基于GPU加速的污染物输移高分辨率数值模型

High-resolution numerical model of pollutant transport based on GPU acceleration

为了高效准确地模拟水污染事件中污染物输移过程,该文引入了一套基于GPU加速的水动力及污染物输移的GAST(GPU Accelerated Surface Water Flow and Associated Transport)高分辨率数值模型,并对水污染事件中污染物的输移进行了模拟。模型采用Godunov格式的有限体积法求解二维浅水方程和污染物输移方程,利用HLLC(Harten-Lax-van Leer-Contact)近似黎曼求解器计算单元网格界面通量,应用MUSCL限坡线性重构和龙格-库塔时间积分法实现了时空二阶精度,有效地解决了输移方程中对流项产生的数值阻尼过大和剧烈的数值振荡等问题,可准确地模拟复杂地形上干湿界面变化。同时模型引入图形处理器GPU(Graphics Processing Unit)加速计算技术。算例结果表明:模型精度高且稳定性好,能有效抑制数值阻尼和数值振荡,大幅提升了计算效率;模型可用于水污染事故的预警和评估,以期为突发水污染事件的决策提供基础数据和科学支撑。

In order to efficiently and accurately simulate the transport law of pollutants in water pollution incidents, present paper introduces a set of GPU-accelerated hydrodynamic and pollutant transport GAST(GPU Accelerated Surface Water Flow and Associated Transport) high-resolution numerical models to simulate the law of pollutant transport in water bodies. The Godunov’s finite volume method is used to solve the two-dimensional shallow water equation and the pollutant transport equation. The HLLC(Harten-Lax-van Leer-Contact) approximate Riemann solver is used to calculate the cell grid interface flux, and the MUSCL limited slope linear reconstruction and Runge-Kutta time integration method are used to achieve the second-order accuracy of time and space. It effectively solves the problems of excessive numerical damping and severe numerical oscillations caused by the convection term in the transport equation, and can accurately simulate the changes of dry and wet interfaces on complex terrain. At the same time, the model introduces GPU(Graphics Processing Unit) to accelerate calculations technology. The results of the calculation example show that the model has high accuracy and good stability, can effectively suppress numerical damping and numerical oscillation, and greatly improves the calculation efficiency of the model. It can be applied to the early warning and assessing of water pollution accidents, and is expected to provide basic data and scientific support for a decision-making of sudden water pollution incidents.