A method for evaluating deformation sensitivity of coal mine goaf

Abstract

The present invention relates to the technical field of surface deformation detection and remote sensing geoscience analysis, and discloses a method for evaluating the deformation sensitivity of coal mine goaf, comprising the following steps: S1, establishing a coal mine goaf deformation sensitivity evaluation index library; S2, using SBAS-InSAR technology to obtain the surface deformation rate of the coal mine goaf; S3, uniformly selecting sample points in the study area, using the coal mine goaf deformation sensitivity evaluation index as the input variable, using the corresponding surface deformation rate as a label, constructing a model training sample data set, and randomly grouping them in a ratio of 7:3, as a training set and a test set respectively. The method for evaluating the deformation sensitivity of coal mine goaf, by combining surface deformation detection technology, remote sensing technology and machine learning methods, can make up for the limitations of conventional methods for evaluating the deformation sensitivity of coal mine goaf based on survey statistical data to a certain extent, and realize a more effective and accurate method for evaluating the deformation sensitivity of coal mine goaf.

Description

Technical Field

[0001] The invention relates to the technical field of surface deformation detection and remote sensing geoscience analysis, and in particular to a method for evaluating the deformation sensitivity of a coal mine goaf area. Background Art

[0002] Small baseline set interferometry is a measurement technology that combines short-time and space baselines with interference pairs with relatively small Doppler center frequency differences to identify the phase information of stable high-coherence points on the surface over long time intervals and establish a time series phase model to estimate the surface displacement. It is widely used in surface deformation detection.

[0003] Among the related technologies, the currently commonly used methods for evaluating the deformation sensitivity of coal mine goaf are mostly based on mathematical statistics methods, and usually combine statistics and geographic information technology to carry out sensitivity mapping; for example, multivariate statistical methods, machine learning techniques and regression analysis are used to calculate and map the sensitivity based on survey and statistical results. However, the evaluation results rely on survey and statistical data and have certain limitations. Field investigations are time-consuming and labor-intensive, and it is difficult to conduct a comprehensive assessment of the deformation sensitivity of goaf areas over a large area. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the shortcomings of the prior art, the present invention provides a method for evaluating the deformation sensitivity of coal mine goaf areas. By using SBAS-InSAR technology and remote sensing geoscience analysis methods, the limitations and shortcomings of conventional field surveys and studies on the deformation sensitivity of coal mine goaf areas based on statistical data can be made up, thereby improving the reliability of the deformation sensitivity evaluation of coal mine goaf areas and solving the problems that the evaluation results rely on survey statistical data and have certain limitations; and the field survey is time-consuming and labor-intensive.

[0006] (II) Technical solution

[0007] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for evaluating the deformation sensitivity of a coal mine goaf, comprising the following steps:

[0008] S1. Establish a coal mine goaf area deformation sensitivity evaluation index database;

[0009] S2. Use SBAS-InSAR technology to obtain the surface deformation rate of coal mine goaf areas;

[0010] S3. Sample points are evenly selected in the study area, and the deformation sensitivity evaluation index of coal mine goaf is used as the input variable, and the corresponding surface deformation rate is used as the label to construct the model training sample data set, and they are randomly grouped in a ratio of 7:3 as the training set and test set respectively;

[0011] S4, construct and train a BP neural network model based on genetic algorithm optimization;

[0012] S5. Evaluate the accuracy of the model training results;

[0013] S6. Adjust the model parameters according to the output results of the neural network model, and then repeat S4 and S5. Repeat this process for 5-10 rounds, and select the model corresponding to the optimal result as the final sensitivity evaluation model.

[0014] S7. Use the optimal assessment model to estimate and generate the spatial distribution of deformation sensitivity of coal mine goaf areas.

[0015] Preferably, the establishment of a coal mine goaf deformation sensitivity evaluation index library in S1 comprises the following steps:

[0016] S11. Collect geographical, environmental and geological spatial elements related to the deformation sensitivity of coal mine goaf areas, including: distribution of aquifer medium types, land cover types, land use types, aquifer or sediment thickness distribution, main geological structure distribution, historical earthquake epicenter distribution and magnitude statistics, groundwater level distribution contours, surface water resource distribution, geological lithology distribution, hydraulic conductivity or water storage coefficient distribution, digital elevation distribution factors or similar indicators;

[0017] S12, grading, classifying or normalizing the above indicators, wherein the digital elevation distribution data are respectively extracted from the slope and aspect data; the fault and surface water resource distribution are respectively generated from the distance distribution data, and normalized, and the remaining indicator data are assigned values ​​according to the grading and classification conditions;

[0018] S13. Establish a coal mine goaf deformation sensitivity evaluation index library including aquifer medium type distribution, land cover type, land use type distribution, aquifer or sediment thickness distribution, distance distribution from faults, historical earthquake epicenter distribution and magnitude, groundwater level distribution contours, distance distribution from water systems, geological lithology distribution, hydraulic conductivity or water storage coefficient distribution, elevation distribution, slope distribution, and slope aspect indicators.

[0019] Preferably, the step of obtaining the surface deformation rate of the coal mine goaf area by using SBAS-InSAR technology in S2 comprises the following steps:

[0020] S21, generating a SAR interference pattern set;

[0021] S22. Time series inversion and deformation rate estimation.

[0022] Preferably, generating a SAR interferogram set in S21 includes first determining a suitable reference image by means of a spatiotemporal threshold and coherence evaluation method, then registering all SAR images to the reference image in combination with precise orbit data, then simulating a reference image intensity map and an inverted terrain phase in combination with external DEM data, performing interference processing of the SAR image and removing the flat ground phase; and finally performing phase unwrapping of the interferogram using a minimum cost flow method to obtain an unwrapped interferogram set.

[0023] Preferably, the time series inversion and deformation rate estimation in S22 include networking the interference pattern using the coherence threshold and the spanning tree rule, inverting the surface time series using the weighted least squares method, estimating the atmospheric delay using the ERA-5 atmospheric model parameters and removing it, then estimating and removing the DEM residual and noise using the wavelet filtering method, generating a surface deformation time series, and finally estimating the surface deformation rate using the least squares method.

[0024] Preferably, constructing the model training sample data set in S3 includes the following steps:

[0025] S31. Generate 2000 evenly distributed sample points in the study area, and use spatial overlay analysis technology to obtain the corresponding index values ​​and surface deformation rates of the points;

[0026] S32. Randomly divide all sample points into training group and test group in a ratio of 7:3.

[0027] Preferably, constructing and training a BP neural network model based on genetic algorithm optimization in S4 comprises the following steps:

[0028] S41, create BP neural network;

[0029] S42, determining the initial weight value and threshold of the network, and encoding them to obtain an initial population;

[0030] S43, training the BP neural network error as a fitness value;

[0031] S44, performing genetic algorithm operations, i.e., selection, crossover, and mutation, to obtain a new population;

[0032] S44, iteratively calculating fitness to obtain an optimal weight threshold;

[0033] S45. Iteratively update the weight threshold until the constraint condition is met and output the result.

[0034] Preferably, in S5, the accuracy of the model training result is evaluated, including by calculating the error between the predicted value and the true value.

[0035] Preferably, the optimal assessment model is used in S7 to estimate and generate the spatial distribution of deformation sensitivity of coal mine goaf areas, including performing pixel-by-pixel inversion based on the optimal model obtained in S6 to obtain the quantitative results of the surface deformation sensitivity assessment of the study area, and dividing the risks into low, medium, high and extremely high according to the quartiles of the assessment result values ​​from large to small to generate the spatial distribution of deformation sensitivity of coal mine goaf areas.

[0036] (III) Beneficial effects

[0037] Compared with the prior art, the present invention provides a method for evaluating the deformation sensitivity of coal mine goaf, which has the following beneficial effects:

[0038] 1. The present invention combines surface deformation detection technology, remote sensing technology and machine learning methods to make up for the limitations of conventional coal mine goaf deformation sensitivity evaluation methods based on survey statistical data to a certain extent, and realizes a more effective and accurate coal mine goaf deformation sensitivity evaluation method.

[0039] 2. The present invention selects more than ten indicators related to the deformation sensitivity of coal mine goaf, covering geographical, environmental and geological factors, and the evaluation results are more scientific and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic flow chart of a first embodiment of a method for evaluating deformation sensitivity of a coal mine goaf area proposed by the present invention;

[0041] Figure 2 The surface deformation rate distribution map of the study area of ​​the second embodiment of the method for evaluating deformation sensitivity of coal mine goaf area proposed by the present invention;

[0042] Figure 3 The sample points of the second embodiment of the method for evaluating the deformation sensitivity of coal mine goaf area proposed by the present invention are distributed in the schematic diagram of the model structure;

[0043] Figure 4 This is a schematic diagram of the surface deformation sensitivity distribution of the goaf area in the second embodiment of the method for evaluating the deformation sensitivity of the goaf area in a coal mine proposed by the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Embodiment 1:

[0046] See attached Figure 1 As shown, a method for evaluating deformation sensitivity of coal mine goaf includes the following steps:

[0047] S1. Establish a coal mine goaf area deformation sensitivity evaluation index database;

[0048] The step S1 of establishing a coal mine goaf area deformation sensitivity evaluation index library includes the following steps:

[0049] S11. Collect geographical, environmental and geological spatial elements related to the deformation sensitivity of coal mine goaf, including the following 10 commonly used factors or similar indicators:

[0050] Distribution of aquifer medium types (e.g. clay, sand, gravel, etc.);

[0051] Distribution of land cover types / land use types;

[0052] Aquifer or sediment thickness distribution;

[0053] Distribution of major geological structures (fractures, faults, etc.);

[0054] Statistics such as the epicenter distribution and magnitude of historical earthquakes;

[0055] Groundwater level distribution contour lines;

[0056] Distribution of surface water resources (rivers, lakes, canals, etc.);

[0057] Geological lithology distribution (geological map);

[0058] Hydraulic conductivity or water storage coefficient distribution;

[0059] Digital elevation map (DEM);

[0060] S12. The above indicators are graded, classified or normalized, wherein the slope and aspect data are extracted from the DEM data; the distance distribution data (i.e., the distance distribution from any position on the surface to the fracture and water system) are generated for the fracture and surface water resource distribution, and normalized; the remaining indicators, such as land use type, aquifer medium type and other data are assigned values ​​according to the classification and classification, such as assigning the construction land type in the land use type to 1, and assigning the clay in the aquifer medium to 5, etc.;

[0061] Finally, a coal mine goaf deformation sensitivity evaluation index library containing the following 12 indicators was established:

[0062] Aquifer medium type distribution (1-5);

[0063] Land cover type / land use type distribution (1-5);

[0064] Aquifer or sediment thickness distribution (1-5);

[0065] Distribution of distance from the fracture (0-1);

[0066] Distribution of historical earthquake epicenters and magnitudes (1-5);

[0067] Groundwater level distribution contour (m);

[0068] The distribution of distance to the water system (0-1);

[0069] Geological lithology distribution (1-5);

[0070] Hydraulic conductivity or water storage coefficient distribution (1-5);

[0071] Elevation distribution (1-5);

[0072] Slope distribution (0-1);

[0073] Slope aspect (1-5);

[0074] S2. Use SBAS-InSAR technology to obtain the surface deformation rate of coal mine goaf areas;

[0075] In S2, the SBAS-InSAR technology is used to obtain the surface deformation rate of the coal mine goaf area, including the following steps:

[0076] S21, generating a SAR interference pattern set;

[0077] S22. Time series inversion and deformation rate estimation.

[0078] The SAR interferogram set is generated in S21, including first determining a suitable reference image by means of a spatiotemporal threshold and a coherence evaluation method, then registering all SAR images to the reference image in combination with precise orbit data, then simulating a reference image intensity map and an inverted terrain phase in combination with external DEM data, performing interferometric processing of the SAR image and removing the flat ground phase; and finally performing phase unwrapping of the interferogram using a minimum cost flow method to obtain an unwrapped interferogram set.

[0079] The time series inversion and deformation rate estimation in S22 includes networking the interference graph using the coherence threshold and the spanning tree rule, inverting the surface time series using the weighted least squares method, estimating the atmospheric delay using the ERA-5 atmospheric model parameters and removing it, then estimating and removing the DEM residual and noise using the wavelet filtering method, generating the surface deformation time series, and finally estimating the surface deformation rate using the least squares method;

[0080] The SAR images are registered, baseline estimated, interferometrically processed and phase unwrapped to obtain a set of SAR interferograms.

[0081] The interference patterns are networked using the coherence threshold and minimum spanning tree rule, and the surface time series are inverted using the weighted least squares method.

[0082] The atmospheric delay is estimated and removed using the ERA-5 atmospheric model parameters;

[0083] Wavelet filtering method is used to estimate and remove DEM residuals and noise to generate a time series of surface deformation;

[0084] The surface deformation rate was estimated using the least squares method.

[0085] S3. Sample points are evenly selected in the study area, and the deformation sensitivity evaluation index of coal mine goaf is used as the input variable, and the corresponding surface deformation rate is used as the label to construct the model training sample data set, and they are randomly grouped in a ratio of 7:3 as the training set and test set respectively;

[0086] Constructing a model training sample data set in S3 includes the following steps:

[0087] S31. Generate 2000 evenly distributed sample points in the study area, and use spatial overlay analysis technology to obtain the corresponding index values ​​and surface deformation rates of the points;

[0088] S32, randomly divide all sample points into a training group and a test group in a ratio of 7:3;

[0089] S4, construct and train a BP neural network model based on genetic algorithm optimization;

[0090] The BP neural network model optimized based on genetic algorithm is constructed and trained in S4, including the following steps:

[0091] S41, create BP neural network;

[0092] S42, determining the initial weight value and threshold of the network, and encoding them to obtain an initial population;

[0093] S43, training the BP neural network error as a fitness value;

[0094] S44, performing genetic algorithm operations, i.e., selection, crossover, and mutation, to obtain a new population;

[0095] S44, iteratively calculating fitness to obtain an optimal weight threshold;

[0096] S45, iteratively updating the weight threshold until the constraint condition is met and outputting the result;

[0097] S5. Evaluate the accuracy of the model training results;

[0098] Initially create a five-layer BP neural network consisting of an input layer, three hidden layers, and an output layer. The number of input layer nodes is set to 12 (consistent with the number of input indicators), and the number of output layer nodes is set to 1. The number of hidden layer nodes needs to be determined through iteration, that is, a loop is used to traverse the hidden layer nodes and training errors within the range. In addition, other parameters need to be set, such as learning rate, number of training times, and target error of training.

[0099] Initialize the genetic algorithm (GA) parameters, set the initial population size, maximum evolutionary generations, crossover probability, mutation probability, etc. The empirical range of weights and thresholds is (-1, 1); then, encode the variable matrix to obtain the initial population, and use the BP neural network error (such as MAE) as the fitness value; perform selection, crossover, and mutation to obtain a new population; iteratively calculate the fitness to determine the optimal weights and thresholds, and determine the corresponding number of hidden layer nodes.

[0100] The corresponding parameters of the optimal solution obtained by GA are passed to the weight matrix and threshold vector of the BP neural network for training, and the training results are output.

[0101] In S5, the accuracy of the model training result is evaluated, including calculating the error between the predicted value and the true value;

[0102] Assume that the predicted value is The true value is y={y1,y2,y3,K,y n}, the calculation formulas of MAE and RMSE are as follows:

[0103]

[0104]

[0105] The calculation results of the two are in the range of [0, +∞). When the predicted value is completely consistent with the true value, it is equal to 0, that is, a perfect model; the larger the error, the larger the value;

[0106] S6. According to the output results of the neural network model, adjust the model parameters (such as the number of hidden layers), and then repeat S4 and S5. Repeat this process for 5-10 rounds, and select the model corresponding to the optimal result as the final sensitivity evaluation model.

[0107] S7. Estimate and generate the spatial distribution of deformation sensitivity of coal mine goaf areas using the optimal assessment model;

[0108] In S7, the optimal assessment model is used to estimate and generate the spatial distribution of deformation sensitivity of coal mine goaf areas, including pixel-by-pixel inversion based on the optimal model obtained in S6 to obtain the quantitative results of the surface deformation sensitivity assessment of the study area, and the risk is divided into low, medium, high and extremely high according to the quartiles of the assessment result values ​​from large to small to generate the spatial distribution of deformation sensitivity of coal mine goaf areas.

[0109] Embodiment 2: Based on embodiment 1, the difference is that;

[0110] See attached Figure 2-3 As shown, an implementation case of a method for evaluating deformation sensitivity of a coal mine goaf in Example 1 is used;

[0111] S1. Collect multispectral remote sensing images such as Sentinel-2, Landsat 8, and Landsat 9 in recent years in Luozhuang District, Linyi City, Shandong Province, Sentinel-1SAR SLC data products, groundwater level contours, hydrogeological data, land use, vegetation coverage, water bodies, and DEM and other raster data products, etc., to build a deformation sensitivity evaluation index database for coal mine goaf areas in Linyi City;

[0112] S2. Using the SBAS-InSAR technical process in S2 of Example 1, obtain high-precision time series deformation measurement results in the LOS (line-of-sight) direction covering a long time series (2016.10 to 2021.12) of the goaf area of ​​the coal mine in Linyi City;

[0113] S3, 2214 sample points were uniformly generated in the study area, and the index values ​​and deformation rates at the corresponding points were obtained as input variables and training targets respectively by using the spatial overlay analysis method, and the training set (1550 sample points) and the test set (664 sample points) were randomly divided according to the ratio of 7:3;

[0114] S4. Construct a neural network model with an input layer, three hidden layers and an output layer. The number of training times is 1000, the learning rate is 0.01, and the minimum error of the training target is 0.0001. The initial population size of GA is 30, the maximum evolutionary generation is 50, the crossover probability is 0.8, the mutation probability is 0.2, and the fitness function is set to the minimum value of MAE, that is, F = min(MAE Trainingset,Testingset );

[0115] The optimal solution variables are output through GA, and then the variables are assigned to the corresponding positions of the weight matrix and threshold vector of the BP neural network, and the optimized BP training and testing are carried out; after iteration, it is determined that the optimal training result is obtained when the number of hidden layer nodes in the BP neural network is 13;

[0116] The MAE and RMSE corresponding to the best training results in S5 and S4 are 2.9213 and 3.6836 respectively;

[0117] S6. The number of hidden layers in the BP neural network was adjusted to 1, 2, 4, 5, and 6 respectively. After 5 rounds of training, 6 optimization models were obtained. By comparison, it was found that the BP neural network model optimized by GA with 5 hidden layers was the optimal model for the sensitivity evaluation of surface stability of coal mine goaf in the study area in this example.

[0118] S7. Using the optimal model determined in S6, conduct surface stability sensitivity evaluation and mapping of the coal mine goaf area in the study area.

[0119] It should be noted that the term "comprises" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, article, or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article, or device that includes the element.

[0120] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating deformation sensitivity of coal mine goaf, characterized in that: The following steps are involved: S1. Establish a coal mine goaf area deformation sensitivity evaluation index database; The step S1 of establishing a coal mine goaf area deformation sensitivity evaluation index library includes the following steps: S11. Collect geographical, environmental and geological spatial elements related to deformation sensitivity of coal mine goaf areas, including: distribution of aquifer medium types, land cover types, land use types, aquifer or sediment thickness distribution, major geological structure distribution, historical earthquake epicenter distribution and magnitude statistics, groundwater level distribution contours, surface water resource distribution, geological lithology distribution, hydraulic conductivity or water storage coefficient distribution, and digital elevation distribution factor indicators; S12, grading, classifying or normalizing the above indicators, wherein the digital elevation distribution data are respectively extracted from the slope and aspect data; the fault and surface water resource distribution are respectively generated from the distance distribution data, and normalized, and the remaining indicator data are assigned values ​​according to the grading and classification conditions; S13. Establish a coal mine goaf deformation sensitivity evaluation index library including aquifer medium type distribution, land cover type, land use type distribution, aquifer or sediment thickness distribution, distance distribution from faults, historical earthquake epicenter distribution and magnitude, groundwater level distribution contours, distance distribution from water systems, geological lithology distribution, water conductivity or water storage coefficient distribution, elevation distribution, slope distribution, and slope aspect indicators; S2. Use SBAS-InSAR technology to obtain the surface deformation rate of coal mine goaf areas, wherein the SAR images are registered, baseline estimated, interferometrically processed, and phase unwrapped to obtain a set of SAR interferograms; the interferograms are networked using the coherence threshold and minimum spanning tree rule, and the surface time series are inverted using the weighted least squares method; the atmospheric delay is estimated and removed using the ERA-5 atmospheric model parameters; the DEM residual and noise are estimated and removed using the wavelet filtering method to generate a surface deformation time series; the surface deformation rate is estimated using the least squares method; S3. Sample points are evenly selected in the study area, and the deformation sensitivity evaluation index of coal mine goaf is used as the input variable, and the corresponding surface deformation rate is used as the label to construct the model training sample data set, and they are randomly grouped in a ratio of 7:3 as the training set and test set respectively; Constructing a model training sample data set in S3 includes the following steps: S31. Generate 2000 evenly distributed sample points in the study area, and use spatial overlay analysis technology to obtain the corresponding index values ​​and surface deformation rates of the points; S32, randomly divide all sample points into a training group and a test group in a ratio of 7:3; S4, construct and train a BP neural network model based on genetic algorithm optimization; The BP neural network model optimized based on genetic algorithm is constructed and trained in S4, including the following steps: S41, create BP neural network; S42, determining the initial weight value and threshold of the network, and encoding them to obtain an initial population; S43, training the BP neural network error as a fitness value; S44, performing genetic algorithm operations, i.e., selection, crossover, and mutation, to obtain a new population; S44, iteratively calculating fitness to obtain an optimal weight threshold; S45, iteratively updating the weight threshold until the constraint condition is met and outputting the result; S5. Evaluate the accuracy of the model training results; Initially, a five-layer BP neural network consisting of an input layer, three hidden layers and an output layer is created. The number of input layer nodes is set to 12, and the number of output layer nodes is set to 1. The number of hidden layer nodes is determined by iteration, that is, a loop is used to traverse the hidden layer nodes and training errors within the range. Initialize the genetic algorithm parameters, set the initial population size, maximum evolutionary generations, crossover probability, mutation probability, and the empirical range of weights and thresholds to (-1, 1); then, encode the variable matrix to obtain the initial population, and use the BP neural network error as the fitness value; perform selection, crossover, and mutation to obtain a new population; iteratively calculate the fitness to determine the optimal weights and thresholds, and determine the corresponding number of hidden layer nodes; The corresponding parameters of the optimal solution obtained by GA are passed to the weight matrix and threshold vector of the BP neural network for training, and the training results are output; S6. Adjust the model parameters according to the output results of the neural network model, and then repeat S4 and S5. Repeat this process for 5-10 rounds, and select the model corresponding to the optimal result as the final sensitivity evaluation model. S7. Estimate and generate the spatial distribution of deformation sensitivity of coal mine goaf areas using the optimal assessment model; In S7, the optimal assessment model is used to estimate and generate the spatial distribution of deformation sensitivity of coal mine goaf areas, including pixel-by-pixel inversion based on the optimal model obtained in S6 to obtain the quantitative results of the surface deformation sensitivity assessment of the study area, and the risk is divided into low, medium, high and extremely high according to the quartiles of the assessment result values ​​from large to small to generate the spatial distribution of deformation sensitivity of coal mine goaf areas.

Images