在调水工程中, 如果泵站站前水位过低, 会危及泵站安全, 如果水位过高, 会危及周边安全, 因此探寻调水工程中河渠湖库水位变化显得尤为重要。以南水北调东线山东段南四湖为研究区域, 寻求不同起调水位、出入流量、泵 站开启时间差的调水方案下泵前水位变化规律。先利用耦合模型对不同的调水方案进行数值模拟, 然后选取 23 组 调水方案及其数值模拟所得的泵前水位作为样本训练 BP 神经网络, 建立 BP 神经网络调蓄水位预测模型并进行验证, 最后利用预测模型对不同调水方案进行泵前水位预测。结果表明, BP 神经网络预测模型具有很强的预测能力, 预测模型结果与耦合模型结果泵前水位基本吻合, 水深相对误差小于 9.15% , 而模型计算效率提升 96.67% 。

In a water diversion project, if the water level before the pump is too low, the security of the pumping station will be endangered; if the water level is too high, the safety of the surrounding area will be endangered. So, it is particularly important to study the change of water level in the reservoir. The paper, based on the Nansi Lake section of the eastern route of the South to North Water Transfer Project in Shandong Province, explores the water level variation pattern before the pump under different conditions of initial water level, inflow and outflow, and opening time differ ence between pumping stations. Firstly, we used the coupled model to numerically simulate different diversion plans. Then we selected 23 water diversion plans and their beforepump water level obtained from numerical simulation as samples to train the BP neural network. Thus, we established a prediction model for water level regulation and had it veri fied. Finally we predicted the water level before the pump in different diversion plans using the model. The results showed that the BP neural network model has a strong predictive power. Its results were basically consistent with the results of the coupled model, with the relative error of water depth less than 9. 15% . Meanw hile the modelcs calculation efficiency improved by 96. 67% .

“十二五”国家科技支撑计划项目 ( 2015BAB07B02) ; 国家自然科学基金项目( 51621092; 51609166)