Integrated pole symmetric mode decomposition method for time-space data analysis

Abstract

The invention discloses an integrated pole symmetric mode decomposition method for time-space data analysis and belongs to the technical field of data analysis, and comprises the following steps: carrying out pole symmetric mode decomposition processing on a time sequence of each grid site to obtain a series of time modes; Unifying the number of time modes for all stations; Selecting several modesof interest from a series of time modes, finding out corresponding values of the modes for a certain concerned moment, accumulating the values of the modes, carrying out spatial synthesis according to stations, and drawing a spatial speckle pattern; Changing the attention moment, and repeating the above steps to obtain a series of spatial speckle patterns, namely spatial modalities; Obtaining time-scales of various time scales over the entire observation area Empty evolutionary speckle pattern. The method is not to solve a specific industry problem, but to develop a new data analysis method,provides technical support for scientific research and exploration of a plurality of application subjects, and has the characteristics of very high transverse property and universality.

Description

technical field [0001] The invention belongs to the technical field of time-space data analysis, and in particular relates to an integrated pole-symmetric mode decomposition method (hereinafter referred to as the IntESMD method) for time-space data analysis. Background technique [0002] The earliest modes of scientific research were experimental, then theoretical, and now computer simulations have been added, while the fourth mode that has recently emerged is caused by the explosion of data, and "data-centric" scientific discoveries are beginning to emerge . "Data" is also called observation value, which is the result of experiment, measurement, observation, investigation, etc., and is often given in the form of quantity. "Data analysis" is the process of purposefully collecting and analyzing data to turn it into information. It has become a new research mode to use data to discover phenomena and reveal laws to improve scientists' understanding of problems. It seems that...

AI Technical Summary

Problems solved by technology

Its disadvantage is that it is only suitable for linearly changing stationary signals; (2) the prevailing wavelet (Wavelet) transform method; it decomposes the signal by taking a local finite wavelet basis, which makes up for the defects of Fourier transform to a certain extent , which can express the time-varying nature of the frequency

However, its theoretical basis is still the principle of linear superposition, which is only applicable to linear signals. Since the wavelet base is orthogonal and the related theories are complete, it has obvious advantages in signal coding, storage and compression; (3) the popular EMD-based Hilbert-Huang (Hilbert-Huang) transformation method; it is a data adaptive analysis method without prior basis functions, and its decomposition mode is not only variable in frequency but also variable in amplitude, suitable for nonlinear, non-linear For stationary signals, the existing problems mainly include: the number of screenings is difficult to determine, the decomposed trend function is too rough, and the Hilbert spectrum analysis method used has inherent defects; (4) the local mean decomposition method as a variant of the empirical mode decomposition; Local mean decomposition can be regarded as a variant of empirical mode decomposition, which is similar to the later AM-FM separation method. Its advantage is that the AM function and the FM function can be directly separated by the screening process, which is beneficial to the frequency calculation. With the help of Hilbert transform; its defect is: first generate discontinuous contour segments and then try to smooth them through the moving average method. This screening method requires the number of smoothing points and the corresponding weighted average coefficient to be set in advance and must be consistent Repeated calls, this is a step backward for adaptive decomposition; (5) pole symmetric mode decomposition method (referred to as ESMD method); the ESMD method we proposed in 2013 borrowed the idea of ​​EMD decomposition, using internal pole symmetric interpolation Replace the outer envelope interpolation and use an optimization strategy to ensure that the trend function and the number of filters are optimal

However, the current version is only suitable for one-dimensional random time series analysis, and cannot be directly used in high-dimensional situations and cannot directly give prediction results

Images