基于GPU和数据同化的深水湖库水温与溶解氧中短期预报

Medium and short term prediction of reservoir temperature and dissolved oxygen in deep lakes and reservoirs based on GPU and data assimilation

深水水库通常存在季节性温度分层,由温度分层引起溶解氧等水质指标的分层还会诱发库区水环境水生态问题。当前在中短期时间尺度上对水库水温和溶解氧进行预报的研究相对较少,提高数学模型的模拟效率与精度对提升中短期预报效果至关重要。本文采用集合卡尔曼滤波算法作为同化方法,基于CE-QUAL-W2模型建立水库水动力水质数学模型,基于OpenACC的GPU并行方法提升模型计算效率,构建大黑汀水库水温与溶解氧的数据同化系统,在中短期时间尺度上开展水库水温与溶解氧高精度、高效率预报。预报结果符合水库水温与溶解氧的中短期变化规律,能够为大黑汀水库的供水与生态安全提供技术支撑。

The seasonal thermal stratification usually exists in deep lakes and reservoirs, and the stratification of water quality variables such as dissolved oxygen caused by temperature stratification will also induce water environment and ecological problems in the reservoir.At present, few research focuses on reservoir water temperature and dissolved oxygen prediction in medium and short term time scale.It is very important toenhance the simulation efficiency and accuracy of mathematical model in order to improve the medium and short term forecast effect.Taking Luanhe-Tianjin Water Diversion Project-Daheiting reservoir as the research object, we built water temperature and dissolved oxygen data assimilation system with Ensemble Kalman filter assimilation algorithm.Based on CE-QUAL-W2,hydrodynamic and water quality mathematical model was established, and the calculation efficiency was improved through the GPU parallel method based on OpenACC model.The precise and efficient prediction of the dynamic changes of water temperature and dissolved oxygen has been realized, which is consistent with the medium and short term variation of water temperature and dissolved oxygen and can provide data support for water supply and ecological security of the reservoir.