泵站动态规划优化模型的全局敏感性分析方法及应用

为深入探究泵站实际运行特征,识别敏感参数,为泵站运行提供技术支撑,以龙门泵站为例,基于拉丁超立方抽样技术对各个参数进行抽样,基于方差的Sobol指数法、互信息法分别分析了泵站优化调度模型的全局敏感性分析。结果表明,瞬时流量为对泵站优化模型最敏感的参数,即瞬时流量对泵站最小能耗影响最大;瞬时流量与其他参数存在较强的交互效应,在构建相关优化模型及对模型进行率定过程中需优先考虑;Sobol法和互信息法均适用于多变量非单调非线性模型,且对参数的概率分布无特殊要求,能够准确定量地识别参数的敏感性。研究成果可为泵站优化模型精确建模及高效率定提供指导。

Global Sensitivity Analysis of Dynamic Programming Optimization Model of Pumping Station and Its Application

In order to deeply investigate the characteristics of the actual operation of the pumping station, identify sensitive parameters, and provide technical support for the operation of the pumping station, the global sensitivity analysis of the pumping station optimal scheduling model was carried out. We sampled every parameter based on the Latin hypercube sampling technique. The variance-based Sobol index method and the mutual information method were used to analyze the influence of the parameters on the output results of the pumping station optimal scheduling. Taking Longmen pumping station as an example, the results show that the instantaneous flow is the most sensitive parameter of the pumping station optimal scheduling model, which means that the instantaneous flow has the greatest influence on the minimum energy consumption of the pumping station. The instantaneous flow has quite strong interaction effect with other parameters, showing that we should give priority to consider it in the process of building and calibrating relative optimal models. The Sobol index method and the mutual information method are both suitable for multivariable non-monotone nonlinearity model without any special requirements for the probability distribution of the parameters, being able to precisely identify the sensitivity of parameters. The research results provide guidance for the precise modeling and high efficiency of calibrating models.