BP神经网络在大坝变形分析中的应用及改进

发布时间：2018-10-10 07:57

【摘要】：水库作为当今社会不可或缺的水利工程枢纽,一直以来发都挥着巨大的工程效益。同时也作为防汛抗洪调度的核心工程,为确保一方平安做出了巨大贡献。水力发电不断为经济社会发展以及工业生产提供着所需能源,为人们的日常饮水灌溉提供了稳定的动力和来源。大坝是水库建设与运营的核心建筑物,大坝在设计时需要考虑结构和高度,而大坝的结构和高度是根据洪水淹没范围来考虑,有时也需要考虑上下游的梯级水电站。建筑环境的复杂多变使得坝体通常承受着巨大的荷载,易发生各种类型的偏移及形变,因此安全监测对水库大坝来说至关重要。在定期进行监测的基础上,也需要结合往期数据对大坝的形变进行分析预测,以便在险情来临之前可以做到提早预防。大坝坝体变形的产生与诸多因素有关,如上下游库水位差,气压,温度,时效等,同时由于这些因素一般具有较强的随机性,各因素之间的关系较为复杂,使得这些因素对坝体的变形量产生的影响无法用确切的定量关系式描述。现阶段,大坝变形监测数据分析的方法多种多样,其中大都是结合长期的变形监测数据并利用统计模型对大坝的安全状况进行分析判断,试验研究表明统计模型在处理类似大坝变形这种非线性模糊系统时存在缺陷和不足,分析预测的结果存在较大的误差,可靠性和稳定性不能达到要求。由于BP网络模型可以处理非线性映射问题,而且在处理该问题上具有相当明显的优势。本课题选用BP神经网络模型来对大坝变形的预警进行分析研究。同时对传统的BP网络模型存在的缺陷进行改进,在传统BP网络模型的基础上加入遗传算法和改进粒子群优化算法,使用这两种优化算法实现了对传统BP网络模型的优化,并将其应用于工程实例。结果表明,基于遗传算法和改进粒子群优化算法的BP模型相较于传统的数值分析方法和单一 BP模型分析方法具有预测精度高,容错能力和泛化能力强等优点。
[Abstract]:Reservoir, as an indispensable water conservancy project, has always been a huge project benefit. At the same time, as the core project of flood control and flood control, it has made a great contribution to ensure the safety of one side. Hydroelectric power supply energy for economic and social development and industrial production, and provide stable power and source for people's daily drinking water irrigation. The dam is the core building of reservoir construction and operation. The structure and height of the dam should be considered when it is designed, and the structure and height of the dam are considered according to the scope of flood inundation, and sometimes the upstream and downstream cascade hydropower stations are also considered. Because of the complex and changeable building environment the dam body usually bears huge loads and is prone to various types of migration and deformation. Therefore safety monitoring is very important to the reservoir dam. On the basis of regular monitoring, it is also necessary to analyze and predict the deformation of the dam with the data of the previous period so as to prevent the dam early before the danger comes. The deformation of dam body is related to many factors, such as water level difference, air pressure, temperature, aging and so on. At the same time, because of the strong randomness of these factors, the relationship between these factors is more complex. The influence of these factors on the deformation of the dam body can not be described by the exact quantitative relation. At present, there are a variety of methods for analyzing dam deformation monitoring data, most of which are combined with long-term deformation monitoring data and statistical model is used to analyze and judge the safety status of dams. The experimental results show that the statistical model has defects and shortcomings in dealing with nonlinear fuzzy systems such as dam deformation, and the results of analysis and prediction have large errors, and the reliability and stability cannot meet the requirements. Because the BP network model can deal with nonlinear mapping problem, it has obvious advantages in dealing with this problem. In this paper, BP neural network model is used to analyze and study the early warning of dam deformation. At the same time, the defects of the traditional BP network model are improved. Genetic algorithm and improved particle swarm optimization algorithm are added to the traditional BP network model. The two optimization algorithms are used to realize the optimization of the traditional BP network model. It is applied to engineering example. The results show that the BP model based on genetic algorithm and improved particle swarm optimization algorithm is superior to the traditional numerical analysis method and single BP model analysis method in high prediction accuracy, fault tolerance and generalization ability.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP183;TV698.11

【相似文献】