Geomagnetic field prediction method based on long-short term memory (LSTM) model recurrent neural network

Abstract

The invention discloses a geomagnetic field prediction method based on a long-short term memory (LSTM) model recurrent neural network. Compared with a spherical harmonic analysis method, the geomagnetic field prediction method has the advantages of simple structure, few parameters to be solved, and simple solving method. The recurrent neural network adopts a long-short term memory model structure,the processing of the temporal relationship of geomagnetic field data is enhanced, and the rule of the variation of the geomagnetic field data over time can be effectively predicted. The complicatedprocessing of the original geomagnetic field data is not required, the geomagnetic field data do not need to be pre-processed spectrally and statistically, and the processing of the data is facilitated. For periodic characteristics of the geomagnetic field data, the length of time of training samples and test samples is selected, on this basis, the structure of the recurrent neural network is designed, so that the neural network can effectively analyze the periodic characteristics of the geomagnetic field, and the prediction accuracy of the geomagnetic field is greatly improved.

Description

technical field [0001] The invention belongs to the technical field of space physics of earth science, and in particular relates to a geomagnetic field prediction method based on a long-short-term memory model recurrent neural network. Background technique [0002] There is a natural magnetic phenomenon inside the earth, which is the geomagnetic field. The analysis and prediction of geomagnetic field data has high application value in engineering, especially in anomaly detection, space physics, navigation and positioning and other fields. Vehicles, metal minerals, submarines, mines and other magnetic targets have their own magnetic fields superimposed on the geomagnetic field, causing the geomagnetic field to appear abnormal in a certain area. In the magnetic field anomaly detection technology, the magnetic field detection equipment will detect the magnetic field anomaly signal, and by extracting and analyzing the abnormal signal, the relevant information (including positio...

AI Technical Summary

Problems solved by technology

The existing methods are limited by the truncation level of the spherical harmonic series. The spatial resolution of the spherical harmonic analysis method is generally not high, which is far from meeting the actual needs of detection and navigation. Although the resolution can be improved by increasing the truncation level, when the spherical harmonic level When the number increases, the number of spherical harmonic coefficients to be solved increases sharply, and the amount of calculation and storage increases sharply

Images