A large height difference scene vertical displacement correction monitoring method based on GNSS monitoring
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID TIBET ELECTRIC POWER CO LTD LHASA POWER GENERATION CO
- Filing Date
- 2026-04-14
- Publication Date
- 2026-07-24
Smart Images
Figure SMS_5 
Figure SMS_10 
Figure SMS_13
Abstract
Description
Technical Field
[0001] This invention relates to the field of surveying and mapping methods, specifically to a vertical displacement correction monitoring method for large elevation difference scenes based on GNSS monitoring. Background Technology
[0002] Currently, the main technologies used in deformation monitoring include: traditional leveling, close-range photogrammetry, synthetic aperture radar (SAR), 3D laser scanning, and Global Navigation Satellite System (GNSS). Each of these technologies has its own characteristics: leveling, while highly accurate, is difficult to implement for dynamic monitoring; photogrammetry and laser scanning are limited by line-of-sight conditions and have significant shortcomings in long-distance monitoring; SAR technology is more suitable for large-scale deformation monitoring, but its temporal resolution is relatively low. In contrast, GNSS technology, with its all-weather capability, high precision, and high degree of automation, has become an important technological means for modern deformation monitoring.
[0003] GNSS deformation monitoring typically employs a single-baseline relative positioning mode, which, theoretically, can achieve millimeter-level positioning accuracy after effectively eliminating systematic errors such as clock bias, orbital errors, and atmospheric delay. However, in complex environments, obstruction and multipath effects degrade observation quality; when there are large elevation differences or excessive distances between baselines, tropospheric delay increases significantly, severely affecting elevation direction accuracy. Current research largely focuses on tropospheric model correction and delay error reduction. Jiang et al. proposed a method that uses additional ground point tropospheric delay prior information as a constraint, employing high-precision tropospheric delay obtained from long-term GNSS observations at ground stations as a constraint. This effectively reduces residual errors caused by large elevation differences. However, this method relies on years of continuous, high-quality GNSS observation data from ground stations to obtain high-precision tropospheric delay prior information, making rapid deployment at newly built or temporary monitoring stations difficult. Based on the sliding window concept, Huang Liangke constructed a high-precision ZWD and ZTD vertical profile grid model that takes into account spatiotemporal factors, improving the accuracy and stability of ZWD and ZTD elevation reduction. However, this model requires a dense three-dimensional grid and continuous sliding window updates, placing high demands on the distribution of GNSS stations within the monitoring area. Yao Yibin et al. established a real-time tropospheric delay model based on CORS and meteorological station data, combining it with the Ntrip protocol to achieve imperceptible RTK (Real-Time Kinematic) delay correction. However, its high dependence on the accuracy of meteorological observation networks and sensors limits its application.
[0004] Static relative positioning, with its millimeter-level or even sub-millimeter-level observation accuracy, has become the preferred solution for long-term continuous deformation monitoring. This method utilizes multiple receivers to simultaneously observe common satellites and employs differential processing between carrier phase observations to effectively eliminate common errors such as orbital errors, clock errors, and atmospheric delays, thereby significantly improving the reliability of the positioning results. Especially in the stability assessment of critical infrastructure such as bridges, dams, and landslides, static relative positioning can establish deformation trend models through long-term baseline calculations, providing quantitative evidence for structural health diagnosis and disaster prevention.
[0005] The basic process of static relative positioning can be divided into three stages: observation, differential processing, and baseline calculation. In the observation stage, GNSS receivers are set up at both the monitoring point to be measured and the known reference point, simultaneously receiving carrier phase and pseudorange observations from multiple satellites. Since both receivers simultaneously observe the same satellite signals, many error sources (such as satellite orbital deviations, clock errors, and atmospheric delays) are strongly correlated in space or time, and can be significantly reduced through differential processing.
[0006] Generally, tropospheric delay is divided into relative tropospheric delay error and absolute tropospheric error. Relative tropospheric delay error characterizes the difference in tropospheric delay between different stations. In mountainous environments with large elevation differences, this error becomes one of the main limiting factors for the accuracy of vertical displacement measurements, and the effect of correcting this error using existing models is very limited.
[0007] Extensive practical calculations show that the average cutoff elevation angle of the GNSS satellites involved in differential calculations is approximately 30°, and a 1mm difference results in an elevation error of about 4mm. When the elevation difference is 100m, the relative tropospheric delay between stations is 2.6~3.8cm; when the elevation difference is 500m, it can reach 13.4~20.5cm. The error in relative tropospheric delay is absorbed into the elevation direction at approximately a 3:1 ratio; even with an elevation difference of only 100m, the impact on elevation accuracy can reach nearly 10cm.
[0008] Currently, tropospheric delay correction mainly employs three methods: empirical modeling, parameter estimation, and external correction. Empirical modeling establishes a mathematical model of atmospheric delay in the zenith direction based on typical meteorological data and ground observation parameters. Commonly used models include the Saastamoinen model, the Hopfield model and its improved forms, and the Black model. The main differences lie in the assumptions regarding vertical refractive index distribution and the form of the mapping function, with the first two models being the most widely used. With accurate meteorological parameters, both the Saastamoinen and Hopfield models can achieve sub-millimeter accuracy in correcting dry delay. However, due to the complex spatiotemporal variations of water vapor, the accuracy of wet delay correction is lower (on the centimeter scale). Comparatively, the Saastamoinen wet delay model has a better accuracy (2-5 cm) than the Hopfield model. The Hopfield model is more sensitive to temperature errors because it includes a squared temperature term, while the Saastamoinen model only relies on a linear temperature term. Overall, the Saastamoinen model has a better correction effect.
[0009] Empirical models approximate the tropospheric delay at the zenith using analytical methods. However, because they cannot perfectly simulate the actual atmospheric delay, approximately 5% error remains even after model correction. In the case of short baselines, if meteorological conditions between stations are similar, differential processing can effectively mitigate the impact of these residual errors. However, when the baseline is long or the meteorological differences between the two ends are significant, the impact of the residual tropospheric delay cannot be ignored and must be considered.
[0010] Therefore, in some monitoring scenarios with large elevation differences, such as when monitoring pumped storage dams, the traditional monitoring methods, such as tropospheric delay models and meteorological station compensation methods, have many limitations in terms of data dependence, deployment cost, and timeliness. Therefore, it is necessary to provide a correction method to correct the vertical displacement of monitoring stations with large elevation differences. Summary of the Invention
[0011] The purpose of this invention is to provide a vertical displacement correction monitoring method for large elevation difference scenarios based on GNSS monitoring.
[0012] To achieve the above objectives, one embodiment of the present invention provides a vertical displacement correction monitoring method for large elevation difference scenarios based on GNSS monitoring, comprising the following steps: Step S1: Set up a reference station and a monitoring station within the monitoring area. The reference station and the monitoring station are at different vertical heights. Determine and record the elevation difference ΔH between each reference station and the monitoring station. Step S2: Deploy GNSS receivers at the reference station and monitoring stations. Using the static relative positioning method, obtain the daily displacement value of each monitoring station after observation, differential processing, and baseline calculation. Take the average of the displacement values calculated for the previous N days at the current monitoring time. This average value is the current reference value for each monitoring station. ; Step S3: Establish a fitting correction model for vertical displacement, and calculate the correction deviation using the fitting correction model; The method for establishing the fitting correction model is as follows: S31. Select at least m+1 monitoring stations with relatively uniform elevation distribution from the monitoring stations as modeling stations, and obtain the elevation value ΔU of each monitoring station and the elevation difference ΔH between each monitoring station and the reference station obtained by the current monitoring station through the GNSS receiver; the relatively uniformity of the present invention means that the distance between two adjacent monitoring stations is relatively close, and the ratio between the maximum spacing and the minimum spacing is 1~1.3:1.
[0013] S32. Substitute the elevation value of each monitoring station and the elevation difference ΔH between each monitoring station and the baseline station into the following model formula: ; Where: ΔU is the elevation value of each monitoring station; The current baseline value for each monitoring station; m is the polynomial order, m is a positive integer from 1 to 4, and m is determined according to the monitoring accuracy requirements; S33. Solving for the optimal coefficients using the least squares method Substituting the optimal coefficients into the model formula yields the formula for calculating the correction deviation. The correction deviation y is then calculated based on this formula.
[0014] The final calculated elevation value U = ΔU - y for each monitoring station is obtained by calculating the correction deviation y and the elevation value ΔU for each monitoring station.
[0015] In the preferred embodiment of the present invention, the elevation difference between each reference station in step S1 is not less than 100 meters, and N in step S2 is 3 days.
[0016] In a preferred embodiment of the present invention, the horizontal distance between each monitoring station in step S1 is less than 15km, and the number of monitoring stations is not less than 6; the reference station is the station with the largest elevation value.
[0017] In a preferred embodiment of the present invention, the elevation difference between each modeling station in step S31 is between 100m and 200m.
[0018] In summary, the present invention has the following advantages: This invention proposes a correction method that utilizes continuous GNSS time-series observation data from multiple monitoring stations to analyze the spatiotemporal correlation between the vertical displacement difference Δh and the elevation difference ΔH between the stations. Furthermore, by constructing a correction model, adaptive correction of the vertical displacement of monitoring stations with large elevation differences within a small area can be achieved. At the same time, the polynomial order in the model can be set according to the monitoring accuracy requirements, which helps to minimize the amount of computation while ensuring accuracy. Detailed Implementation
[0019] This invention provides a vertical displacement correction monitoring method for large elevation difference scenarios based on GNSS monitoring, comprising the following steps: Step S1: Set up a reference station and a monitoring station within the monitoring area. The reference station and the monitoring station are at different vertical heights. Determine and record the elevation difference ΔH between each reference station and the monitoring station. Step S2: Deploy GNSS receivers at the reference station and monitoring stations. Using the static relative positioning method, obtain the daily displacement value of each monitoring station after observation, differential processing, and baseline calculation. Take the average of the displacement values calculated for the previous N days at the current monitoring time. This average value is the current reference value for each monitoring station. ; Step S3: Establish a fitting correction model for vertical displacement, and calculate the correction deviation using the fitting correction model; The method for establishing the fitting correction model is as follows: S31. Select at least m+1 monitoring stations with relatively uniform elevation distribution from the monitoring stations as modeling stations, and obtain the elevation value ΔU of each monitoring station and the elevation difference ΔH between each monitoring station and the reference station obtained by the current monitoring station through the GNSS receiver. S32. Substitute the elevation value of each monitoring station and the elevation difference ΔH between each monitoring station and the baseline station into the following model formula: ; Where: ΔU is the elevation value of each monitoring station; The current baseline value for each monitoring station; m is the polynomial order, m is a positive integer from 1 to 4, and m is determined according to the monitoring accuracy requirements; S33. Solving for the optimal coefficients using the least squares method Substituting the optimal coefficients into the model formula yields the formula for calculating the correction deviation. The correction deviation y is then calculated based on this formula.
[0020] The final calculated elevation value U = ΔU - y for each monitoring station is obtained by calculating the correction deviation y and the elevation value ΔU for each monitoring station.
[0021] In step S1, the elevation difference between each benchmark station is not less than 100 meters, and in step S2, N is 3 days.
[0022] In step S1, the horizontal distance between each monitoring station is less than 15km, and the number of monitoring stations is no less than 6; the benchmark station is the station with the largest elevation value.
[0023] In step S31, the elevation difference between each modeling station is between 100m and 200m. Example
[0024] Step S1: Select a power plant scenario and set up 10 monitoring stations YH01~YH10 within the monitoring area. YH01~YH10 together form a monitoring network. YH01~YH10 are distributed at different altitudes, with varying elevation differences between stations, ranging from approximately 28m to approximately 600m. Since station YH04 has the highest altitude, it is used as the baseline station, and the others are used as monitoring stations for the experiment. The elevation differences between the baseline station YH04 and other stations are shown below: YH04 is the baseline station, and the rest are monitoring stations.
[0025] Step S2: Deploy GNSS receivers at the reference station and monitoring stations. Using the static relative positioning method, obtain the daily displacement value of each monitoring station after observation, differential processing, and baseline calculation. Take the average of the displacement values calculated for the previous N days at the current monitoring time. This average value is the current reference value for each monitoring station. .
[0026] The model formula of this invention requires a baseline value to be obtained in advance, and the monitoring of the baseline value needs to be calculated from data from several days prior. For example, the average of the monitoring data obtained from the previous three days is used as the baseline value. When monitoring on the fourth day, the data from the previous three days can be used as the calculation parameters for the baseline value. When monitoring on the fifth day, the data from the fourth day is included in the calculation parameters for the baseline value, that is, the monitoring data from the second, third, and fourth days are used as the basis for calculation.
[0027] The baseline station and monitoring stations collected observation data from November 5th to November 20th, 2024, with a sampling interval of 10 seconds and a cutoff elevation angle set to 15°. Other variables that could affect positioning accuracy were controlled. The horizontal distance between all stations was within 15km, therefore, errors due to baseline length were not considered. The data was post-processed to obtain the vertical displacement results for each monitoring station over 24 hours. Then, the data from the previous three days for each station were calculated separately, and the average value was used to obtain the baseline value Δu for the vertical displacement of each monitoring station. Among all monitoring stations, this invention selected six stations with relatively uniform elevation distribution: YH05, YH06, YH07, YH09, YH10, and YH02, for modeling. The ΔU-Δu and ΔH for each station from November 11th to November 18th (a total of 8 days) were substituted into the formula, with m = 1, 2, 3, and 4. Polynomial regression fitting should appropriately reduce the polynomial order to minimize divergence caused by excessively high order leading to large independent variable values. Therefore, the highest polynomial order used in this invention is 4. The daily model coefficients are fitted based on the least squares principle; the least squares calculation method is existing technology.
[0028] For example, using the least squares method to solve for the optimal coefficients ,make
[0029] The sum of squared errors is expressed as:
[0030] In matrix representation, it can be written as a Vandermonde matrix:
[0031] make , , The regression equation can then be expressed as: The optimal coefficient A is obtained by solving for: To evaluate the polynomial fit, the coefficient of determination is used. The evaluation is conducted using the following calculation formula:
[0032] This invention solves for the optimal coefficients using the least squares method. Substituting the optimal coefficients into the model formula yields the formula for calculating the correction deviation. The correction deviation y is then calculated based on this formula.
[0033] The final calculated elevation value U = ΔU - y for each monitoring station is obtained by calculating the correction deviation y and the elevation value ΔU for each monitoring station. After calculating the elevation value U, the accuracy of the model of this invention can be verified by obtaining the residual RMSE and the coefficient of determination using existing methods. This allows us to obtain the fitting index.
[0034] That is, this invention fits the model coefficients for each day based on the least squares principle and statistically analyzes the fitting effect. The fitting evaluation indexes of the four models are shown below. The four models referred to here are the models obtained when m takes the values 1, 2, 3, and 4. When the values of m are different, it means that the order is different, and thus different coefficients will be obtained.
[0035] In the above indicators, linear means m=1, quadratic polynomial means m=2, cubic polynomial means m=3, and quartic polynomial means m=4.
[0036] As shown in the table above, the higher the polynomial order, the more significantly the model fitting accuracy and interpretability are improved. The quartic polynomial achieved the lowest residual RMSE and the highest coefficient of determination across all dates, especially on November 11th and 12th, when linear models and lower-order polynomials performed weakly, increasing the coefficient of determination to 0.8633 and 0.9105 respectively. The cubic polynomial followed closely, reducing the RMSE to around 3.0 mm and increasing the coefficient of determination to 0.9164-0.9293 on most dates, and also exhibiting good robustness to abnormally volatile days (such as November 18th). The quadratic polynomial showed significant improvement over the linear model, with the average R² increasing from around 0.87 to around 0.90, but its marginal benefit was limited under highly nonlinear or noisy conditions. As observation conditions stabilized, such as from November 15th to 17th, the additional benefits of the quartic and cubic models gradually converged: the RMSE improvement was less than 0.3 mm, and the coefficient of determination R² improvement was less than 0.03. Overall, the quartic polynomial combines the highest accuracy with the strongest robustness, making it suitable for scenarios with extremely high accuracy requirements; when there are stricter constraints on computing resources or real-time performance, the cubic or quadratic polynomial can provide residuals of 3-4 mm and an interpretability of ≥0.90 in most cases.
[0037] Based on different elevation differences, the fitted values calculated using the four models and the residuals of the elevation displacement deviations are summarized and compared. The results are shown in the table below: The results show that when the elevation difference between stations is 0-100m, the positioning accuracy is high, and modeling is not very meaningful and can be ignored. When the elevation difference is 100-200m, the corresponding data for YH06 shows that the residual RMSE of the quadratic and cubic polynomials are comparable, approximately 0.6mm, indicating good and almost identical fitting effects. When the elevation difference is 200-300m, the corresponding data for YH07 shows that the quadratic polynomial has a good fitting effect, with a residual RMSE of approximately 0.7mm. When the elevation difference is 300-400m, the data for YH08 shows that the quartic polynomial has a good fitting effect, with a residual RMSE of 5.5mm. When the elevation difference is above 600m, the corresponding data for YH02 shows that the quartic polynomial has the best fitting effect, with a residual RMSE of 0.1mm. In summary, both linear and nonlinear models show good fitting effects. Although the fitting effects of each model vary under different elevation differences, the values are basically equivalent.
[0038] Experimental Example: Multi-model Performance Validation This experiment studies the performance of four models in correcting elevation displacement. YH01, YH03, and YH09 in the pumped storage scenario monitoring stations are selected as research objects. Based on the daily model and the elevation difference between the monitoring station and the benchmark station, the elevation displacement correction is fitted and corrected on the post-processed elevation displacement data of the three monitoring stations to verify the error correction effect of the linear and nonlinear models in this study.
[0039] (1) Linear regression model In the regression model of this study, when m=1, The elevation differences between the three monitoring stations YH01, YH03, and YH09 and the benchmark station are substituted into the model fitting value, which is also the correction value for elevation displacement. The elevation displacement is then corrected, and the external compliance accuracy of the corrected vertical displacement is statistically analyzed. The external compliance accuracy of ΔU before and after the linear model correction is shown in the table below.
[0040] The table above shows that after linear model correction, the external conformity accuracy of YH01, which has an elevation difference of 219m from the benchmark station, is improved by about 18.3%. After correction, the external conformity accuracy of YH03, which has an elevation difference of 393m from the benchmark station, is improved by about 94.5%. After correction, the external conformity accuracy of YH09, which has an elevation difference of 380m, is improved by about 86.3%.
[0041] (2) Quadratic polynomial model When m=2, the polynomial model In the model coefficients for each day , , Substituting the elevation difference of each monitoring station into the model fitting value, the elevation displacement is corrected. The external consistency accuracy of ΔU before and after the quadratic polynomial correction is shown below: The table above shows that after correction using a quadratic polynomial model, the external conformity accuracy of YH01, which has an elevation difference of 219m from the benchmark station, is improved by about 18.3%. After correction, the external conformity accuracy of YH03, which has an elevation difference of 393m from the benchmark station, is improved by about 95.5%. After correction, the external conformity accuracy of YH09, which has an elevation difference of 380m, is improved by about 86.3%.
[0042] (3) Cubic polynomial model When m=3, the polynomial model The model coefficients for each day are then corrected to elevation displacement using the same method as for the quadratic polynomial. The external compliance accuracy is then statistically analyzed. The external compliance accuracy of ΔU before and after the cubic polynomial correction is shown in the table below: The table above shows that after correction using a cubic polynomial model, the external conformity accuracy of YH01, which has an elevation difference of 219m from the benchmark station, is improved by about 35%. After correction, the external conformity accuracy of YH03, which has an elevation difference of 393m from the benchmark station, is improved by about 96%. After correction, the external conformity accuracy of YH09, which has an elevation difference of 380m, is improved by about 82.9%.
[0043] (4) Quadratic polynomial model When m=3, the polynomial model formula is as follows: ; The model coefficients for each day are then corrected to elevation displacement using the same method as for the quadratic polynomial. The external compliance accuracy is then statistically analyzed. The external compliance accuracy of ΔU before and after the fourth-order polynomial correction is shown in the table below: The table above shows that after correction using a fourth-order polynomial model, the external conformity accuracy of YH01, which has an elevation difference of 219m from the benchmark station, is improved by about 53%. After correction, the external conformity accuracy of YH03, which has an elevation difference of 393m from the benchmark station, is improved by about 91.5%. After correction, the external conformity accuracy of YH09, which has an elevation difference of 380m, is improved by about 81.8%.
[0044] In summary, all four polynomial models of this invention demonstrate practical effectiveness in correcting vertical displacement accuracy, significantly reducing the ΔU error. Higher-order models exhibit stronger expressive power in error fitting, resulting in greater accuracy improvements in scenarios with large elevation differences, such as YH03 / YH09; while in scenarios with low elevation differences, such as YH01, lower-order models are sufficient for most requirements. Overall, the accuracy gain from increasing model order decreases progressively, and it is recommended to choose the simplest model possible while still meeting accuracy requirements. Experimental results show that polynomial regression correction based on elevation difference characteristics can reduce vertical displacement errors to the centimeter level (an accuracy improvement of approximately 60%), thereby significantly enhancing the reliability of GNSS vertical measurements.
[0045] With the continuous development of GNSS technology, monitoring high-precision deformable bodies such as large-scale engineering projects using GNSS technology has become an important means. Providing all-weather, uninterrupted, high-precision, real-time, stable, and reliable monitoring services for deformable bodies has become a trend. However, in GNSS deformation monitoring of pumped storage power stations, when there is a large elevation difference between the reference station and the monitoring station, the correlation of spatial correlation errors (such as atmospheric delay error) between the stations decreases, leading to a significant decrease in the monitoring accuracy in the elevation direction. To address this problem, this paper, combined with the actual scenario of pumped storage power stations, proposes a corresponding correction method based on the distribution law of elevation displacement monitoring deviation under large elevation difference environments. This invention proposes a multinomial regression model correction method based on the vertical displacement difference and elevation difference of monitoring stations. The vertical displacement difference is extracted using GNSS monitoring data from multiple monitoring stations. ) and elevation difference ( Based on the spatiotemporal correlation of vertical displacement, a polynomial regression model was constructed to correct the vertical displacement of monitoring stations with large elevation differences within a small area. The results show that when the elevation difference is between 100 and 300 m, the quadratic polynomial model performs best, with a residual RMSE of approximately 0.7 mm; when the elevation difference is between 300 and 400 m, the quartic polynomial model performs best, with a residual RMSE of 5.5 mm; when the elevation difference is between 400 and 500 m, the quartic polynomial model has a fitting residual RMSE of 3.2 mm; and when the elevation difference is above 600 m, the quartic polynomial model has a fitting residual RMSE of 0.1 mm. After correcting the vertical displacement using linear and nonlinear models, the accuracy is significantly improved, with a peak correction rate reaching 96%.
[0046] While specific embodiments of the present invention have been described in detail, this should not be construed as limiting the scope of protection of this patent. Various modifications and variations that can be made by those skilled in the art without inventive effort within the scope described in the claims still fall within the scope of protection of this patent.
Claims
1. A method for vertical displacement correction monitoring in scenarios with large elevation differences based on GNSS monitoring, characterized in that, Includes the following steps: Step S1: Set up a reference station and a monitoring station within the monitoring area. The reference station and the monitoring station are at different vertical heights. Determine and record the elevation difference ΔH between each reference station and the monitoring station. Step S2: Deploy GNSS receivers at the reference station and monitoring stations. Using the static relative positioning method, obtain the daily displacement value of each monitoring station after observation, differential processing, and baseline calculation. Take the average of the displacement values calculated for the previous N days at the current monitoring time. This average value is the current reference value for each monitoring station. D ; Step S3: Establish a fitting correction model for vertical displacement, and calculate the correction deviation using the fitting correction model; The method for establishing the fitting correction model is as follows: S31. Select at least m+1 monitoring stations with relatively uniform elevation distribution from the monitoring stations as modeling stations, and obtain the elevation value ΔU of each monitoring station and the elevation difference ΔH between each monitoring station and the reference station obtained by the current monitoring station through the GNSS receiver. S32. Substitute the elevation value of each monitoring station and the elevation difference ΔH between each monitoring station and the baseline station into the following model formula: ； Where: ΔU is the elevation value of each monitoring station; Δ The current baseline value for each monitoring station; m is the polynomial order, m is a positive integer from 1 to 4, and m is determined according to the monitoring accuracy requirements; S33. Solving for the optimal coefficients using the least squares method Substituting the optimal coefficients into the model formula yields the formula for calculating the correction deviation. The correction deviation y is then calculated based on this formula. ； The final calculated elevation value U = ΔU - y for each monitoring station is obtained by calculating the correction deviation y and the elevation value ΔU for each monitoring station.
2. The vertical displacement correction monitoring method for large elevation difference scenarios based on GNSS monitoring as described in claim 1, characterized in that: In step S1, the elevation difference between each benchmark station is not less than 100 meters, and in step S2, N is 3 days.
3. The vertical displacement correction monitoring method for large elevation difference scenarios based on GNSS monitoring as described in claim 1, characterized in that: In step S1, the horizontal distance between each monitoring station is less than 15km, and the number of monitoring stations is no less than 6; the benchmark station is the station with the largest elevation value.
4. The vertical displacement correction monitoring method for large elevation difference scenarios based on GNSS monitoring as described in claim 1, characterized in that: In step S31, the elevation difference between each modeling station is between 100m and 200m.