例2 大坝监测信息的时变Kalman预测模型.docVIP

大 坝 监 测 信 息 的 时 变 Kalman 预 测 模 型 摘 要 基于对大坝监测信息预测模型时变性的要求，在模型LS参数求解过程中引入遗忘因子，通过建立遗忘矩阵突出现在及近期资料对模型的贡献，提出了能够实现模型参数实时更新的IWRLS算法。在此基础上，为使预测模型体现物理含义的同时实现滤波操作，在Kalman滤波方程组中融入统计模型、ARMA等多种方法：以自变量建立状态方程，并以ARMA反映其状态变化特性；以因变量建立观测方程，并以统计模型表征其物理含义。由此建立了考虑白色观测噪声的时变Kalman预测模型，具有滤波、参数时变、分量分离、拟合及预测的综合功能。实例分析表明，时变Kalman模型拟合及预测精度均优于传统统计模型，为大坝监测资料的预测分析提供了新思路。 关键词：监测信息；时变；Kalman滤波；预测 中图分类号：TV Time-varying Kalman Model for Dam Monitoring Information Prediction Abstract Based on time-varying requirements of prediction model for dam monitoring information, in the LS algorithm for solving the model parameters, the forgetting factor is introduced to set up a forgotten matrix to give prominence to the contributions of recent data, then the IWRLS algorithm is made to achieve updating model parameters real-time. On this basis, in order to reflect the physical meaning and complete the filtering operation at the same time, statistical model and ARMA are introduced into the Kalman filter equations. In the equations, state equation is established by self-variable where reflect the state characteristics with ARMA, and observation equation is established by dependent variable and where reflect physical meaning with statistical models. So considering the white noise, the time-varying Kalman prediction model is established with the comprehensive functions such as filtering, time-varying parameters, component separation, fitting and forecasting. Case analysis shows that the fitting and forecast accuracy of time-varying Kalman model are superior to traditional statistical models. It provides a new idea for the prediction of dam monitoring data analysis. Key words：monitoring information; time-varying; Kalman filter; prediction 目 录 摘 要 - 2 - Abstract - 3 - 引 言 - 5 - 1　时变参数估值 - 5 - 2　Kalman滤波模型及预报器 - 7 - 3　时变Kalman预测模型的建立 - 9 - 3.1　状态方程与观测方程的建立 - 9 - 3.2　参 数 的 确 定 - 10 - 3.3　时变Kalman模型预报器 - 12 - 4　实 例 分 析 - 13 - 图1 水库蓄水期上游水位变化过程线 - 13 - 表1 拟合模型特征值表 - 14 - （a）传统统计模型 （b）时变Kalma