论文部分内容阅读
针对监测数据中不可避免的含有随机噪声以及监测时间往往不是等时距的问题,该文提出了一种基于卡尔曼滤波的非等时距加权灰色线性组合方法:首先对原始变形监测数据进行卡尔曼滤波处理,有效剔除随机噪声的影响;然后利用滤波后的数据建立非等时间间隔的加权灰色线性组合模型进行预测;最后进行对比分析。该模型不仅具有线性回归以及GM(1,1)的特点,而且克服了传统GM(1,1)的不足。实验结果表明,基于卡尔曼滤波的非等时距加权灰色线性组合模型不仅可以有效的剔除监测数据中的随机噪声,而且提高了模型精度,具有一定的参考价值。
Aiming at the problem that monitoring data contains unavoidable random noise and the monitoring time is not always time-equal, a non-equal-time-weighted gray linear combination method based on Kalman filtering is proposed in this paper: firstly, the original deformation monitoring data is processed by Karl. Mann filtering effectively eliminates the effects of random noise. Then, the filtered data is used to establish a weighted grey linear combination model with unequal time intervals for prediction. Finally, a comparative analysis is performed. This model not only has the characteristics of linear regression and GM(1,1), but also overcomes the deficiency of traditional GM(1,1). The experimental results show that the non-equal-time-weighted grey linear combination model based on Kalman filter not only can effectively eliminate the random noise in the monitoring data, but also improve the model accuracy, which has certain reference value.