大坝监测量最佳统计模型优选方法

现有报导的大坝监测量数学模型主要以拟合精确性为目标，忽略模型简单性原则的定量分析。本文从系统论角度出发，提出一种以模型拟合良好性、检验有效性以及模型简单性三方面为优选原则的大坝监测量最佳统计模型优选方法。首先将监测数据系列划分为回归数据与检验数据，依据先验知识确立统计模型优选集合，然后通过回归或优化方法，基于回归数据系列辨识统计模型参数，接着基于检验数据系列，对参数辨识的统计模型进行检验分析，最后将拟合精度、模型因子数和检验效果进行标准化，计算决策量，依据决策量大小优选最佳统计模型。结合梅山大坝11~#坝垛水平位移监测资料展示了本文提出的监测量最佳统计模型优选方法。结果表明，不同测点监测数据对应不同的统计模型，经模型优选所得出的最佳统计模型不仅具有良好的拟合性与检验有效性，而且具备模型简单性。

Optimization method of optimal statistical model of dam monitoring data

The existing reported mathematical models of dam monitoring data mainly take the fitting accuracy as the objective function, ignoring the quantitative analysis of model simplicity principle. According to the perspective of system theory, a method for optimizing the optimal statistical model of dam monitoring data is proposed in this paper on the basis of the optimization principles of good model fitting, test validity and model simplicity. First, monitoring data series are divided into regression data and test data, and the set of statistical model is determined on the basis of priori knowledge. Then the parameters of statistical models are identified based on the regression data by using the method of regression or optimization. Next,the identified statistical models are verified based on the test data. Finally,the fitting accuracy,model factor number and test effect are standardized, and based on the calculated decision content, the optimal statistical model is obtained. Combined with the monitoring data of horizontal displacement of the No. 11^# buttress of the Meishan dam,the model optimization method is implemented in detail. The research results show that the monitoring data of different measuring points are corresponding to different statistical models,the optimal statistical model not only has good fitting and validity of test,but also has the simplicity of model.