顾及共模误差的基线时间序列季节项重构

Reconstruction of Season Item of the Baseline Time Series with Consideration of Common Mode Error

针对基线时间序列的共模误差影响季节项重构的问题,以美国OKAR-TXLI基线为例,首先采用一元线性回归拟合并去除时间序列的趋势项,以避免趋势项影响季节项重构;然后运用小波的多尺度分析法分解去除趋势项后的序列,结合分量频谱特点确定并重构季节项;之后用PCA法剔除时间序列的共模误差并重构滤波后的季节项;最后对比滤波前后季节项的功率谱及振幅变化分析共模误差对重构结果的影响,并分析共模误差的特性验证其影响的合理性。结果表明,剔除共模误差后,N、E和U方向上季节项功率谱与原始序列功率谱更接近,能保留更多原始序列信息;且滤波后季节项振幅在不同周期内变化不尽相同,但其均值呈现出变小的现象,振幅均值分别减小了11.1%、15.0%和16.0%。

Aiming at the problem that the common mode error of baseline time series affects the reconstruction of season iterm,taking the American OKKARTXLI baseline as the example,firstly,the trend item of OXAR-TXLI baseline time series is removed by using unary linear regression,then the wavelet multi-scale analysis method is used to decompose time series after removing the trend item,and season item is reconstructed according to the spectral characteristics of components,After that,the PCA method is used to eliminate the common mode error of time series.Finally,the power spectrum and amplitude of the season item before and after filtering are compared,and the rationality of the influence of common mode error on the seasonal term is analyzed.The results show that after eliminating the common mode error,the power spectrums of the seasonal term in N,E and U directions are closer to the original time series power spectrums,and more information of the original time series can be retained.The amplitudes of the season item after filtering vary in different periods,but those mean values have decreased,and the average amplitudes decrease by 11.1%,15.0%and 16.0%,respectively.