU技术对水文资料的插值和研究流域的产汇流采取并行计算,利用非并行递归原理对马斯京根的河道汇流进行计算。结果表明:基于GPU和C语言相结合的降雨量插值计算效率明显高于普通CPU的计算效率,采用递归法进行马斯京根汇流演算的计算效率是利用产汇流次序表的0.6-1.38倍。
Distributed hydrological model considers always to study the hydrological characters elements including climate change and rainfall runoff of a catchment,and the precision and range of hydrological processes simulation increased constantly,leading to a sharp increase in the model calculation,the model problems of slow calculation velocity and low calculation efficiency gradually revealed. The rapid development of GPU parallel computing technology makes ordinary computer efficient operated in high-efficiency,this paper studied the interpolation of hydrological data and runoff of a catchment adoption and calculation based on CPU technology,using non-parallel recursion principle to calculate the Muskingum flow runoff. It appears that using combination of GPU and C Language to calculate the rainfall interpolation,the calculation efficiency is obviously higher than ordinary CPU alone,adopting the recursion method to calculate the Muskingum runoff,its calculation efficiency is 0. 6-1. 38 times equivalent with obtained with runoff sequence table.
作者
宏瑾靓
HONG Jin-Liang(Dalian Hydrology Bureau of Liaoning province, Dalia;Chin;116023)
出处
《黑龙江水利科技》
2018年第4期15-18,68,共5页
Heilongjiang Hydraulic Science and Technology
关键词
GPU并行运算
产汇流
分布式模型
大凌河
GPU parallel computing
runoff generation and confluence
distributed type model
DalingRiver