基于长短期记忆模型的汉江中下游藻类防控调度

Reservoir operating for phytoplankton prevention and control in middle-lower reaches of Han River based on Long Short-Term Memory model

创建基于长短期记忆模型的藻类防控调度方法,可提高藻类的模拟精度,提升水库调度效益。以汉江中下游藻类及丹江口水库为研究对象,构建包括经济和生态目标函数的藻类防控调度模型,采用水体综合营养指数和浮游植物密度的长短期记忆模型建立生态目标函数,最后采用布谷鸟优化算法求解。结果显示:基于长短期记忆模型的综合营养指数和浮游植物密度模拟模型的模拟值与实测值在0.01水平相关性显著,藻类模拟效果良好;与常规调度相比,提出的藻类防控调度方案多年平均供水量可提高0.45%,多年平均发电量可提高1.06%,浮游植物平均密度降低1.57%,在有效控制汉江中下游藻类的同时,可进一步提升丹江口水库调度综合效益。研究成果可以为河流藻类防控提供技术支撑。

Reservoirs improve the spatial and temporal distribution of water resources and create huge socioeconomic benefits,while also providing a powerful tool for operating the downstream ecological environment.In the middle and lower reaches of the Han River,water bloom tends to occur in late winter and early spring.Operating through hydrodynamic methods can destroy the growth conditions of phytoplankton and effectively prevent and control phytoplankton growth.However,the formation mechanism of phytoplankton growth is complex and difficult to simulate accurately,which makes the reservoir operating for phytoplankton growth prevention and control inefficient.The problem of missing mechanism can be solved to some extent by machine learning method for phytoplankton simulation.A Long Short-Term Memory-based operating method is suitable for phytoplankton growth prevention and control,which can improve the accuracy of phytoplankton growth simulation and enhance the efficiency of reservoir operating.Taken the water bloom problem in the middle and lower reaches of Han River and Danjiangkou Reservoir as the research object,a phytoplankton growth prevention and control operating model including economic and ecological objective functions;the LSTM model of comprehensive nutritional index and phytoplankton density of water bodies is used to establish ecological objective functions;and the cuckoo optimization algorithm is used to solve the problem.The results show that:(1)the simulated values of comprehensive nutritional index and phytoplankton density simulation model based on LSTM model are significantly correlated with the measured values at 0.01 level,and the phytoplankton density simulation effect is good;(2)compared with the conventional operating,the proposed phytoplankton prevention and control operating scheme can increase the multi-year average water supply by 0.45%,the multi-year average power generation by 1.06%,and the average phytoplankton density is reduced by 1.57%,which further enhances the comprehensive benefits of Danjiangkou Reservoir operating while effectively controlling phytoplankton in the middle and lower reaches of Han River.The non-inferior solution is obtained from the operating chart of Danjiangkou Reservoir for phytoplankton growth prevention and control.The operating chart contains the flood limit line,normal storage level line,first reduced water supply line,second reduced water supply line,restricted water supply line and dead water level line,and the reservoir operating according to this chart can effectively prevent and control the phytoplankton growth in the middle and lower reaches of Han River.The research results can provide technical support for the prevention and control of river phytoplankton.