万家寨水库建成后上游河段冰情特性分析及冰情信息模拟预报模型研究

发布时间：2018-07-05 10:39

本文选题：万家寨上游河段 + 冰情特性　；参考：《内蒙古农业大学》2017年硕士论文

【摘要】：在万家寨水利枢纽修建前,距坝96km以上河段为稳定封冻河段,以下河段冬季一般不封河,河道仅有岸冰和流冰花,整个冬季以流凌为主,封、开河时期不产生冰塞冰坝等险情,不发生凌汛灾害。1998年万家寨水库建成后,改变了万家寨上游河道天然形态,使冰情特征发生了很大变化。由于水库蓄水,库区水面比降和回水末端流速减小,冰凌输移能力明显减小,下泄的大量冰花滞留在库区,成为首封地点,然后向上游延伸,使得在封开河期卡冰结坝现象频发,冰凌灾害严重。本文首先根据1987-2016年黄河万家寨上游河段的历史实测冰情资料,统计分析了该河段的冰情特征以及万家寨水库运行后出现的新冰情,并对影响该河段冰情的河道形态、气象条件、水库调度和水力因素等方面进行分析。其次将支持向量机模型(SVR)、BP神经网络模型和偏最小二乘回归模型(PLS)应用于万家寨上游河段的流凌日期、封河日期和开河日期预报中。最后利用改进累积负气温法和统一度-日法对万家寨上游河段的冰厚进行了模拟及验证。研究表明万家寨上游河段冰情特征规律是:封河顺序由库区自下而上,封冻形式为平立封相交;开河时库区立封河段及立封河段以下解冻较上游晚,以上河段由下游向上游开河,开河形式为文开、武开、半文半武。1998年万家寨水库运行后,河道形态及水力因素的改变,气温的升高导致冰情发生了明显的变化,主要表现为:流凌、封河日期推迟,开河日期提前,不封冻河段变成稳定封冻河段,封冻长度增加,封河期水位升高,冰塞冰坝险情增加。万家寨上游河段的流凌日期、封河日期和开河日期预报和冰厚模拟的结果表明:所建立的三个模型精度均满足预报要求,预报结果良好,其中偏最小二乘回归模型预测效果最好,对流凌、封河和开河日期预报精度均为100%,均属于甲等预报方案;支持向量机模型预测效果次之;BP神经网络预测效果相对较差。统一度-日法模型对万家寨上游河段冰厚的模拟精度明显高于改进累积负气温法,可更好的运用于该河段冰厚模拟中,为当地防凌减灾工作提供可靠的理论依据和有效的技术指导。
[Abstract]:Before the construction of Wanjiazhai Water Conservancy Project, the section above the dam 96km was a stable frozen reach, and the following reaches generally did not seal the river in winter. The river channel had only shore ice and floating ice flowers, and the whole winter was dominated by the flow of ice and ice, and the closure of the river did not result in dangerous conditions such as ice slugs and ice dams during the period of Kaihe River. After the completion of Wanjiazhai Reservoir in 1998, the natural form of river course in the upper reaches of Wanjiazhai has been changed and the ice characteristics have changed greatly. As a result of the reservoir water storage, the water surface specific drop and the return end velocity decrease, and the ice transport capacity is obviously reduced. A large number of ice flowers released from the reservoir area remain in the reservoir area, become the first seal site, and then extend upstream. During the period of Fengkai River, the phenomenon of carapace and damming occurs frequently, and the ice disaster is serious. Based on the historical measured ice data of the upper reaches of the Yellow River Wanjiazhai from 1987 to 2016, the ice regime characteristics of the reach and the new ice conditions after the operation of the Wanjiazhai Reservoir are statistically analyzed, and the river channel shape and meteorological conditions that affect the ice condition of the reach are also discussed in this paper. Reservoir operation and hydraulic factors were analyzed. Secondly, the support vector machine (SVR) BP neural network model and partial least square regression model (PLS) are applied to the prediction of river flow, closure and Kaihe in the upper reaches of Wanjiazhai. Finally, the ice thickness of the upper reaches of Wanjiazhai is simulated and verified by the improved accumulated negative temperature method and the unified degree-day method. The study shows that the ice regime characteristics of the upper reaches of Wanjiazhai are as follows: the sequence of river closure is from the bottom to the top of the reservoir area, and the freezing form is Pingli seal intersecting, and the thawing of the Lifeng reach and below the Lifeng reach in the Kaihe reservoir area is later than that in the upper reaches. In the form of Wenkai, Wukai, and half-Wenjiazhai in 1998, after the operation of Wanjiazhai Reservoir, the changes of river channel shape and hydraulic factors, and the increase of air temperature, resulted in obvious changes in ice conditions. The main performance is that the river flow, the date of river closure is delayed, the date of opening river is advanced, the section without freezing is changed into a stable frozen reach, the length of the frozen river increases, the water level rises during the period of river closure, and the dangerous situation of ice slug ice dam increases. The results of the prediction and ice thickness simulation show that the accuracy of the three models all meet the requirement of forecast, and the forecast results are good. The partial least square regression model has the best prediction effect, and the forecasting accuracy of convection, river closure and Kaihe is 100, which belongs to the first class prediction scheme, and the prediction effect of support vector machine model is inferior to that of BP neural network. The accuracy of simulation of ice thickness in the upper reaches of Wanjiazhai is obviously higher than that of the improved cumulative negative temperature method, which can be better used in the simulation of ice thickness in the upper reaches of Wanjiazhai River. To provide reliable theoretical basis and effective technical guidance for local ice prevention and disaster reduction work.
【学位授予单位】：内蒙古农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TV875

【参考文献】