Transform-based low-cost GNSS ionosphere disturbance monitoring and early warning method and system

Abstract

The invention discloses a low-cost GNSS (Global Navigation Satellite System) ionosphere disturbance monitoring and early warning method based on Transform, which is characterized in that a low-cost GNSS receiver is used for extracting total electron content (TEC) and processing data of a plurality of monitoring stations in a region in real time in combination with an ROTI index, and ionosphere disturbance is quickly predicted and early warning delay is remarkably reduced through a dynamic attention mechanism and multi-scale feature extraction of the Transform. Compared with a traditional method, more comprehensive ionosphere state information is provided by integrating observation sequences of all receivers in an area, meanwhile, the dependence on expensive high-frequency equipment is greatly reduced, and the monitoring cost is reduced. According to the invention, the architecture based on deep learning has high expandability, can be seamlessly integrated to the existing GNSS network, and provides an economical, efficient and high-timeliness solution for global ionosphere monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of ionospheric disturbance monitoring and early warning, and in particular to a low-cost GNSS ionospheric disturbance monitoring and early warning method and system based on Transformer. BACKGROUND

[0002] In modern communication and navigation technology, the ionosphere plays a key role, and its electron density changes directly affect radio wave propagation, having a significant impact on the signal quality and positioning accuracy of satellite navigation systems such as GPS. Ionospheric disturbances, such as traveling ionospheric disturbances and equatorial ionospheric bubbles, can cause abnormal electron density distribution, interfering with radio communication and navigation systems. Existing ionospheric monitoring techniques have obvious limitations: they rely on high-cost ground stations and high-frequency receivers, making it difficult to achieve continuous coverage worldwide, especially in remote areas such as the ocean; the early warning system is slow to respond, making it difficult to meet real-time requirements; in addition, traditional methods usually only use the observation data of a single receiver and satellite pair, resulting in low data utilization and difficulty in fully describing regional ionospheric conditions, and there is a long time delay. SUMMARY

[0003] To overcome the shortcomings of the prior art, the present application provides a low-cost GNSS ionospheric disturbance monitoring and early warning method based on Transformer, which uses low-cost GNSS receivers to extract total electron content (TEC) and combines with ROTI indicators to process the data of multiple monitoring stations in the region in real time, and through the dynamic attention mechanism and multi-scale feature extraction of Transformer, the ionospheric disturbance is quickly predicted, significantly reducing the early warning delay.

[0004] According to an aspect of the present application, a low-cost GNSS ionospheric disturbance monitoring and early warning method based on Transformer is provided, comprising: Obtaining observation data of a plurality of low-cost GNSS receivers and satellite pairs and preprocessing the data; Inputting the preprocessed data into a trained disturbance monitoring and early warning model to output the ionospheric disturbance level at the next time; wherein the training of the disturbance monitoring and early warning model comprises: Obtaining long-time series of high-frequency device data and low-cost GNSS receiver observation data; Extracting ionospheric observation values from the high-frequency device data, using the ROTI index determination method to select the observation values in the ionospheric disturbance region, and labeling them according to different levels; at the same time, extracting ionospheric observation values from the low-cost GNSS receiver observation data; Using the selected ionospheric disturbance area and disturbance time as conditions, the ionospheric observation values extracted from the low-cost GNSS receiver observation data are marked according to the corresponding disturbance area and disturbance time, and the marked value is the ionospheric disturbance level of the high-frequency observation value in this time and range; All GNSS observation data are characterized by receiver-satellite-penetration point area-ionospheric short-term historical observation sequence, different features are stacked according to time sequence as input, and the level of ionospheric disturbance occurring at the next time is taken as the label as output. A deep learning model based on multi-input, multi-scale spatio-temporal feature fusion of Transformer is trained, and the trained model is used for disturbance monitoring and early warning.

[0005] As a further technical solution, the ionospheric observation values are extracted from the high-frequency device data, including: Based on the carrier phase smoothing pseudorange method, a combination of pseudorange and carrier phase is constructed, the combined ambiguity is estimated, and the TEC is extracted by combining the carrier phase equation.

[0006] As a further technical solution, the method further comprises: Real-time monitoring of TEC change rate, dynamic adjustment of arc length.

[0007] As a further technical solution, the ROTI index determination method is used to screen out ionospheric disturbance area observation values and mark them according to different levels, including: Based on the observation data, the ionospheric observation values are extracted, and the TEC change rate is calculated; In the set time interval, the standard deviation of the TEC change rate is calculated to obtain the ROTI value; Based on the ROTI value, the ionospheric disturbance level is defined.

[0008] As a further technical solution, using the selected ionospheric disturbance area and disturbance time as conditions, the ionospheric observation values extracted from the low-cost GNSS receiver observation data are marked according to the corresponding disturbance area and disturbance time, including: Calculate the ROTI of each high-frequency device observation path to determine the disturbance area and disturbance time; Align the IPP coordinates and time of low-cost GNSS observation with the disturbance area and time of high-frequency devices. If the IPP of low-cost GNSS observation falls within the disturbance area and the time overlaps, the corresponding ROTI level is assigned; Add a field mark to each low-cost GNSS observation record.

[0009] As a further technical solution, the deep learning model based on multi-input, multi-scale spatio-temporal feature fusion of Transformer includes: a time series encoding and fusion module for using multi-scale convolution to extract and fuse features of different time scales for the input TEC and ROTI; a spatial feature embedding and fusion module for embedding discrete features and encoding continuous features of the input receiver, satellite and IPP network features, and fusing the discrete features and the continuous features; a context feature fusion module for encoding the ROTI, TEC mean, solar / magnetic index and time features of the surrounding receivers; a multi-head fusion and classification module for splicing the time fusion features, spatial fusion features and context fusion features, and outputting the disturbance level probability and prediction label.

[0010] As a further technical solution, the preprocessing comprises: obtaining observation data of a plurality of low-cost GNSS receivers and satellite pairs; extracting ionospheric observation values from the low-cost GNSS receiver observation data, and calculating TEC sequences, TEC change rates and ROTI values; stacking different features according to time sequences as input, with all observation data of the low-cost GNSS receiver as features according to receiver-satellite-penetration point area-ionospheric short-term historical observation sequences.

[0011] According to an aspect of the present application, a low-cost GNSS ionospheric disturbance monitoring and early warning system based on a Transformer is provided, comprising: a data preprocessing module for obtaining observation data of a plurality of low-cost GNSS receivers and satellite pairs and preprocessing the data; a monitoring and early warning module for inputting the preprocessed data into a trained disturbance monitoring and early warning model, and outputting ionospheric disturbance levels at the next time; wherein the training of the disturbance monitoring and early warning model comprises: obtaining long-time sequence high-frequency device data and low-cost GNSS receiver observation data; extracting ionospheric observation values from the high-frequency device data, using the ROTI index determination method to filter out observation values in the ionospheric disturbance region, and labeling according to different levels; at the same time, extracting ionospheric observation values from the low-cost GNSS receiver observation data; using the filtered ionospheric disturbance region and disturbance time as conditions, marking the ionospheric observation values extracted from the low-cost GNSS receiver observation data according to the corresponding disturbance region and disturbance time, and the marking value is the ionospheric disturbance level of the high-frequency observation value in this time and range; All GNSS observation data is characterized by receiver-satellite-penetration point area-ionospheric short-term history observation sequence, different features are stacked according to time sequence as input, the level of ionospheric disturbance occurring at the next moment is taken as output, a deep learning model based on multi-input, multi-scale spatio-temporal feature fusion of Transformer is trained, and the trained model is used for disturbance monitoring and early warning.

[0012] According to an aspect of the present application, a low-cost GNSS ionospheric disturbance monitoring and early warning device based on Transformer is provided, comprising a memory and a processor, the memory storing program instructions executed by the processor, and the processor calling the program instructions to execute the low-cost GNSS ionospheric disturbance monitoring and early warning method based on Transformer.

[0013] According to an aspect of the present application, a non-transitory computer readable storage medium is provided, which stores computer instructions for executing the low-cost GNSS ionospheric disturbance monitoring and early warning method based on Transformer.

[0014] The present application provides a low-cost GNSS ionospheric disturbance monitoring and early warning method based on Transformer, which extracts total electron content (TEC) using low-cost GNSS receivers and realizes high-precision disturbance prediction combined with ROTI index, compared with the prior art, the beneficial effects of the present application are: (1) Regional data integration capability The present method breaks through the limitation of relying on a single receiver in traditional ionospheric monitoring, and innovatively uses ionospheric observation data of all low-cost GNSS receivers in the region, including TEC sequence, rate of change and ROTI value, to construct a comprehensive spatio-temporal feature set. Traditional methods usually only analyze the time sequence of a single receiver and satellite pair, which is difficult to capture the regional characteristics of ionospheric disturbances, such as ionospheric anomalies propagating across regions. This method integrates the observation results of multiple receivers in the region through weighted fusion based on ROTI intensity, fully considers the disturbance characteristics and signal quality differences, and provides more comprehensive and continuous ionospheric state information. This high data utilization method significantly improves the monitoring accuracy, especially in areas such as the ocean where receivers are sparsely distributed, showing strong regional collaboration capability.

[0015] (2) Significantly reduce the existing monitoring time delay Traditional ionospheric monitoring is limited by the observation frequency and data accumulation requirements of a single station, often leading to minute-level processing delays, making it difficult to meet the needs of real-time early warning. The method uses a multi-scale spatio-temporal Transformer model to replace traditional models, with the ability to process multiple low-cost GNSS receiver station data in parallel within a region, allowing for rapid analysis of TEC and ROTI data streams, real-time prediction of ionospheric disturbance levels, including undisturbed, mild, moderate, and severe states. The Transformer efficiently identifies key moments of rapid ionospheric changes, such as significant increases in ROTI, through a dynamic attention mechanism, reducing the delay from data collection to early warning to seconds. This high timeliness ensures that the system can respond promptly to ionospheric anomalies, significantly reducing the potential impact on communication and navigation systems.

[0016] (3) Cost-effectiveness The method significantly reduces the cost of ionospheric monitoring equipment procurement and maintenance by using low-cost GNSS receivers instead of expensive high-frequency equipment, while maintaining high-precision TEC extraction and ROTI disturbance monitoring capabilities. The multi-scale spatio-temporal Transformer-based technical architecture has excellent scalability, allowing easy integration into existing global GNSS monitoring networks without additional hardware upgrades. Through regional data integration and adaptive processing optimization, the method achieves monitoring results comparable to high-frequency equipment on low-cost devices, providing an economical and efficient solution for global ionospheric monitoring, especially in resource-limited remote areas, demonstrating significant cost advantages.

[0017] (4) History-driven prediction model The core innovation of the invention is to use historical ionospheric observation data from low-cost GNSS receivers, including TEC, rate of change, and ROTI sequences, to train a multi-scale spatio-temporal Transformer model, enabling accurate prediction and early warning of ionospheric disturbances. Traditional methods rely on real-time data from high-frequency receivers, which are costly and have limited coverage, while the method uses deep learning techniques to mine disturbance patterns from historical data, allowing the construction of high-performance prediction models without expensive equipment. The Transformer model effectively captures short-term mutations and long-term trends in ionospheric disturbances through multi-scale feature extraction and dynamic attention mechanisms, improving the robustness and accuracy of predictions, providing a low-cost, efficient early warning solution for communication, navigation, and space weather monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly described in the following. Obviously, the accompanying drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can be obtained based on these drawings without creative labor.

[0019] Figure 1 A flowchart of a low-cost GNSS ionospheric disturbance monitoring and early warning method based on a Transformer provided by an embodiment of the present application. DETAILED DESCRIPTION

[0020] It should be noted that: In view of the current ionospheric monitoring technology relying on high-cost ground stations and high-frequency receivers, it is difficult to achieve continuous coverage in the global range, especially in remote areas such as oceans; the early warning system is slow to respond, and it is difficult to meet the real-time demand; and only the observation data of a single receiver and satellite pair are used, the data utilization rate is low, it is difficult to fully describe the regional ionospheric state, and there is a long time delay, the present application proposes an ionospheric disturbance monitoring and early warning method based on multi-scale space-time Transformer (MST-Transformer). The method uses low-cost GNSS receivers to extract total electron content (TEC) and combines ROTI indicators to process the data of multiple monitoring stations in the region in real time, quickly predict ionospheric disturbances through the dynamic attention mechanism and multi-scale feature extraction of the Transformer, and significantly reduce the early warning delay. Compared with the traditional method, the present method integrates the observation sequences of all receivers in the region to provide more comprehensive ionospheric state information, while greatly reducing the dependence on expensive high-frequency equipment and reducing the monitoring cost. The deep learning-based architecture has high scalability and can be seamlessly integrated into existing GNSS networks, providing an economical and efficient, high-time-efficient solution for global ionospheric monitoring.

[0021] The terms "include" and "have" and any variations thereof in the specification and claims of the present application and the above-described drawings are intended to cover the non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units, without being limited to clearly listed steps or units, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present application. In addition, the technical features in each embodiment or in a single embodiment provided by the present application can be combined with each other at will to form new technical solutions, and such combination is not restricted by the order of steps and / or structure mode, but should be based on the fact that it can be realized by those of ordinary skill in the art. When the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, and is not within the protection scope of the present application.

[0023] Please refer to Figure 1 The low-cost GNSS ionospheric disturbance monitoring and early warning method of the Transformer provided by the embodiments of the present application has the following flow: (1) Prepare high-frequency receiver and other high-frequency device data and low-cost GNSS receiver observation data of long time series.

[0024] (2) Extract ionospheric observation values from the observation values of the high-frequency device. The extraction method uses the optimized carrier phase smoothing pseudorange method, which is applicable to both high-frequency and low-frequency devices. The specific principle is as follows: The phase smoothing pseudorange method combines pseudorange and phase observation, uses the high-precision carrier phase to smooth pseudorange noise, and improves the extraction accuracy of ionospheric TEC. The core idea is to use the characteristics that the phase and pseudorange are affected by the ionosphere in the same size and opposite sign, and to weaken the noise effect by continuous arc segment average ambiguity. The phase smoothing pseudorange method observation equation is simplified as:

[0025] Among them: i: represents different frequencies, which can be distinguished by k; : receiver r to satellite s dual-frequency pseudorange geometrically independent combination observation value; : dual-frequency carrier phase geometrically independent combination observation value; (unit: , is the frequency, unit: Hz), which represents the frequency-dependent coefficient of ionospheric delay; : TEC along the signal path (unit: TECU); c: speed of light (unit: m / s) Differential Code Bias of receiver and satellite; Differential Phase Bias of receiver and satellite; Frequency Wavelength (unit: m); Frequency Integer ambiguity.

[0026] For simplicity, define:

[0027] Then the observation equation is simplified as:

[0028] where, and denote the observation noise of pseudorange and carrier phase respectively (including multipath effect).

[0029] The noise level of pseudorange observation is high, while the noise of carrier phase is low (usually millimeter level), but contains ambiguity term . Within a continuous observation arc segment without cycle slip, is constant. By constructing the combination of pseudorange and carrier phase, the pseudorange noise is smoothed and TEC is extracted.

[0030] Add the pseudorange and carrier phase observations:

[0031] Simplify as:

[0032] General hardware bias and are stable in short time, while can be corrected using products, and the mean of noise term is close to zero. Take the average of continuous arc segments (M epochs) to estimate the combined ambiguity:

[0033] k represents the kth epoch, and the estimated is finally substituted into the carrier phase equation:

[0034] Solve for TEC:

[0035] Then, the ionospheric mapping function is calculated using the following projection function (MSLM, Modified Single Layer Model) based on the ionospheric thin layer assumption, and then the ionospheric delay of the oblique path is converted to the ionospheric delay in the vertical direction, and the calculation formula is shown in the following formula (1.9):

[0036] R represents the radius of the earth, and H represents the ionospheric height. In order to ensure the reliability of TEC extraction, the following dynamic smoothing arc length adjustment data quality control method is proposed for the problems of signal loss, multipath effect, rapid change of ionosphere and abnormal observation, etc. The fixed arc length M cannot adapt to the rapid change of ionosphere (such as geomagnetic storm) or stable scene. In this embodiment, the TEC change rate is monitored in real time, and M is dynamically adjusted to optimize the smoothing effect.

[0037] Rough TEC estimation:

[0038] Change rate calculation:

[0039] Wherein, is the interval between epochs.

[0040] Arc length adjustment:

[0041] Wherein: Epoch, applicable to rapid change; Epoch, applicable to stable scene; TECU / sec (adjusted according to the environment).

[0042] Smoothing processing: in order to avoid frequent jumping of M, exponential weighted average is adopted:

[0043] Wherein, Update ambiguity estimation:

[0044] Implementation details: Initialization: set .

[0045] Threshold selection: high latitude area TECU / s, mid-low latitude TECU / s.

[0046] Boundary handling: if the arc segment is interrupted due to a cycle slip, reset .

[0047] This method adaptively adjusts M, taking into account the responsiveness of rapid changes and the accuracy of stable scenes; experiments show that the TEC error decreases from 0.3 TECU for fixed arc segments to 0.1 TECU; unlike traditional fixed window methods, it has higher environmental adaptability.

[0048] (3) Use the ROTI index determination method to screen out ionospheric disturbance area observation values, label them according to different levels, and store them for subsequent model training.

[0049] The ROTI index determination method is as follows, which is applicable to both high and low frequency devices: ROTI is a method for monitoring ionospheric irregular structure activity, and its core is to evaluate the dynamic characteristics of the ionosphere by calculating the rate of change of total electron content (TEC). The following is a brief description of the ROTI method: Data collection: first collect GNSS observation data, then extract ionospheric observation values (TEC) through certain methods (non-difference non-combination PPP method).

[0050] Calculate the TEC rate of change (ROT): for each observation point, calculate the rate of change of TEC in the adjacent time interval. ROT can be calculated by the following formula:

[0051] where, and are the TEC values at the current time and the previous time, and Δt is the time interval.

[0052] Finally, within a certain time interval, calculate the standard deviation of ROT to get the ROTI value. ROTI reflects the fluctuation degree of TEC rate of change, and higher ROTI value usually indicates the activity of ionospheric irregular structure.

[0053] Based on the ROTI value, define the ionospheric disturbance level (reference literature and practical application): No disturbance: TECU / min; Mild disturbance: TECU / min; Moderate disturbance: TECU / min; Severe disturbance: TECU / min; (4) Extract ionospheric observations from low-cost GNSS observation data using the same method as high-frequency GNSS receivers, and save the information of ionospheric observations (including ionospheric piercing point latitude and longitude, time, receiver, and satellite, etc.) for subsequent model data training.

[0054] (5) Use the ionospheric disturbance area and disturbance time extracted by high-frequency equipment as screening conditions for labeling, and mark the ionospheric observations extracted by high-frequency GNSS according to the corresponding disturbance area and time. The marking value is the ionospheric disturbance level of the high-frequency observation in this time and range.

[0055] Labeling process: (a) High-frequency device data processing: Calculate the ROTI of each high-frequency device observation path.

[0056] Determine the disturbance area (centered on IPP, radius 20 km) and time period (5-minute resolution).

[0057] (b) Low-cost GNSS data matching: Align the IPP coordinates and time of low-cost GNSS observations with the disturbance area and time of high-frequency equipment.

[0058] If the IPP of low-cost GNSS observation falls within the disturbance area and the time overlaps, assign the corresponding ROTI level ({0, 1, 2, 3}).

[0059] (c) Data labeling: Add field [ROTI_Level] to each low-cost GNSS observation record.

[0060] Update the data set format: [Time, Receiver_ID, Receiver_Lat, Receiver_Lon, Satellite_PRN, IPP_Lat, IPP_Lon, TEC, ROTI_Level] (6) Extract all low-cost GNSS observation data (including the part confirmed to have occurred disturbance) together, and extract the features according to (receiver-satellite-piercing point area-ionospheric short-term historical observation sequence). Stack different features according to time sequence, and train the next time ionospheric disturbance level as label.

[0061] Objective: To construct a deep learning model that fuses multi-input, multi-scale spatiotemporal features to predict the level of ionospheric perturbation at the next time step (based on ROTI, a classification task, outputting {0, 1, 2, 3}). To capture the rapid perturbation characteristics reflected by ROTI, the following innovative design is proposed to enhance the model's complexity and predictive power.

[0062] ROTI, as a perturbation index, exhibits high temporal sensitivity and spatial heterogeneity, making it difficult for traditional models (such as LSTM and CNN) to simultaneously capture its long-term and short-term dependencies and regional correlations. This innovation proposes a multi-scale spatiotemporal Transformer (MST-Transformer), which combines multi-scale feature extraction, dynamic attention mechanisms, and spatial embedding to optimize the modeling of ROTI perturbation patterns.

[0063] (d) Architecture components: lTime series encoder: enter: .

[0064] Extracting features at different time scales using multi-scale 1D convolution:

[0065] in, It corresponds to short (1-3 minutes), medium (3-5 minutes), and long (5-7 minutes) modes.

[0066] l Fusion of multi-scale features:

[0067] l-Spatial feature embedding: Input: Receiver, satellite, IPP grid features.

[0068] Embedded discrete features:

[0069] l Encoding continuous features: Including receiver latitude and longitude Satellite elevation angle and azimuth IPP latitude and longitude :

[0070] l-Spatial feature fusion:

[0071] l Contextual feature fusion: Inputs: ROTI and TEC mean values ​​of the surrounding receivers, solar / geomagnetic index, and time characteristics.

[0072] Encoding:

[0073] Multi-head fusion and classification: Concatenated features:

[0074] Output disturbance level probability:

[0075] Predicted label:

[0076] (e) Architecture advantages: Multi-scale convolution captures short, medium, and long-term fluctuations of ROTI.

[0077] ROTI-guided dynamic attention highlights disturbance hotspots.

[0078] Spatial embedding integrates regional correlation, adapting to spatial heterogeneity of ROTI.

[0079] (7) After parameter tuning, the model can finally predict the ionospheric disturbance level in a certain area based on the input of multiple receivers and observed satellite pairs, i.e., introducing continuous historical short-term ionospheric observation data in the surrounding area, to predict the future ionospheric disturbance situation in this area and determine the level of ionospheric disturbance phenomenon.

[0080] (f) Prediction process: Input: Observed data of multiple low-cost receivers and satellite pairs: Current and historical TEC, ROT, ROTI sequences; Low-cost receiver, satellite, IPP information; Contextual features (surrounding ROTI, solar / geomagnetic indices).

[0081] Feature extraction: Processed according to the method during training.

[0082] Inference: MST-Transformer outputs the probability of the next ROTI level.

[0083] Output: Predicted disturbance level (0-3) and confidence.

[0084] The implementation basis of each embodiment of the present application is realized through the programmed processing of a device with processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present application are packaged into various modules. Based on this actual situation, on the basis of the above-mentioned embodiments, the embodiment of the present application provides a low-cost GNSS ionospheric disturbance monitoring and early warning system based on Transformer, which is used to execute the low-cost GNSS ionospheric disturbance monitoring and early warning method based on Transformer in the above-mentioned method embodiment.

[0085] The system comprises: a data preprocessing module, used for acquiring and preprocessing observation data of a plurality of low-cost GNSS receivers and satellite pairs; a monitoring and early warning module, used for inputting the preprocessed data into a trained disturbance monitoring and early warning model and outputting an ionospheric disturbance level at the next time; wherein the training of the disturbance monitoring and early warning model comprises: acquiring long-time sequence high-frequency device data and low-cost GNSS receiver observation data; extracting ionospheric observation values from the high-frequency device data, using ROTI index determination method, screening out observation values in the ionospheric disturbance region, and labeling according to different levels; at the same time, extracting ionospheric observation values from the low-cost GNSS receiver observation data; using the screened ionospheric disturbance region and disturbance time as conditions, marking the ionospheric observation values extracted from the low-cost GNSS receiver observation data according to the corresponding disturbance region and disturbance time, and the marking value is the ionospheric disturbance level occurred in this time and range by the high-frequency observation value; stacking different features according to the time sequence as input, and the level of ionospheric disturbance occurred at the next time as label as output, training the deep learning model based on the multi-input, multi-scale spatio-temporal feature fusion constructed by Transformer, and outputting the trained model for disturbance monitoring and early warning.

[0086] The low-cost GNSS ionospheric disturbance monitoring and early warning system based on Transformer provided by the embodiment of the present application faces the present situation that the existing ionospheric monitoring technology relies on high-cost ground stations and high-frequency receivers, it is difficult to realize continuous coverage in the global range, especially in remote areas such as oceans; the early warning system is slow in response, it is difficult to meet the real-time demand; and only the observation data of a single receiver and satellite pair is used, the data utilization rate is low, it is difficult to comprehensively describe the regional ionospheric state, and there is a long time delay. The foregoing several modules are used, the total electron content (TEC) is extracted by the low-cost GNSS receiver and combined with the ROTI index, the data of a plurality of monitoring stations in the region is processed in real time, the dynamic attention mechanism and multi-scale feature extraction of Transformer are used to quickly predict the ionospheric disturbance, and the early warning delay is significantly reduced.

[0087] It should be noted that the system embodiments provided by the present application are used to implement the methods in the above method embodiments, and are also used to implement the methods in other method embodiments provided by the present application, the difference is only that the corresponding function modules are set, and the principle is basically the same as that of the above system embodiments provided by the present application. As long as the person skilled in the art improves the modules in the above system embodiments on the basis of the above system embodiments, refers to the specific technical solutions in other method embodiments, obtains the corresponding technical means by combining technical features, and the technical solutions composed of these technical means, on the premise of ensuring the practicability of the technical solutions, the corresponding system class embodiments are obtained, which are used to implement the methods in other method class embodiments.

[0088] Based on the same inventive concept as the foregoing embodiments, the present embodiment also provides a low-cost GNSS ionospheric disturbance monitoring and early warning device based on a Transformer, comprising a memory and a processor, the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the low-cost GNSS ionospheric disturbance monitoring and early warning method based on the Transformer.

[0089] Based on the same inventive concept as the foregoing embodiments, the present embodiment also provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions make the computer execute the low-cost GNSS ionospheric disturbance monitoring and early warning method based on the Transformer.

[0090] In summary, compared with the traditional method, the present application has the following advantages: (1) Low cost and wide distribution: The present method uses simple and low-cost GNSS monitoring stations for ionospheric disturbance monitoring, avoiding the dependence on expensive high-frequency receivers in traditional methods. This makes the monitoring stations can be arranged in a wider area, especially in the ocean area which is difficult to cover by traditional monitoring means, significantly reducing the economic burden of monitoring.

[0091] (2) Regional data integration capability: The present method uses the ionospheric observation value sequence of all receivers in the region as training data, while the traditional method can only determine whether ionospheric disturbance occurs according to the time sequence of a single receiver and satellite pair. This method has high data utilization rate and can provide more comprehensive ionospheric state information.

[0092] (3) High timeliness based on Transformer: To realize real-time monitoring and early warning of ionospheric disturbances, a multi-scale spatio-temporal Transformer (MST-Transformer) model is adopted to process and analyze the TEC and ROTI data streams of low-cost GNSS monitoring stations in real time. Through multi-scale feature extraction and dynamic attention mechanism, the model efficiently captures short-term mutations and long-term trends of ionospheric disturbances, providing disturbance level prediction (no disturbance, mild, moderate, severe) with second-level response. Compared with traditional methods, the high timeliness of MST-Transformer significantly reduces the potential impact on communication and navigation systems, especially suitable for scenarios with high real-time requirements such as marine navigation.

[0093] (4) Adaptive adjustment of smoothing window: To meet the different needs of ionospheric rapid changes and stable scenarios, an adaptive adjustment of smoothing window carrier phase smoothing pseudo-range algorithm is proposed. By monitoring the ionospheric TEC change rate in real time, the smoothing window length M is dynamically adjusted. This method takes into account the responsiveness of rapid disturbances and the accuracy of stable scenarios, significantly improving the TEC calculation accuracy of low-cost GNSS receivers in widely distributed networks.

[0094] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit it; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solution deviate from the technical solutions of the embodiments of the present application.

Claims

1. A low-cost GNSS ionospheric disturbance monitoring and warning method based on Transformer, characterized in that, The method comprises: Obtaining observation data of a plurality of low-cost GNSS receivers and satellite pairs and preprocessing the observation data; Inputting the preprocessed data into a trained disturbance monitoring and early warning model to output an ionospheric disturbance level at the next time; wherein the training of the disturbance monitoring and early warning model comprises: Obtaining long-time series of high-frequency device data and low-cost GNSS receiver observation data; Extracting ionospheric observation values from the high-frequency device data, using the ROTI index determination method to screen out observation values in the ionospheric disturbance region, and labeling according to different levels; at the same time, extracting ionospheric observation values from the low-cost GNSS receiver observation data; Using the screened ionospheric disturbance region and disturbance time as conditions, marking the ionospheric observation values extracted from the low-cost GNSS receiver observation data according to the corresponding disturbance region and disturbance time, and the marking value is the ionospheric disturbance level of the high-frequency observation value in the time and range; Stacking different features according to the time sequence as input, and the level of ionospheric disturbance at the next time as label, training a deep learning model based on Transformer and constructed with multiple inputs and multi-scale spatio-temporal feature fusion, and outputting the trained model for disturbance monitoring and early warning.

2. The low-cost GNSS ionospheric disturbance monitoring and warning method based on Transformer according to claim 1, characterized in that, The method further comprises: Extracting ionospheric observation values from the high-frequency device data, comprising:

3. The low-cost GNSS ionospheric disturbance monitoring and warning method based on Transformer according to claim 2, characterized in that, Based on the carrier phase smoothing pseudo-range method, constructing a combination of pseudo-range and carrier phase, estimating the combination ambiguity and combining the carrier phase equation to extract TEC. The method further comprises:

4. The low-cost GNSS ionospheric disturbance monitoring and warning method based on Transformer according to claim 1, characterized in that, Monitoring the TEC change rate in real time and dynamically adjusting the arc length. Using the ROTI index determination method to screen out observation values in the ionospheric disturbance region and label according to different levels, comprising: Extracting ionospheric observation values based on observation data and calculating the TEC change rate; In a set time interval, calculate the standard deviation of the TEC change rate to obtain the ROTI value; 5. The low-cost GNSS ionospheric disturbance monitoring and warning method based on Transformer according to claim 4, characterized in that, Defining the ionospheric disturbance level based on the ROTI value. Using the screened ionospheric disturbance region and disturbance time as conditions, marking the ionospheric observation values extracted from the low-cost GNSS receiver observation data according to the corresponding disturbance region and disturbance time, comprising: Calculate the ROTI of each high-frequency device observation path to determine the disturbance region and disturbance time; Align the IPP coordinates and time of low-cost GNSS observation with the disturbance region and time of high-frequency devices, and if the IPP of low-cost GNSS observation falls into the disturbance region and the time overlaps, assign the corresponding ROTI level; 6. The low-cost GNSS ionospheric disturbance monitoring and warning method based on Transformer according to claim 1, characterized in that, Add a field mark to each low-cost GNSS observation record. The deep learning model based on Transformer and constructed with multiple inputs and multi-scale spatio-temporal feature fusion, comprising: A time series encoding and fusion module for using multi-scale convolution to extract and fuse features of different time scales for input TEC and ROTI; a spatial feature embedding and fusion module, configured to perform discrete feature embedding on input receiver, satellite and IPP network features, encode continuous features, and perform spatial feature fusion on the discrete features and the continuous features; a context feature fusion module, configured to encode ROTI, TEC mean, solar / magnetic index and time features of surrounding receivers; a multi-head fusion and classification module, configured to splice time fusion features, spatial fusion features and context fusion features, and output disturbance level probability and prediction labels.

7. The low-cost GNSS ionospheric disturbance monitoring and warning method based on Transformer according to claim 1, characterized in that, The preprocessing comprises: obtaining observation data of a plurality of low-cost GNSS receivers and satellite pairs; extracting ionospheric observation values from the low-cost GNSS receiver observation data, and calculating TEC sequences, TEC change rates and ROTI values; stacking different features according to time sequences as input.

8. A low-cost GNSS ionospheric disturbance monitoring and warning system based on Transformer, characterized in that, The data preprocessing module is configured to obtain observation data of a plurality of low-cost GNSS receivers and satellite pairs and perform preprocessing. The monitoring and early warning module is configured to input the preprocessed data into a trained disturbance monitoring and early warning model, and output ionospheric disturbance levels at the next time; wherein the training of the disturbance monitoring and early warning model comprises: obtaining long-time sequence high-frequency device data and low-cost GNSS receiver observation data; extracting ionospheric observation values from the high-frequency device data, using the ROTI index determination method to filter out observation values in the ionospheric disturbance region, and labeling according to different levels; at the same time, extracting ionospheric observation values from the low-cost GNSS receiver observation data; using the filtered ionospheric disturbance region and disturbance time as conditions, marking the ionospheric observation values extracted from the low-cost GNSS receiver observation data according to the corresponding disturbance region and disturbance time, and the marking value is the ionospheric disturbance level of the high-frequency observation value in this time and range; stacking different features according to time sequences as input, and the next time ionospheric disturbance level as label, training a deep learning model based on Transformer multi-input and multi-scale spatio-temporal feature fusion, and outputting the trained model for disturbance monitoring and early warning. The non-transitory computer readable storage medium stores computer instructions, and the computer instructions cause the computer to perform the Transformer-based low-cost GNSS ionospheric disturbance monitoring and early warning method according to any one of claims 1 to 7. 9.A low-cost GNSS ionospheric disturbance monitoring and warning device based on a Transformer, characterized in that, The non-transitory computer readable storage medium stores computer instructions, and the computer instructions cause the computer to perform the Transformer-based low-cost GNSS ionospheric disturbance monitoring and early warning method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, comprising: ​

Images