Research On Medium And Short-Term Prediction Of Temperature And Oxygen In Deep Water Reservoir Based On GPU And Data Assimilation

Seasonal temperature stratification usually exists in deep-water reservoirs.The change of water temperature not only affects the movement law of water body,but also affects the metabolism of aquatic organisms.The stratification of water quality indicators such as dissolved oxygen caused by temperature stratification will also induce water environmental and ecological problems in the reservoir area,threatening water supply safety and downstream ecosystem health.At present,the research on reservoir water temperature and dissolved oxygen mostly focuses on its long-term evolution law,and the research on medium and short-term prediction is relatively less.Improving the simulation efficiency and accuracy of the mathematical model is crucial to improve the medium and short-term prediction effect.This paper took Daheiting Reservoir as the research object,which is the project of Luanhe River to Tianjin.The long-term and high-frequency monitoring data of the reservoir were used as the assimilation data source.The ensemble Kalman filter algorithm was used as the assimilation method.The mathematical model of reservoir hydrodynamic water quality was established based on CE-QUAL-W2 model.The GPU parallel method based on OpenACC was used to improve the calculation efficiency of the model.The data assimilation system of water temperature and dissolved oxygen in Daheiting Reservoir was constructed.The high-precision and high-efficiency prediction of the dynamic changes of water temperature and dissolved oxygen in the reservoir in the medium and short-term time scale was realized,which provides data support for reservoir water environment management.The main results include :(1)Selecting the appropriate OpenACC introduction can effectively reduce the time consumption of data transmission between the host and GPU,and improve the GPU parallel acceleration effect.The combination of kernels lead and copy and copyin sublanguages selected in this paper can effectively reduce data transmission time,and effectively improve the GPU parallel efficiency.When the number of model grids is less than 600,it takes too much time in the data transmission stage,resulting in CPUGPU collaborative computing time is more than the traditional CPU serial computing.When the number of grids exceeds 600,the time-consuming of GPU parallel computing begins to be lower than that of CPU serial computing.Since then,the more model grids are,the greater the advantage of GPU parallel computing is.(2)The ensemble Kalman filter(EnKF)was selected as the data assimilation method to construct the water temperature and dissolved oxygen data assimilation system of Daheiting Reservoir.The in-situ observation data of Daheiting Reservoir were combined with the W2 model to correct the simulation results of water temperature and dissolved oxygen.The results of water temperature assimilation showed that the average simulation deviations of water temperature at 1 m,7 m and 13 m undewater at the monitoring station of the front section of the dam were 1.01°C,0.31°C and 0.21°C,respectively,which were 68.8%,51.6% and 41.2% higher than those without data assimilation.The assimilation results of dissolved oxygen showed that the average simulation deviations of dissolved oxygen at 1 m,7 m and 13 m underwater at the monitoring station of the front section of the dam were 0.38 mg/L,0.22 mg/L and 0.18mg/L,respectively,which were 71.2%,63.6% and 54.4% higher than those without data assimilation.(3)The data assimilation system constructed in this paper was used to predict the water temperature and dissolved oxygen concentration of Daheiting Reservoir for 1-10 days in the medium and short term.The results show that the average deviation of water temperature prediction is within 0.60°C and that of dissolved oxygen prediction is within 0.54mg/L in different prediction time.The prediction results are consistent with the short-term and medium-term variation of water temperature and dissolved oxygen in Daheiting Reservoir,which can provide technical support for water supply and ecological security of Deepwater Reservoir.