Coal mine area settlement prediction method based on improved GM (1, 1) model and SBAS-InSAR

Abstract

The invention discloses a coal mine area settlement prediction method based on an improved GM (1, 1) model and SBAS-InSAR, and the method comprises the steps: obtaining the satellite image data of a target coal mine area, and processing the satellite image data to obtain a coherence coefficient graph; multiple errors are further removed through two times of earth surface deformation inversion; calculating to obtain a settlement rate map of the target coal mine area; selecting a main observation point in the most obvious settlement area of the target coal mine area by using the graph, and extracting data of the main observation point in different periods; and finally, establishing an improved gray GM (1, 1) model by using the data to predict the settlement amount of the target coal mine area, and setting a threshold value for verification to enable the model to reach the required precision. Through data verification, according to the method, the observation data of the SBAS-InSAR and the improved gray GM (1.1) model are combined to predict the settlement amount of the mining area, high prediction precision can be continuously kept, the calculation efficiency is high, and data support is provided for early warning of goaf settlement in advance.

Description

Technical Field

[0001] The present invention belongs to the field of coal mine terrain change detection, and specifically relates to a method for predicting coal mine area subsidence based on an improved GM(1,1) model and SBAS-InSAR. Background Art

[0002] The continuous mining of coal resources will form irregular mined-out areas. These mined-out areas will cause subsidence and damage to the mined-out areas and the surrounding surface due to the stress damage of the overlying strata, resulting in adverse effects such as threats to life safety, damage to the ecological environment, and induction of geological disasters. How to monitor and early warn of sudden disasters and reduce the losses caused by major hazards has always been the focus of attention.

[0003] InSAR technology has advantages such as high spatial resolution, high sensitivity to deformation, and wide coverage. It has been deeply studied in surface deformation monitoring, geological disaster monitoring, and glacier movement, and has been widely used especially in surface deformation monitoring. Among them, SBAS-InSAR is an InSAR time series analysis method based on multi-master images. By setting spatio-temporal baseline thresholds in obtaining millimeter-level accurate surface slow deformation information, the influence of spatial and temporal decorrelation can be reduced. Then, through the singular value decomposition (SVD) method, the deformation phase and other error phases can be separated, which can effectively improve the accuracy of surface deformation monitoring in coal mine areas. In addition, the grey GM(1,1) prediction model requires less information, has a small amount of computation, and high modeling accuracy, and is an effective tool for dealing with prediction problems of small samples and small data volumes. However, in some areas with large fluctuations in settlement data, the prediction accuracy will be low.

[0004] Therefore, how to improve the grey GM(1.1) model to improve the prediction accuracy and use the observation data of SBAS-InSAR for deformation prediction of coal mine mined-out areas is the research direction of the present invention. Summary of the Invention

[0005] In view of the problems existing in the above-mentioned prior art, the present invention provides a method for predicting coal mine area subsidence based on an improved GM(1,1) model and SBAS-InSAR. The improved grey GM(1.1) model is established by using the observation data of SBAS-InSAR to predict the subsidence amount of the mining area, which can continuously maintain a high prediction accuracy and provide data support for early warning of mined-out area subsidence.

[0006] To achieve the above object, the technical solution adopted by the present invention is: a method for predicting coal mine area subsidence based on an improved GM(1,1) model and SBAS-InSAR, and the specific steps are as follows:

[0007] Step 1: Obtain satellite image data of the target coal mine area.

[0008] Step 2: Collect the ALOS DEM data of 12.5 m in the target coal mine area in Step 1 to remove the topographic phase of the satellite image data in Step 1.

[0009] Step 3: For the satellite image data processed in Step 2, select the super master image according to the principle of the number of primary and secondary image pairs, and generate a spatio-temporal baseline map and a temporal baseline map around the super master image.

[0010] Step 4: Generate a coherence coefficient map based on the spatio-temporal baseline map and the temporal baseline map, perform filtered interferometry and phase unwrapping processing on the satellite image data, and select effective GCP control points to eliminate orbit errors and slope phases.

[0011] Step 5: First, perform the first surface deformation inversion on the satellite image data processed in Step 4 to obtain the preliminary settlement deformation rate in the target coal mine area, and remove the remaining topographic phase, elevation error, and orbit error in the interference phase; then, perform the second surface deformation inversion on the basis of the obtained settlement deformation rate, calculate the displacement in the time series, and further remove the error of the atmospheric phase by combining the atmospheric high-pass filtering and low-pass filtering methods.

[0012] Step 6: For the results of the two inversions in Step 5, calculate the corresponding average deformation rate and temporal deformation amount in the target coal mine area through the SVD algorithm, perform geocoding, and obtain the settlement rate map of the target coal mine area.

[0013] Step 7: According to the settlement rate map of the target coal mine area, select an observation point in the area with the most obvious settlement in the target coal mine area as the main observation point, extract the data of the main observation point at different times, and use the part of the data with an earlier time sequence as the training set in chronological order, and the rest as the test set.

[0014] Step 8: Establish an improved grey GM(1,1) model using the training set in Step 7.

[0015] Step 9: Use the improved grey GM(1,1) model established in Step 8 to predict the settlement amount of the main observation point determined in Step 7; compare it with the actual settlement amount data at different times in the test set. If the error between the predicted value and the actual value is less than the set threshold, retain the current model for predicting the settlement amount of other observation points in the future; otherwise, continuously repeat Steps 8 and 9 until the error between the model predicted value and the actual value is less than the set threshold and stop.

[0016] Furthermore, the specific content of Step 7 is as follows:

[0017] ① Establish an improved grey GM(1,1) model through matlab software, and its formula is:

[0018] Take the data in the training set as the original data, and take the square root of each element of the original data to obtain a new sequence of numbers:

[0019]

[0020] Then sum them up to get:

[0021]

[0022] Based on the newly generated data sequence, establish a first-order differential equation:

[0023]

[0024] Generate the mean value of the accumulated data to obtain B and the constant term vector Y n , and obtain the matrix composed of a and u

[0025]

[0026] z (1) (k)=0.5x (1) (k - 1)+0.5x (1) (k), k = 2, 3…n

[0027]

[0028] Substitute into the first-order differential equation to derive the time response formula:

[0029]

[0030] Perform cumulative subtraction reduction to obtain the prediction model. Square the predicted result x (0) (t + 1) to restore the predicted value to the scale of the original data

[0031]

[0032] ② Then calculate the residual:

[0033]

[0034] ③ Substitute the calculated residual data into the improved grey GM(1,1) model established in step ①, and repeat the calculation process of step ① to obtain the predicted value of the residual

[0035]

[0036] ④ Perform residual correction to obtain the final predicted value:

[0037]

[0038] Compared with the prior art, the present invention first obtains satellite image data of a target coal mining area, generates a spatio-temporal baseline map and a temporal baseline map, and then obtains a coherence coefficient map, performs filtered interference and phase unwrapping processing on it, and selects effective GCP control points to eliminate orbital errors and slope phases; then further removes various errors through two surface deformation inversions; then obtains a settlement rate map of the target coal mining area through the SVD algorithm; uses this map to select main observation points in the area with the most obvious settlement in the target coal mining area, and extracts data of the main observation points at different times; finally, uses the obtained data to establish an improved grey GM(1,1) model to realize the prediction of the settlement amount of the target coal mining area, and sets a threshold for verification, and obtains the required prediction model after meeting the requirements; through data verification, the method of combining the observation data of SBAS-InSAR and the improved grey GM(1,1) model in the present invention for predicting the settlement amount of the mining area can continuously maintain a high prediction accuracy and has a high calculation efficiency, providing data support for early warning of goaf settlement. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is the spatio-temporal baseline map generated in the embodiment of the present invention;

[0040] Figure 2 is the temporal baseline map generated in the embodiment of the present invention;

[0041] Figure 3 is the imaging map of the first surface deformation inversion in the embodiment of the present invention;

[0042] Figure 4 is the imaging map of the second surface deformation inversion in the embodiment of the present invention;

[0043] Figure 5 is the deformation rate map generated in the embodiment of the present invention;

[0044] Figure 6 is the location map of the P1 settlement point selected in the embodiment of the present invention;

[0045] Figure 7 is the comparison map of the settlement prediction value and the actual surface deformation in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] The present invention will be further described below.

[0047] Embodiment: The specific steps are as follows:

[0048] Step 1: Download the C-band sentinel-1A radar satellite image set launched by the European Space Agency, and select a target coal mining area in the satellite image set.

[0049] Step 2: Collect the ALOS DEM data of 12.5 m in the target coal mine area in Step 1 to remove the topographic phase of the satellite image data in Step 1.

[0050] Step 3: For the satellite image data processed in Step 2, select the super master image according to the principle of the number of primary and secondary image pairs, and generate a spatio-temporal baseline map and a temporal baseline map around the super master image, as Figure 1 and 2 shown.

[0051] Step 4: Generate a coherence coefficient map based on the spatio-temporal baseline map and the temporal baseline map, perform filtered interferometry and phase unwrapping processing on the Sentinel-1 satellite image data, and select effective GCP control points to eliminate orbit errors and slope phases.

[0052] Step 5: First, perform the first surface deformation inversion on the satellite image data processed in Step 4 as Figure 3 shown, obtain the preliminary settlement deformation rate of the target coal mine area, and remove the residual topographic phase, elevation error, and orbit error in the interference phase; then, based on the obtained settlement deformation rate, perform the second surface deformation inversion as Figure 4 shown, calculate the displacement in the time series, and further remove the error of the atmospheric phase by combining the atmospheric high-pass filtering and low-pass filtering methods.

[0053] Step 6: For the results of the two inversions in Step 5, use the SVD algorithm to calculate the corresponding average deformation rate and time series deformation amount of the target coal mine area, perform geocoding, and obtain the settlement rate map of the target coal mine area as Figure 5 shown.

[0054] Step 7: According to the settlement rate map of the target coal mine area, select the P1 observation point as the main observation point in the area with the most obvious settlement in the target coal mine area as Figure 6 shown, and extract 18 periods of data of the P1 observation point. Take the first 13 periods of data as the training set for modeling, and the last 5 periods of data as the test set for testing the accuracy of the model settlement prediction value.

[0055] Step 8: Establish an improved grey GM(1,1) model using the training set in Step 7, specifically:

[0056] ① Establish an improved grey GM(1,1) model through the matlab software, and its formula is:

[0057] Take the data in the training set as the original data, and take the square root of each element of the original data to obtain a new sequence:

[0058]

[0059] Then perform the cumulative summation on it to obtain:

[0060]

[0061] Establish a first-order differential equation based on the newly generated data sequence:

[0062]

[0063] Perform mean generation of B and constant term vector Y on the accumulated generated data n , and obtain the matrix composed of a and u

[0064]

[0065] z (1) (k) = 0.5x (1) (k―1) + 0.5x (1) (k), k = 2, 3…n

[0066]

[0067] Substitute into the first-order differential equation to derive the time response formula:

[0068]

[0069] Perform cumulative subtraction reduction on it to obtain the prediction model, and square the predicted result x (0) (t + 1) to restore it to the predicted value of the original data scale

[0070]

[0071] ② Then calculate the residual:

[0072]

[0073] ③ Substitute the calculated residual data into the improved grey GM(1,1) model established in step ①, and repeat the calculation process of step ① to obtain the predicted value of the residual

[0074]

[0075] ④ Perform residual correction to obtain the final predicted value:

[0076]

[0077] Step Nine: Use the improved grey GM(1,1) model established in Step Eight to predict the settlement amount of the P1 observation point determined in Step Seven; and compare it with the actual settlement amount data of the last 5 periods in the test set as Figure 7 shown in Table 1;

[0078] Table 1: Predicted values and actual monitoring data at Observation Point P1

[0079]

[0080] The set threshold is 0.2. As can be seen from Figure 7 Table 1, the errors between the predicted values and the actual monitoring data in the 5 periods are all less than 0.2, indicating that the model established in the embodiment of the present invention has a relatively high prediction accuracy for the settlement amount and meets the requirements.

[0081] In addition, the posterior difference ratio C can be used to analyze the error of the predicted data. The specific formula for the posterior difference ratio is as follows:

[0082]

[0083] S e : standard deviation of residuals, S x : standard deviation;

[0084]

[0085] Through analysis, it can be found that the predicted data has excellent prediction accuracy and can well predict the settlement amount of the goaf, so as to realize the early warning of the goaf settlement and facilitate personnel to take corresponding measures in advance.

[0086] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and retouches can be made, and these improvements and retouches should also be regarded as the protection scope of the present invention.

Claims

1. A method for predicting coal mining area subsidence based on improved GM (1,1) model and SBAS-InSAR, characterized in that: The specific steps are: Step 1: Obtain satellite image data of the target coal mining area; Step 2: Collect 12.5m ALOSDEM data of the target coal mining area in step 1 to remove the terrain phase of the satellite image data in step 1; Step 3: The satellite image data processed in step 2 is used to select a super main image according to the principle of the number of main and secondary image pairs, and a spatiotemporal baseline map and a time baseline map are generated around the super main image; Step 4: Generate a coherence coefficient map based on the space-time baseline map and the time baseline map, perform filtering interference and phase unwrapping processing on the satellite image data, and select valid GCP control points to eliminate orbit errors and slope phases; Step 5: Perform the first surface deformation inversion on the satellite image data processed in step 4 to obtain the preliminary settlement deformation rate of the target coal mining area, and remove the terrain phase, elevation error and orbit error remaining in the interference phase; then perform the second surface deformation inversion based on the settlement deformation rate obtained, calculate the displacement on the time series, and combine the atmospheric high-pass filtering and low-pass filtering methods to further remove the atmospheric phase error; Step 6: For the two inversion results in step 5, the average deformation rate and time series deformation amount corresponding to the target coal mining area are calculated by SVD algorithm, and geocoding is performed to obtain the settlement rate map of the target coal mining area; Step 7: According to the settlement rate map of the target coal mining area, select an observation point in the area with the most obvious settlement of the target coal mining area as the main observation point, extract data of the main observation point in different periods, and use the data at the front of the time sequence as the training set and the rest as the test set; Step 8: Use the training set in step 7 to establish an improved grey GM (1,1) model; Step 9. Use the improved grey GM (1,1) model established in step 8 to predict the settlement of the main observation points determined in step 7; and compare it with the actual settlement data of different periods in the test set. If the error between the predicted value and the actual value is less than the set threshold, retain the current model for subsequent settlement prediction of other observation points; otherwise, continue to repeat steps 8 and 9 until the error between the model predicted value and the actual value is less than the set threshold.

2. The method for predicting coal mining area subsidence based on the improved GM (1,1) model and SBAS-InSAR according to claim 1 is characterized in that: The step seven is specifically as follows: ① The improved grey GM (1,1) model is established by Matlab software, and its formula is: Take the data in the training set as the original data, and take the square root of each element of the original data to get a new sequence: Then add them up to get: Establish a first-order differential equation based on the newly generated data series: Generate the mean of the accumulated data B and the constant term vector Y n , find the matrix composed of a and u With (1) (k)=0.5x (1) (k―1)+0.5x (1) (k),k=2,3…n Will Substitute the first-order differential equation to derive the time response formula: Perform cumulative reduction to obtain the prediction model and convert the prediction result x (0) (t+1) is squared to restore the predicted value of the original data scale ②Then calculate the residual: ③ Substitute the calculated residual data into the improved grey GM (1,1) model established in step ①, and repeat the calculation process in step ① to obtain the predicted value of the residual ④ Perform residual correction to obtain the final prediction value:

Images