A ground surface deformation prediction method, device, medium and product

Abstract

The application discloses a ground surface deformation prediction method and device, medium and product, and relates to the technical field of geological disaster monitoring and surveying. The method comprises the following steps: according to target factors under a research area, outputting a future period cumulative deformation variable by using a trained ground surface deformation prediction model, wherein the target factors are data with an influence degree on the future period cumulative deformation variable greater than an influence degree threshold, the target factors comprise time sequence deformation data of a historical period, topographic data of a future period and environmental data of the future period, and the ground surface deformation prediction model is a model for optimizing weights and bias parameters of a traditional BP neural network by using a particle swarm algorithm and a genetic algorithm. Since the factors such as topography and environment influencing the ground surface deformation are considered, and the defect that the traditional neural network model is prone to falling into local optimization is overcome, the prediction precision of the neural network is improved, and thus high-precision ground surface deformation prediction can be realized.

Description

Technical Field

[0001] The present application relates to the field of geological disaster monitoring and mapping technology, and in particular to a surface deformation prediction method, equipment, medium and product. Background Art

[0002] Its high resolution and all-weather monitoring capabilities make InSAR technology a crucial tool for surface monitoring. While InSAR can effectively identify the distribution of surface hazards, it cannot predict the risk of future geological disasters. However, neural network technology, a key branch of artificial intelligence, is widely used in surface deformation prediction. Combining time-series InSAR technology with neural network prediction techniques can provide early warning and prevention of deformation hazards in mining areas.

[0003] Currently, there are many methods for surface prediction models using InSAR technology combined with neural networks, but there is still a problem of insufficient accuracy in predicting surface deformation. Summary of the Invention

[0004] The purpose of this application is to provide a surface deformation prediction method, equipment, medium and product that can improve the accuracy of surface deformation prediction.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for predicting surface deformation, comprising:

[0007] Obtain the target factors for the study area, where the target factors are data whose impact on the cumulative deformation in the future period is greater than the impact threshold. The target factors include time-series deformation data of the historical period, terrain data of the future period, and environmental data of the future period. The terrain data includes slope, roughness, and undulation, and the environmental data includes vegetation coverage and average annual precipitation.

[0008] According to the target factor, the accumulated deformation amount in the future period is output using a trained surface deformation prediction model, wherein the surface deformation prediction model is an improved BP neural network, which refers to optimizing the weights and bias parameters of the traditional BP neural network using a particle swarm algorithm and a genetic algorithm on the basis of the traditional BP neural network.

[0009] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the surface deformation prediction method described in the first aspect.

[0010] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the surface deformation prediction method described in the first aspect.

[0011] In a fourth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the surface deformation prediction method described in the first aspect.

[0012] According to the specific embodiments provided in this application, this application has the following technical effects:

[0013] The present application provides a surface deformation prediction method, device, medium and product. The method includes outputting the cumulative deformation variable of the future period using a trained surface deformation prediction model based on the target factor of the study area, wherein the target factor is data whose influence on the cumulative deformation variable of the future period is greater than the influence degree threshold. The target factor includes time series deformation data of the historical period, terrain data of the future period and environmental data of the future period. The surface deformation prediction model is a model that optimizes the weights and bias parameters of the traditional BP neural network using a particle swarm algorithm and a genetic algorithm. Since factors such as terrain and environment that affect surface deformation are taken into account, and the disadvantage that the traditional neural network model is prone to falling into local optimality is overcome, the prediction accuracy of the neural network is improved, thereby achieving high-precision surface deformation prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 This is an application environment diagram of a surface deformation prediction method in Example 1 of the present application.

[0016] Figure 2 This is a flow chart of a surface deformation prediction method in Example 1 of the present application.

[0017] Figure 3 Schematic diagram of the InSAR line-of-sight deformation field in Example 1 of the present application.

[0018] Figure 4 This is a comparison chart of four discretization methods of the example detection factor in Example 1 of this application.

[0019] Figure 5 This is a schematic diagram of the results of detecting the influence of a single factor on surface deformation in Example 1 of the present application.

[0020] Figure 6 This is a schematic diagram of the results of the influence of the interaction between detection factors on surface deformation in Example 1 of the present application.

[0021] Figure 7 Schematic diagram of the comparison between the predicted and actual results and the fitting results in Example 1 of the present application.

[0022] Figure 8 A schematic diagram of the structure of a computer device provided in Example 2 of the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0024] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0025] Example 1

[0026] Research has found that in the current surface prediction model method using InSAR technology combined with neural networks, traditional neural networks are prone to falling into local optimal solutions and are sensitive to initial weights and biases. In addition, the prediction model only uses time-series InSAR monitoring data and does not consider the impact of environmental factors such as precipitation and topography on surface deformation.

[0027] To address this issue, this example proposes a surface deformation prediction method that combines InSAR technology with a GA-PSO-BP (particle swarm optimization algorithm, PSO; genetic algorithm, GA) neural network. This method directly optimizes the network's basic parameters, thereby avoiding the overfitting problem caused by redundant convolution kernels in CNNs and better adapting to the high-precision prediction requirements of small-sample coal mine data. This method utilizes geographic detectors to optimize time-series deformation and environmental factor datasets, integrating the global search capabilities and fast convergence of GA and PSO to construct a surface deformation prediction model for mining areas, achieving high-precision surface deformation prediction.

[0028] The surface deformation prediction method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. The terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the target factors under the study area to the server 104. After the server 104 receives the target factors, the server 104 outputs the cumulative deformation amount in the future period based on the target factors using the trained surface deformation prediction model, wherein the surface deformation prediction model is an improved BP neural network, and the improved neural network refers to the optimization of the weights and bias parameters of the traditional BP neural network using the particle swarm algorithm and the genetic algorithm on the basis of the traditional BP neural network. The server 104 can feed back the obtained cumulative deformation amount in the future period to the terminal 102. In addition, in some embodiments, the surface deformation prediction method can also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 can directly use the surface deformation prediction method to process the target factors in the study area, or the server 104 can obtain the target factors in the study area from the data storage system and use the surface deformation prediction method to process them.

[0029] Terminal 102 may include, but is not limited to, various desktop computers, laptops, smartphones, tablet computers, IoT devices, and portable wearable devices. IoT devices may include smart speakers, smart TVs, smart air conditioners, and smart car devices. Portable wearable devices may include smart watches, smart bracelets, and head-mounted devices. Server 104 may be implemented as a standalone server or a server cluster consisting of multiple servers, or may be a cloud server.

[0030] This embodiment provides a method for predicting surface deformation, which is executed by a computer device, specifically a terminal or a server, and can also be executed by a terminal and a server together. In this embodiment, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the process, including the following steps 201 to 202.

[0031] Step 201: Obtain the target factors in the study area, wherein the target factors are data whose influence on the cumulative deformation in the future period is greater than the influence threshold. The target factors include time series deformation data of the historical period, terrain data of the future period, and environmental data of the future period. The terrain data includes slope, roughness, and undulation, and the environmental data includes vegetation coverage and average annual precipitation.

[0032] Step 202: Output the cumulative deformation amount in the future period using a trained surface deformation prediction model based on the target factor, wherein the surface deformation prediction model is an improved BP neural network. The improved neural network refers to optimizing the weights and bias parameters of the traditional BP neural network using a particle swarm algorithm and a genetic algorithm based on the traditional BP neural network.

[0033] The training process of the surface deformation prediction model includes:

[0034] (a) obtaining a sample data set, wherein the sample data set includes a historical sample radar image sequence, a historical sample terrain time series data, and a historical sample environment time series data;

[0035] (b) using remote sensing technology to process the historical sample radar image sequence to obtain InSAR line-of-sight historical time series deformation data;

[0036] (c) using a discretization method to process the InSAR line-of-sight historical time series deformation data, the historical sample terrain time series data, and the historical sample environment time series data to obtain a preprocessed data set;

[0037] (d) using the preprocessed dataset as an independent variable and the historical cumulative deformation variable as a dependent variable, and calculating the contribution intensity, wherein the historical cumulative deformation variable is a variable value calculated based on the historical time series deformation data of the InSAR line of sight, and the contribution intensity includes the contribution intensity of each variable to the dependent variable and the contribution intensity of the interaction between the independent variables to the dependent variable;

[0038] (e) Using the independent variable whose contribution intensity is greater than the contribution intensity threshold as feature data and the historical accumulated deformation variable as label data to train the surface deformation prediction model.

[0039] In order to make the above process of this embodiment more clear to those skilled in the art, it is explained in detail below.

[0040] In view of the shortcomings of the existing models in small sample areas of coal mines in that they cannot quantify the contribution of complex environmental factors and have limited generalization ability under small sample conditions, making it difficult to meet the demand for high-precision prediction in small sample areas, this embodiment provides a surface deformation prediction method that combines InSAR observation data with neural networks based on the data characteristics of small sample coal mine areas. The method includes four parts: solving InSAR line of sight, optimizing surface deformation driving factors, GA-PSO-BP prediction, and analysis and verification. Figure 2 As shown, specifically including:

[0041] (1) Calculating the InSAR line-of-sight deformation field

[0042] S1: Use the SBAS-InSAR (Small Baseline Subset InSAR) method to perform time series processing on InSAR data (in this embodiment, the InSAR data is a Sentinel-1B radar image series) to obtain the InSAR line-of-sight time series deformation field (i.e., InSAR line-of-sight historical time series deformation data). Specifically, the following steps are performed:

[0043] 1) Baseline screening of InSAR data was performed based on pre-set spatiotemporal baseline coherence thresholds to obtain several image pairs that met the spatiotemporal baseline coherence threshold constraints. Specifically, baseline estimation was used to filter the acquired images to select the primary image, selecting July 4, 2020, as the public primary image. The temporal baseline coherence threshold was set to 240 days, and the spatial baseline coherence threshold was set to 45% of the total length.

[0044] 2) Group several image pairs that meet the spatiotemporal baseline coherence threshold constraints to obtain several small baseline sets.

[0045] 3) Perform multi-view processing on the small baseline set to obtain a multi-view processed baseline set. In this embodiment, the multi-view ratio is set to 4:1. The purpose of multi-view is to reduce noise.

[0046] 4) The Goldstein adaptive filtering algorithm is used to perform interference phase filtering on the multi-view processed baseline set to generate a filtered baseline set.

[0047] 5) Use minimum cost flow (MCF) to perform phase unwrapping on the filtered baseline set to generate an unwrapped phase matrix. In this embodiment, the unwrapping coherence threshold is set to 0.25 (only high coherence regions are retained).

[0048] 6) Map the unwrapped phase matrix to the common master image time reference and construct the small baseline subset phase observation equation.

[0049] 7) Perform singular value decomposition (SVD) and reconstruction on the overall observation equation of the phase observation equations of all small baseline subsets, jointly invert the deformation phase components and time series constraints of each subset, and iteratively solve the reconstructed observation equation using the weighted least squares method to obtain the millimeter-level time series deformation field of the radar line of sight (LOS) (i.e., the historical time series deformation data of the InSAR line of sight), as shown in Figure 7. Figure 3 shown.

[0050] (2) Optimizing surface deformation driving factors

[0051] S2: Downsample the InSAR line-of-sight historical time series deformation data described in S1, and calculate the terrain environment factor value at the sampling point (i.e., the value corresponding to the historical sample terrain time series data and the historical sample environment time series data).

[0052] The InSAR line-of-sight historical time series deformation data were downsampled and screened to 5140 points. Considering the influence of terrain and environmental factors on the cumulative surface deformation y (i.e., the historical cumulative deformation), this example selected elevation (X1), slope (X2), aspect (X3), plane curvature (X4), roughness (X5), relief (X6), vegetation cover (X7), average annual precipitation (X8), and average annual surface deformation rate (X9) as well as 10 time series deformations (X10-X19) as driving factors x. Based on these factors, the terrain environmental factor values ​​at the sampling points were calculated.

[0053] The factors for slope (X2), aspect (X3), curvature (X4), roughness (X5), and relief (X6) are all calculated from elevation (X1). Vegetation cover (X7) is calculated using Sentinel-2 imagery. Annual average precipitation data is from the China Meteorological Network. Time series deformation is the historical time series deformation data from the InSAR line of sight as described in S1. Cumulative surface deformation is a variable calculated from the historical time series deformation data from the InSAR line of sight as described in S1. The specific calculation methods for each factor are as follows:

[0054]

[0055]

[0056]

[0057]

[0058]

[0059]

[0060]

[0061] Where, Indicates the altitude (elevation value) of a point, u and v represent the plane coordinates of the point, which are the east-west direction and the north-south direction respectively. and are the gradients of elevation in the east-west and north-south directions, is the elevation value of each pixel in the local area, is the average elevation of the local area, is the total number of pixels in the local area, is the normalized difference vegetation index, is the minimum value of the normalized vegetation index, is the maximum value of the normalized vegetation index, is the near-infrared band, It is the red band.

[0062] S3: Use the sampling points obtained in S2 as basic data for geographic exploration.

[0063] The temporal deformation and topographic environmental factor values ​​described in S2 are used as independent variables, and the accumulated surface deformation is used as the dependent variable to prepare the basic data for driving factor detection.

[0064] Since traditional geographic detectors use manual grading of factor values, which can easily cause data distortion, this embodiment first uses a discretization method to process the above-mentioned independent variables, converting continuous geographic factors (such as slope and elevation) into discrete categories, reducing noise interference and enhancing the model's ability to capture nonlinear relationships. This can 1) reduce the curse of dimensionality: continuous variables under small samples can easily lead to model overfitting, and discretization can compress the feature space; 2) enhance feature separability: coal mine deformation often shows a threshold effect (such as a sudden increase in landslide risk after the slope exceeds a certain value), and discretization can highlight this segmentation law; 3) reduce subjective bias: traditional manual division (such as a fixed threshold) may ignore the local geological characteristics of the coal mine, and automated discretization (such as the natural breakpoint method) can adaptively classify according to data distribution.

[0065] The discretization methods in this embodiment include equal interval method, natural breakpoint method, quantile method and geometric interval method. Figure 4 As shown in Figure 2, four discretization methods are used to process the independent variables X11-X13. The expression corresponding to the equal interval method is:

[0066] ;

[0067] in, Indicates the value of the i-th dividing point, i ranges from 1 to An integer representing the demarcation point number. is the original data set, is the preset number of groups, and are the maximum and minimum values ​​of the data respectively.

[0068] The expression corresponding to the natural breakpoint method is:

[0069] ;

[0070] in, is the within-group variance, Represents the serial number of the group, y belongs to Group A single data value of for The data set of the group, No. The mean of the group data.

[0071] The expression corresponding to the quantile method is:

[0072] ;

[0073] in, Indicates the The value of the dividing point, a is from 1 to An integer representing the demarcation point number. is the original data set, is the preset number of groups, For data No. Quantile.

[0074] The corresponding expression of the geometric interval method is:

[0075] ;

[0076] in, Indicates the value of the bth dividing point, b is from 1 to An integer representing the demarcation point number. is the original data set, is the preset number of groups, and are the maximum and minimum values ​​of the data respectively.

[0077] S4: Use the geographic detector to detect driving factors on the basic data obtained after discretization in S3, obtain the contribution intensity of each driving factor and the interaction between factors to the cumulative surface deformation, and determine the optimal set of surface deformation driving factors based on the contribution intensity.

[0078] Geographic detectors are used to detect the driving effect of independent variables on dependent variables, including the detection of single factor influence and factor interaction, and ultimately to optimize the driving factors. Factor detectors are used to analyze the degree of influence of independent variables on the spatial distribution of dependent variables and identify the driving force of different driving factors on the spatial distribution of dependent variables. The expression is:

[0079] ;

[0080] Where h is the number of layers of the driving factor x, h=1,2,…,L; and are the number of units in the whole area and the variance of y values ​​respectively; q is the driving force measurement value, and its range is [0,1]. The larger the q value is, the stronger the driving force of the driving factor x on the dependent variable y is, and vice versa. The single factor detection results are as follows: Figure 5 shown.

[0081] Interaction detection is used to detect the degree of influence of two different influencing factors on the cumulative deformation of the surface under their joint action. The specific process includes: first, calculating the q values ​​of the two influencing factors X1 and X2 respectively through factor detection, then calculating q (X1∩X2) under the interaction of X1 and X2, and determining their interaction relationship by comparing q (X1), q (X2) and q (X1∩X2) (Table 1). The results of the interaction detection between factors are shown in Figure 1. Figure 6 shown.

[0082] Table 1. Judgment methods and relationships of interaction effects of driving factors

[0083]

[0084] The preferred driving factors in this embodiment include natural characteristic factors X2, X5, X6, X7, X8 and deformation factors X9 to X19, a total of 16 factors, for subsequent prediction.

[0085] (3) Surface deformation prediction (i.e., GA-PSO-BP prediction)

[0086] S5: Take the driving factor set (16 factors) described in S4 as input and the accumulated surface deformation as the label, set a reasonable neural network threshold for GA-PSO-BP, and compare it with traditional GA-BP and PSO-BP to establish the optimal prediction model and predict the accumulated surface deformation.

[0087] The implementation of the GA-PSO-BP neural network involves the following key steps: First, the set of driving factors described in S4 is divided into training and test sets in a 7:3 ratio. This dataset is preprocessed and normalized to ensure input data quality and consistency. Second, the weights and biases of the neural network are initialized to provide a foundation for the optimization process. Next, the parameters of the GA and PSO algorithms, including population size, crossover probability, mutation probability, and inertia weight, are configured to establish the initial settings. Finally, parameters are defined for each layer of the GA and PSO algorithms to enable seamless integration and optimization within the neural network framework. The GA and PSO algorithms are used to optimize the neural network parameters (Table 2).

[0088] Table 2 Parameter configuration of each layer of GA and PSO algorithms

[0089]

[0090] The detailed steps of the optimization process are as follows:

[0091] a. Use GA and PSO algorithms to optimize the weights and bias parameters of the neural network. The optimized weights and biases are expressed by the following formula:

[0092] ;

[0093] ;

[0094] ;

[0095] ;

[0096] Where, Represent the weight matrices from the input layer to the hidden layer and from the hidden layer to the output layer, respectively. Represent the bias vectors of the hidden layer and the output layer respectively, Represents the set of all parameters, n, h, and o represent the number of nodes in the input layer, hidden layer, and output layer, respectively. The reshape function is used to reshape the parameter vector into a two-dimensional shape suitable for the weight matrix.

[0097] b. Calculate the output of the neural network and calculate the error, and use the error function (mean square error) to evaluate the performance of the neural network. The evaluation index can be expressed as:

[0098] ;

[0099] Where: is the fitness function; For the The network output value of the training sample; For the The expected output value of the training samples; is the number of training sets.

[0100] c. Based on the value of the error function, update the GA and PSO populations or particles to gradually optimize the parameters of the neural network. The population with the best adaptability is directly copied into the next generation. For the population with moderate adaptability, two particles are randomly selected and crossover is performed with a crossover probability Pc to select the optimal individual. The population with poor adaptability is mutated with a mutation probability Pm. The speed of particles i and j is crossovered, that is:

[0101] ;

[0102] ;

[0103] Where, and is a random value between [0,1], which is used to control the ratio of speed and position crossover.

[0104] and represents the velocity and position of particle i at iteration step t, and is the corresponding value of another randomly selected particle j, and the new velocity and position generated by the crossover operation of particle i are recorded as and The new velocity and position of particle j generated by the crossover operation are recorded as and The fitness value of the offspring particles after the crossover operation is compared with the fitness value of the parent particles, and the particles with larger fitness value are retained for the next iteration.

[0105] d. Repeat the above steps until the following termination conditions are met: when the number of iterations reaches the preset maximum value, or the algorithm error is lower than a specific threshold.

[0106] (4) Result analysis and verification

[0107] S6: Evaluate the accuracy of the prediction results.

[0108] In order to evaluate the prediction performance of the model, the root mean squared error (RMSE) and mean absolute error (MAE) are used as indicators to evaluate the prediction performance of the model. The root mean squared error and mean absolute error are calculated as follows:

[0109]

[0110] Where: is the total sample size, is the InSAR monitoring value, is the predicted value.

[0111] GA-BP, PSO-BP and GA-PSO-BP were used to predict the results, and the root mean square error (RMSE) and mean absolute error (MAE) of the three prediction results were statistically analyzed. As can be seen from Table 3, the absolute coefficient R2 of the training set and the test set of the three models all reached 0.99, indicating that the three models all showed strong fitting ability. In terms of error indicators, the GA-PSO-BP model performed slightly better than the GA-BP and PSO-BP models in root mean square error (RMSE) and mean absolute error (MAE). The RMSE of the test set was 2.68 mm and the MAE was 2.01 mm, indicating that its prediction accuracy was high. Fitting the predicted value with the InSAR monitoring value shows that ( Figure 7), the absolute coefficient R2 of the two reached 0.99, with a high correlation. Therefore, the GA-PSO-BP prediction model based on time series InSAR technology has good reliability in mining area deformation research.

[0112] Table 3 Comparison of the accuracy of three prediction models

[0113]

[0114] This example proposes a surface deformation prediction method that combines InSAR observation data with a neural network. This method not only obtains a high-precision surface deformation field, but also, after considering the driving factors of surface deformation, constructs a GA-PSO-BP prediction model (i.e., a surface deformation prediction model) that integrates the global search capability of the GA algorithm and the fast convergence of the PSO algorithm.

[0115] (1) In order to address the problem that previous deformation prediction only used historical deformation data to predict future deformation without considering factors such as topography and environment that affect surface deformation, this embodiment uses a geographic detector with optimal parameters after discretization to quantitatively analyze the driving factors of surface deformation in the mining area, optimize the driving factors, and predict surface deformation based on InSAR deformation data and driving factors, thereby improving the scientificity and reliability of the prediction model.

[0116] (2) To address the problem of slow convergence and easy local optima in traditional neural network prediction methods, this embodiment integrates the advantages of the GA algorithm's global search capability and the PSO's fast convergence into a GA-PSO-BP prediction model. This improves the prediction accuracy and convergence speed of the BP neural network, overcomes its disadvantage of being easily trapped in local optima, and constructs a high-precision surface prediction model.

[0117] (3) By comparing and analyzing the prediction effects of GA-BP and PSO-BP algorithms, a surface deformation prediction model with higher accuracy and stronger robustness is constructed, providing key technical support for deformation monitoring and early warning in mining areas.

[0118] Considering the particularity of small sample coal mine area data, this embodiment:

[0119] 1) We address data scarcity by combining geographic detectors with discretization methods for dimensionality reduction, and by using improved neural networks to improve sample efficiency. (Data scarcity: Traditional big data-driven models are prone to overfitting due to limited sample sizes in coal mining areas due to high monitoring costs, insufficient historical records, and the sudden nature of disasters.) Specifically:

[0120] Through geographic detectors, the explanatory power of various factors (such as slope and precipitation) on deformation is quantified, key influencing factors are automatically screened, redundant features are eliminated, and the dimensionality disaster under small samples is reduced.

[0121] Four discretization methods (equal interval / natural breakpoint / quantile / geometric interval) convert continuous variables into categorical variables, compress the feature space, and reduce the model's requirement for sample size.

[0122] Traditional BP networks are prone to overfitting under small samples, while PSO-GA collaborative optimization avoids the gradient descent method's dependence on a large amount of iterative data by intelligently searching weights / biases.

[0123] 2) By integrating multiple target factors and enhancing nonlinear expression through discretization, we address nonlinearity and spatial heterogeneity (nonlinearity: coal mine deformation is affected by multiple factors such as goaf subsidence and groundwater extraction, resulting in highly nonlinear data; spatial heterogeneity: coal mine geological structures vary greatly, and deformation patterns in local areas may vary significantly). Specifically:

[0124] Simultaneously input time series deformation data (historical settlement patterns), terrain data, and environmental data to enhance the model's ability to model complex mechanisms through multimodal data complementarity.

[0125] Discretization transforms continuous nonlinear relationships into piecewise linear relationships, allowing simple neural networks to fit complex patterns.

[0126] 3) Through dynamic encoding of historical deformation data and PSO-GA optimization of anti-disturbance, the problem of temporal instability (temporal instability: mining activities lead to sudden changes in deformation rate, and small samples cannot cover all working conditions) is solved. Specifically:

[0127] Taking historical time series deformation data as input factors, the time dependency is implicitly learned through neural networks, avoiding the strict requirements of traditional time series models on the length of small sample time series.

[0128] The swarm intelligence of PSO and the mutation mechanism of GA enable the model to cover more potential solution space in parameter search and maintain robustness even if there are outliers caused by mutations in mining activities in the sample.

[0129] In summary, this embodiment solves the problem of factor representation distortion under small sample conditions through automated discretization, and uses hybrid intelligent optimization to break through the local optimal limitations of traditional neural networks, ultimately achieving the "low data, high precision" deformation prediction requirements in coal mine scenarios.

[0130] Compared with the prior art, this embodiment has the following advantages:

[0131] Improved prediction accuracy: Through the joint optimization of GA and PSO, the prediction accuracy of the BP neural network was improved, and the root mean square error (RMSE) of the test set was reduced to 2.68 mm and the mean absolute error (MAE) was reduced to 2.01 mm.

[0132] Scientific screening of key factors: The geographic detector method with optimal parameters after discretization is used to quantify driving factors, reduce factor data distortion, and ensure the scientificity and reliability of model input.

[0133] Enhanced model robustness: Through comparative analysis, the GA-PSO-BP model shows better fitting ability and stability than the GA-BP and PSO-BP models.

[0134] High practical application value: The research results verify the effectiveness of the prediction model in surface deformation monitoring and early warning, and provide important technical support for the early prevention of surface disasters.

[0135] This embodiment addresses the unique complexity of surface deformation in coal mining areas, which is affected by the multi-scale coupling of mining disturbances and topographic and environmental factors (such as slope changes affecting stress distribution, precipitation infiltration exacerbating crack development, and vegetation cover reflecting surface stability). Traditional BP neural networks, due to their fully connected structure, are more suitable for processing low-dimensional time-series and environmental feature data. However, they are limited by the risk of overfitting under small sample conditions and the local convergence characteristics of traditional backpropagation algorithms, making it difficult to achieve reliable generalization performance. The surface deformation prediction model utilizes a GA to globally optimize the initial weights of the neural network and combines it with a PSO to dynamically optimize the network bias parameters. This model retains the architectural advantages of the BP neural network in processing low-dimensional time-series and environmental feature data while significantly improving the model's convergence and generalization capabilities under small sample conditions. Furthermore, to address the difficulty of quantifying factor contributions due to the scarcity of mining area data, a geographic detector is used to integrate multiple discretization methods to achieve adaptive optimization of classification thresholds. This effectively overcomes the subjective influence of traditional human reclassification and reduces factor data distortion. This allows for a more accurate representation of the dynamic coupling relationship between key factors such as the topographic environment and accumulated deformation, achieving high-precision surface deformation prediction.

[0136] The surface deformation prediction method proposed in this example, which combines InSAR observation data with a neural network, is highly scientific and applicable. This method not only monitors surface deformation but also builds a prediction model using factors driving surface deformation, such as the terrain environment, thereby achieving a more complete integrated monitoring and prediction model. This method is easy to implement and has broad application in the fields of geological disaster and engineering deformation monitoring and prediction.

[0137] Example 2

[0138] This embodiment provides a computer device, which can be a server or a terminal. Its internal structure diagram can be as follows: Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store data in the surface deformation prediction method. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a surface deformation prediction method in Example 1 is implemented.

[0139] Those skilled in the art will understand that Figure 7 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the above-mentioned method embodiments when executing the computer program.

[0140] Example 3

[0141] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the method for predicting ground surface deformation in embodiment 1 is implemented.

[0142] Example 4

[0143] This embodiment provides a computer program product, including a computer program. When the computer program is executed by a processor, the method for predicting ground surface deformation in embodiment 1 is implemented.

[0144] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0145] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0146] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0147] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0148] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting ground deformation, characterized in that: The surface deformation prediction method comprises: Obtain the target factors for the study area, where the target factors are data whose impact on the cumulative deformation in the future period is greater than the impact threshold. The target factors include time-series deformation data of the historical period, terrain data of the future period, and environmental data of the future period. The terrain data includes slope, roughness, and undulation, and the environmental data includes vegetation coverage and average annual precipitation. Outputting the cumulative deformation amount in the future period using a trained surface deformation prediction model according to the target factor, wherein the surface deformation prediction model is an improved BP neural network, which is based on a traditional BP neural network and uses a particle swarm algorithm and a genetic algorithm to optimize weights and bias parameters of the traditional neural network; The training process of the surface deformation prediction model includes: Acquire a sample data set, wherein the sample data set includes a historical sample radar image sequence, a historical sample terrain time series data, and a historical sample environment time series data; The historical sample radar image sequence is processed using remote sensing technology to obtain InSAR line-of-sight historical time series deformation data; Using a discretization method to process the InSAR line-of-sight historical time series deformation data, the historical sample terrain time series data, and the historical sample environment time series data to obtain a preprocessed data set; The preprocessed data set is used as an independent variable and the historical accumulated deformation variable is used as a dependent variable to calculate the contribution intensity, wherein the historical accumulated deformation variable is a variable value calculated based on the historical time series deformation data of the InSAR line of sight, and the contribution intensity includes the contribution intensity of each variable to the dependent variable and the contribution intensity of the interaction between the independent variables to the dependent variable; Using the independent variable with contribution intensity greater than the contribution intensity threshold as feature data and the historical accumulated deformation variable as label data to train the surface deformation prediction model; The contribution strength is characterized by a driving force value, and the driving force value is calculated as follows: ； in, q is the driving force magnitude; j is the number of strata of the independent variable, j =1,2,…,L; Indicates the independent variable j The number of units in each layer; Indicates the independent variable j The variance of the dependent variable values ​​for each stratum; and are the number of units in the entire region and the variance of the dependent variable values, respectively.

2. The surface deformation prediction method according to claim 1, characterized in that: The discretization methods include equal interval method, natural breakpoint method, quantile method and geometric interval method.

3. The surface deformation prediction method according to claim 1, characterized in that: The calculation formula of the weight is: ； ； in, Represent the weight matrices from the input layer to the hidden layer and from the hidden layer to the output layer respectively; represents the set of all parameters; n, h, and o represent the number of nodes in the input layer, hidden layer, and output layer, respectively; the reshape function represents the function used to organize the parameter vector into a two-dimensional shape of the weight matrix.

4. The surface deformation prediction method according to claim 1, characterized in that: The calculation formula of the bias parameter is: ； ； in, Bias vectors for hidden layer and output layer respectively; Represents the entire parameter set; h and o represent the number of nodes in the hidden layer and output layer respectively.

5. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the surface deformation prediction method according to any one of claims 1 to 4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting ground deformation according to any one of claims 1 to 4 is implemented.

7. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for predicting ground deformation according to any one of claims 1 to 4 is implemented.

Images