基于变形监测管理系统中预测模型的研究

预警报警功能作为建(构)筑物变形监测信息管理系统中的重要功能,其可靠性和精确度显得尤为重要。建(构)筑物变形监测管理系统中所用到的灰色模型与回归分析模型具有良好的可靠性和稳定性,在模型基础上进一步延伸,提高其预测精度,针对多尺度时间序列各尺度单支预测与整合重构的基础上,对变形监测数据进行小波分解成各尺度的子序列,根据各子序列性质配置适合的回声状态网络并进行单支预测,预测子序列经过权值计算重构成预测的序列。实验表明:小波分解理论基础上的回声状态模型预测精度较高。

Research on Forecast Model Based on Deformation Monitoring Management System

The early warning function is an important function in the information management system for the deformation monitoring of buildings (structures),and its reliability and accuracy are particularly important.The grey model and regression analysis model used in the construction (structure) monitoring system for building deformation have good reliability and stability.It is further extended on the basis of the model to improve its prediction accuracy.For each scale of a multi-scale time series based on prediction and integration and reconstruction,wavelet decomposition of deformation monitoring data into sub-sequences of various scales,and a suitable echo state network is configured according to the properties of each sub-sequence and single-segment prediction is performed.The predicted sub-sequences are reconstructed by weight calculation.Experiments show that the echo state model based on wavelet decomposition theory has higher prediction accuracy.