SOA-KELM-based method and system for predicting height of water-conducting fissure zone of coal seam roof of coal mine

Abstract

The invention discloses an SOA-KELM-based height prediction method and system for a water-conducting fracture zone of a coal seam roof of a coal mine, and the method comprises the steps: initializing the parameters of an SOA algorithm, and setting a fitness function as a root-mean-square error between the expected output and actual output of a training sample of a KELM model; optimizing the regularization coefficient and kernel function parameters of the KELM model based on an SOA algorithm to obtain an optimal parameter combination; according to the optimal parameter combination, an SOA-KELM prediction model is established; and evaluating and verifying the performance of the SOA-KELM prediction model by adopting a test set sample, and outputting the height of the coal mine coal seam roof water-conducting fracture zone. According to the method, automatic optimization of the key parameters of the KELM model is realized in combination with the SOA algorithm, the performance defect caused by manual presetting of the key parameters of the KELM model is avoided, and the model prediction precision is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of coal mine coal seam roof water-conducting fracture zone height prediction, and in particular relates to a coal mine coal seam roof water-conducting fracture zone height prediction method and system based on SOA-KELM. Background Art

[0002] After coal seam mining, overburden stress redistributes, causing deformation, movement, and damage to the surrounding rock in the goaf. Once the resulting water-conducting fracture zones connect to the overlying aquifer or surface water, they become pathways for water from the overlying aquifer or surface water to enter the coal mining face or goaf, causing water inrush accidents from the coal seam roof, seriously threatening mine safety. To sever the hydraulic connection between the water body and the working face, a thick, waterproof safety coal pillar is typically left between the water body and the water-conducting fracture zone during coal mining beneath water. Accurately predicting the height of the water-conducting fracture zone under varying overburden conditions is crucial for preventing and controlling mine roof water hazards and ensuring safe coal mining. It is also an effective way to reduce the thickness of the waterproof coal pillar, increase the upper limit of safe mining, and free up idle coal resources. Currently, there are four main methods for determining the height of water-conducting fracture zones: calculation, including theoretical calculations and empirical formulas; physical similarity simulation or numerical simulation; field measurement, primarily employing geophysical exploration techniques such as borehole flushing fluid, borehole sound velocity, borehole ultrasonic imaging, color borehole television systems, underground water-conducting fracture zone height observation instruments, and magnetotelluric methods; and prediction based on comprehensive multi-factor analysis. These methods have played an important role in predicting the height of water-conducting fracture zones and ensuring safe coal mining under water.

[0003] However, the existing technology still has many shortcomings in practical applications. Although the empirical formula calculation is simple, the influencing factors considered are single, and it is difficult to reflect the comprehensive effects of multiple influencing factors. In addition, the engineering geological and hydrogeological conditions of each mine are different and complex, which makes it difficult to ensure the accuracy of the calculation. Although physical similarity simulation tests and numerical simulation tests can provide a certain reference, there is a deviation between the simulation conditions and the actual working conditions, and it is difficult to fully and accurately reflect the actual situation. Although the field measurement method has reliable data and high accuracy, it is cumbersome to operate, has a large workload, and is costly. It is also limited by field conditions and is difficult to promote and apply on a large scale. In addition, the height of the water-conducting fracture zone is affected by many factors such as the occurrence conditions of the ore layer, the geological background and the mining process. It has the characteristics of complexity, difficulty in quantification, and nonlinearity. It is currently difficult to describe it with an accurate mathematical model, which further limits the application effect of the existing technology in actual engineering. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes a method and system for predicting the height of water-conducting fracture zones in coal seam roofs of coal mines based on SOA-KELM to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for predicting the height of water-conducting fracture zones in coal seam roofs of coal mines based on SOA-KELM, comprising:

[0006] Determine the main controlling factors affecting the height of the water-conducting fracture zone in the coal seam roof and collect the corresponding measured data;

[0007] Divide the collected measured data into training set samples and test set samples, and standardize the data;

[0008] Initialize the parameters of the Seagull Optimization Algorithm (SOA) and set the fitness function to the root mean square error between the expected output and the actual output of the training sample of the kernel extreme learning machine (KELM) model.

[0009] Based on the SOA algorithm, the regularization coefficient and kernel function parameters of the KELM model are optimized to obtain the optimal parameter combination;

[0010] Establishing a SOA-KELM prediction model based on the optimal parameter combination;

[0011] The test set samples were used to evaluate and verify the performance of the SOA-KELM prediction model, and the height of the water-conducting fracture zone in the coal seam roof of the coal mine was output.

[0012] Preferably, the main control factors include the selection of coal seam roof type, mining depth, coal seam inclination, mining thickness, and working face inclined length.

[0013] Preferably, the parameters of the initialization seagull optimization algorithm SOA include setting the number of seagull populations, the maximum number of iterations, the optimization parameter dimensions, and the upper and lower bounds of the parameters.

[0014] Preferably, the formula of the fitness value function is:

[0015]

[0016] Where: y i is the actual value; is the predicted value.

[0017] Preferably, the process of optimizing the regularization coefficient and kernel function parameters of the KELM model includes:

[0018] Calculate the initial seagull fitness value and sort them;

[0019] Update the seagull position and recalculate the fitness value;

[0020] Determine whether the maximum number of iterations of the algorithm has been reached or whether the optimal fitness has been reached. If so, output the current optimal parameter combination. If not, continue to update the seagull position and recalculate the fitness value.

[0021] Preferably, the updating of the seagull position includes updating the position based on the seagull's migration behavior and foraging behavior.

[0022] Preferably, when updating the position based on the migratory behavior of the seagull, the seagull is moved to the current optimal position by controlling the variables.

[0023] Preferably, when updating the position based on the foraging behavior of the seagull, the seagull continuously attacks the prey in a spiral descending arrangement, and updates the position according to the corresponding motion trajectory formula.

[0024] Preferably, the process of establishing the SOA-KELM prediction model includes:

[0025] The optimal parameter combination is substituted into the KELM model to construct a model for predicting the height of the water-conducting fracture zone in the coal seam roof.

[0026] In a second aspect, the present invention discloses a coal mine coal seam roof water-conducting fracture zone height prediction system based on SOA-KELM, which is used to implement the method described in the first aspect, comprising:

[0027] The data acquisition module is used to determine the main controlling factors affecting the height of the water-conducting fracture zone in the coal seam roof and collect the corresponding measured data;

[0028] The data processing module is used to divide the collected measured data into training set samples and test set samples, and standardize the data;

[0029] The parameter setting module is used to initialize the parameters of the Seagull Optimization Algorithm (SOA) and set the fitness function to the root mean square error between the expected output and the actual output of the training sample of the kernel extreme learning machine (KELM) model.

[0030] Parameter optimization module, which is used to optimize the regularization coefficient and kernel function parameters of the KELM model based on the SOA algorithm to obtain the optimal parameter combination;

[0031] A model building module, used to establish a SOA-KELM prediction model based on the optimal parameter combination;

[0032] The result prediction module is used to evaluate and verify the performance of the SOA-KELM prediction model using test set samples and output the height of the water-conducting fracture zone in the coal seam roof of the coal mine.

[0033] Compared with the prior art, the present invention has the following advantages and technical effects:

[0034] The present invention provides a method for predicting the height of the water-conducting fracture zone in the coal seam roof of a coal mine based on SOA-KELM. First, the main control factors affecting the height of the water-conducting fracture zone in the coal seam roof are determined, and corresponding measured data are collected; secondly, the collected measured data are divided into training set samples and test set samples, and the data are standardized; then, the parameters of the Seagull Optimization Algorithm (SOA) are initialized, and the fitness function is set as the root mean square error between the expected output and the actual output of the training sample of the kernel extreme learning machine (KELM) model; further, the regularization coefficient and kernel function parameters of the KELM model are optimized based on the SOA algorithm to obtain the optimal parameter combination; thirdly, a SOA-KELM prediction model is established according to the optimal parameter combination; finally, the performance of the SOA-KELM prediction model is evaluated and verified using the test set samples, and the height of the water-conducting fracture zone in the coal seam roof of the coal mine is output.

[0035] This invention, combined with the SOA algorithm, automatically optimizes the key parameters of the KELM model, avoiding performance drawbacks caused by artificially presetting key parameters and effectively improving the model's prediction accuracy. The invention is practical and its accuracy meets the requirements for predicting the height of water-conducting fracture zones in modern mine engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0037] Figure 1 This is a model convergence curve diagram of an embodiment of the present invention;

[0038] Figure 2 Schematic diagram of comparison results between actual values and predicted values of training samples according to an embodiment of the present invention;

[0039] Figure 3 Schematic diagram of the prediction results of the KELM model and the support vector regression model (SVR) for training samples according to an embodiment of the present invention;

[0040] Figure 4 Schematic diagram of prediction results of three models for training samples according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0042] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0043] Example 1

[0044] S1. Determine the main controlling factors affecting the height of the water-conducting fracture zone in the coal seam roof and collect the corresponding measured data;

[0045] Furthermore, the main control factors include the selection of coal seam roof type, mining depth, coal seam inclination, mining thickness, and working face inclined length.

[0046] Specifically, the main controlling factors affecting the development height of the coal seam roof fracture zone were determined. Five indicators, namely, coal seam roof type X1, mining depth X2, coal seam inclination X3, mining thickness X4, and working face oblique length X5, were selected as the main controlling factors affecting the development height (Y) of the coal seam roof fracture zone. Measured data on the height of the water-conducting fracture zone in the coal seam roof of various mining areas in my country were collected, with a total of 53 sets of measured data, as shown in Table 1.

[0047] Table 1

[0048]

[0049]

[0050] S2. Divide the collected measured data into training set samples and test set samples, and standardize the data;

[0051] Specifically, the training set samples and test set samples are randomly divided into 8:2 ratios, where groups 1 to 42 are training set samples and groups 43 to 53 are test set samples, and the training set and test set data are standardized.

[0052] Standardization processing: Based on the research data information, establish the prediction indicator matrix X:

[0053] X=(x ij ) n*p

[0054] Among them, n is the number of measured sample groups of water-conducting fracture zone height; p is the number of main controlling factors affecting the development height of water-conducting fracture zone.

[0055] In order to eliminate the influence of different dimensions of each indicator and the errors caused by the variation of each indicator or the large difference in values, it is necessary to standardize the original data of each indicator. [7] , the standardization formula is shown in Eq.

[0056]

[0057] in, is the average value of the jth main control factor; is the standard deviation of the jth main control factor.

[0058] S3. Initialize the parameters of the Seagull Optimization Algorithm (SOA) and set the fitness function to the root mean square error between the expected output and the actual output of the training sample of the kernel extreme learning machine (KELM) model.

[0059] Furthermore, the parameters of the initialization seagull optimization algorithm SOA include setting the number of seagull populations, the maximum number of iterations, the optimization parameter dimensions, and the upper and lower bounds of the parameters.

[0060] Specifically, given that the Seagull Optimization Algorithm (SOA) has the advantages of simple principle, low computational complexity, strong global search capability, and fast iterative convergence speed, it has obvious advantages in solving optimal problems. The Seagull Algorithm (SOA) is used to optimize the key parameters of KELM, realize automatic optimization of the key parameters of KELM, avoid performance defects caused by artificial presets, and effectively improve the prediction accuracy of the model.

[0061] The Extreme Learning Machine (ELM) is a single hidden layer feedforward neural network (SLFN) structure. Compared to traditional learning algorithms that iteratively seek (ω, b, β) by descending the gradient, the SLFN randomly generates input weights w and hidden layer biases b. When w and b are randomly determined, the hidden layer output matrix H is also uniquely determined. That is, the output layer weight β can be expressed as β = H + T, where H + is the generalized inverse matrix of H. In fact, ELM learning is to find the least squares solution of the minimum norm of the linear system Hβ=T.

[0062] Suppose there are Q arbitrary samples (X i ,T i ) T (i=1,2,…,Q), where there are n input variables and m output variables, i.e., sample input X i =[x i1 ,x i2 ,…,x in ] T ∈R n , expected output T i =[t i1 ,t i2 ,…,t in ] T ∈R m .

[0063] The kernel extreme learning machine (KELM) replaces the random mapping of the hidden layer by introducing a stable kernel mapping. According to the Karush-Kuhn-Tucker (KKT) theory, the regularization coefficient C and the diagonal matrix I are introduced to solve the least squares solution of Hβ=T, and the result is:

[0064]

[0065] Then the approximation function learned by KELM can be expressed as:

[0066]

[0067] Where T is the output expectation, x is the input vector. Define the kernel function matrix Ω ELM =HH T , by the element Ω ELMi,j =h(x i )·h(x j )=K(x i ,x j ), where K(x i ,x j ) is the kernel function, x i and x j is the input vector for training, then the approximation function learned by KELM can be written as:

[0068]

[0069] K(x i ,x j ) is the kernel function. To improve the prediction effect, the radial basis kernel function with strong generalization ability is selected:

[0070]

[0071] Where: σ is the kernel function parameter.

[0072] The KELM model uses a kernel function to map the input data samples of the influencing factors of the height of the nonlinear coal seam roof water-conducting fracture zone into a high-dimensional space, and converts the original input data into a more discriminative feature representation to better capture the nonlinear relationship.

[0073] In this embodiment, the factors affecting the height of the water-conducting fracture zone in the coal seam roof are complex and nonlinear. Based on the theoretical basis of kernel functions, the kernel extreme learning machine (KELM) is used as the baseline prediction model. The reference of the kernel function makes KELM more powerful than ELM in function approximation, but increases the sensitivity to parameter selection. When using KELM to predict the height of the water-conducting fracture zone in the coal seam roof, the large-scale input data increases the computational complexity of the model. In order to ensure the stability of the algorithm and the prediction accuracy, it is necessary to set an appropriate regularization coefficient C and kernel function parameter σ. Since the regularization coefficient C and kernel function parameter σ in KELM are generally preset manually, the prediction effect is completely dependent on the experience of the staff, which limits the application effect of the prediction model in actual engineering sites.

[0074] S4. Based on the SOA algorithm, the regularization coefficient and kernel function parameters of the KELM model are optimized to obtain the optimal parameter combination;

[0075] Furthermore, the process of optimizing the regularization coefficient and kernel function parameters of the KELM model includes:

[0076] Calculate the initial seagull fitness value and sort them;

[0077] Specifically, the Seagull Algorithm requires initializing a flock of seagulls during optimization, i.e., a set of initial solutions. Each solution corresponds to a set of KELM hyperparameters (regularization parameter and kernel parameter). The fitness value is used to evaluate the performance of the KELM model corresponding to each set of hyperparameters, using the mean squared error (MSE) of the validation set.

[0078]

[0079] Where: y i is the actual value; is the predicted value.

[0080] The specific process includes:

[0081] (1) Initialize the seagull population: randomly generate N seagull individuals, each of which represents a set of hyperparameters (regularization parameters and kernel parameters);

[0082] (2) For each seagull individual, train the KELM model using the current hyperparameter configuration;

[0083] (3) Evaluate the performance of the model on the validation set and calculate the fitness value, that is, calculate the mean square error (MSE) value;

[0084] (4) Record the fitness value of each seagull individual;

[0085] (5) Sort all seagulls according to their fitness values, usually from best to worst, or the sorting order can be determined based on whether the problem is to minimize or maximize the objective function.

[0086] Update the seagull position and recalculate the fitness value;

[0087] Determine whether the maximum number of iterations of the algorithm has been reached or whether the optimal fitness has been reached. If so, output the current optimal parameter combination (regularization coefficient C and kernel parameter σ). If not, continue to update the seagull position and recalculate the fitness value.

[0088] Furthermore, the updating of the seagull position includes updating the position based on the seagull's migration behavior and foraging behavior.

[0089] Furthermore, when updating the position based on the migratory behavior of the seagull, the seagull is moved to the current optimal position by controlling the variables.

[0090] Furthermore, when updating the position based on the seagull's foraging behavior, the seagull continuously attacks the prey in a spiral descending arrangement, and the position is updated according to the corresponding motion trajectory formula.

[0091] Specifically, the Seagull Optimization Algorithm (SOA) is a novel population-based search algorithm that can be used to solve optimization problems in various fields. The algorithm simulates the migration and foraging behavior of a seagull population to implement both global and local search functions. Global search is used to quickly locate the optimal solution range, while local search is used to find the optimal solution. Each seagull's position represents a potential solution to the problem. The mathematical model is as follows:

[0092] (1) Migration behavior

[0093] To find abundant food (the optimal solution to the problem), seagulls migrate to habitats suitable for survival. This migration behavior affects the global search capability of the SOA algorithm. To avoid collisions between seagulls during migration, a new variable A is introduced to control the position of the seagulls:

[0094] D s (t) = A × G s (t)

[0095]

[0096] Where: s = 1, 2, ..., pop, (pop is the initial population size); t is the current number of iterations; tmax is the maximum number of iterations; D S To avoid collision, the seagull should be in position; G S is the current position of the seagull; f C A is a constant that controls the range of variation of the value of A and decreases linearly with the increase of the number of iterations.

[0097] After avoiding collisions, the gulls should move toward the best individual;

[0098] M s (t)=B[G bs (t)-G s (t)]

[0099] B=2×A 2 ×rand

[0100] E s (t)=|D s (t)+M s (t)|

[0101] Where: M S (t) is the current best position of the seagull, Seagull G bs (t) direction of movement; E S (t) is the distance between the seagull and the current best seagull; rand is a random number between [0,1]; B is the convergence factor for balancing development and exploration.

[0102] (2) Foraging behavior

[0103] The seagulls continuously attack their prey in a spiral descending formation. Their attacking behavior affects the local development capability of the SOA algorithm. Their movement trajectory is described as follows:

[0104]

[0105] Where x, y, and z are the position vectors of the seagull in three-dimensional space; r is the flight radius of the seagull controlled by the spiral coefficients u and v, where u and v are equal to 1; k is the angle of the seagull's random attack in the range [0, 2π]; and e is the base of the natural logarithm.

[0106] After the migration and foraging behavior, the overall updated position of the seagull is G s (t+1):

[0107] G s (t+1)=E s (t)×(x,y,z)+G bs (t)

[0108] In this embodiment, when establishing the KELM prediction model, the reference of the kernel function makes KELM have a more powerful function approximation ability than ELM, but increases the sensitivity to parameter selection. When using KELM to predict the height of the water-conducting fracture zone at the top of the coal seam, large-scale input data increases the complexity of the model calculation. In order to ensure the stability of the algorithm and the prediction accuracy, it is necessary to set a suitable regularization coefficient C and kernel parameter σ. Since the regularization coefficient and kernel parameters in KELM are generally preset manually, the prediction effect depends entirely on the experience of the staff, which limits the application effect of the prediction model in the actual engineering site. Therefore, in order to avoid the subjectivity of manually setting parameters and improve the accuracy and efficiency of predicting the height of the water-conducting fracture zone at the top of the coal seam, the SOA algorithm is used to optimize the regularization coefficient C and kernel parameter σ of the KELM model.

[0109] S5. Establishing a SOA-KELM prediction model based on the optimal parameter combination;

[0110] Furthermore, the process of establishing the SOA-KELM prediction model includes:

[0111] The optimal parameter combination is substituted into the KELM model to construct a model for predicting the height of the water-conducting fracture zone in the coal seam roof.

[0112] S6. Use the test set samples to evaluate and verify the performance of the SOA-KELM prediction model, and output the height of the water-conducting fracture zone in the coal seam roof of the coal mine.

[0113] Example:

[0114] The 42 sets of training sample data {X1, X2, X3, X4, X5} were used as input features, and the height of the water-conducting fracture zone {y} of the 42 sets of training sample data was used as the model output. The KELM parameters were optimized using the Seagull algorithm. The number of seagulls, pop, was set to 20; the maximum number of iterations, T, was set to 300; the dimension, dim, was set to 2, optimizing two parameters (the regularization coefficient, C, and the kernel function parameter, σ); the lower bound, lb, was set to [1, 1], and the upper bound, ub, was set to [100, 100]. The mean square error (MSE) between the expected and actual outputs of the KELM model's training samples was used as the fitness function, calculated as follows:

[0115]

[0116] Where: y i is the actual value; is the predicted value.

[0117] After initializing each parameter and setting the fitness function, the SOA algorithm was used to optimize the KELM parameters. The fitness value of each seagull was calculated based on the fitness function. If the updated fitness value was better than the previous value, it was replaced and used as the starting time for the next iteration. Through successive iterations, if the obtained extreme value was less than the set threshold or the maximum number of iterations was reached, the optimal parameters were output and used to build the model. The entire optimization process was implemented using the MATLAB 2018b platform.

[0118] The fitness change curve during training is as follows Figure 1 As shown in the figure, the fitting results of the trained model for the training samples are as follows Figure 2 As shown in the figure, the fitness value decreases with increasing iterations. After 78 iterations, the algorithm reaches a plateau, and the fitness value remains unchanged in subsequent iterations, indicating that the model has reached its highest prediction accuracy. At this point, the MSE value for the training sample is 0.004271, and the goodness-of-fit R² is 0.972. The SOA algorithm optimizes the KELM to obtain a regularization coefficient C and kernel function parameter σ of 98.2936 and 1, respectively.

[0119] While the established model has excellent fitting ability, attention should also be paid to whether it exhibits overfitting. This refers to the model's high prediction accuracy around the training samples, but low prediction accuracy for new sample data. Therefore, to verify the accuracy and stability of the established model, model validation was performed using test sample data. Eleven reserved test samples (numbered 43 to 53) were used to test the model's performance. The prediction results for the test samples are shown in Table 2. Table 2 shows that the absolute value of the difference between the predicted and actual values in the test samples ranged from a maximum of 8.7586 m to a minimum of 0.8766 m, with an average absolute error of 4.2462 m. The relative error ranged from a maximum of 16.13% to a minimum of 1.46%, with an average relative error of 9.38%. This indicates that the model has good predictive ability for new samples.

[0120] Table 2

[0121]

[0122] To further verify the effectiveness of the prediction model, a comparative analysis was conducted with two other models, including the traditional KELM model and the SVR model. Among them, the traditional KELM model uses the original data of the five main control factors as the model input and the height of the water-conducting fracture zone as the model output. The regularization coefficient C and the kernel function parameter σ in the model are set by trial and error. After multiple tests, C = 20 and σ = 1 are taken. At this time, the goodness of fit R between the predicted value and the actual value of the training sample is 0. 2=0.8469, indicating that the model training effect is good; the SVR model uses the original data of the five main control factors as the model input and the height of the water-conducting fracture zone as the model output. The penalty factor S and the kernel function parameter σ in the model are set by trial and error. After multiple tests, S=10 and σ=1 are taken. At this time, the goodness of fit R between the predicted value of the training sample and the actual value is 2 =0.9269, indicating that the model training effect is good.

[0123] The prediction results of these two models for training samples are as follows Figure 3 As shown, combined Figure 2 ,It can be seen that the SOA-KELM model has the best ,prediction effect for the training samples, and its goodness of fit is significantly ,higher than the KELM model and the SVR model.

[0124] The prediction results of the three models for the test samples are as follows: Figure 4 and as shown in Table 3. All three models demonstrate a certain degree of predictive ability for new samples. For the 11 test samples, the SOA-KELM model provided five optimal prediction results, while the KELM model and SVR model each provided three optimal prediction results. The mean absolute errors for the SOA-KELM, KELM, and SVR models for the test sample data were 2.4264 m, 13.1184 m, and 5.4846 m, respectively. The mean relative errors for the test sample data were 9.38%, 28.96%, and 11.65%, respectively. This indicates that the SOA-KELM model has better prediction accuracy and robustness for new sample data than the KELM and SVR models, providing an effective approach and method for accurately predicting the height of water-conducting fracture zones in coal seam roofs. The established SOA-KELM model has certain practicality, and its accuracy meets the requirements for predicting the height of water-conducting fracture zones in actual modern mine engineering.

[0125] Table 3

[0126]

[0127]

[0128] Beneficial effects of this embodiment:

[0129] This embodiment proposes a new method for predicting the height of water-conducting fracture zones in coal seam roofs of coal mines based on SOA-KELM. This method is practical and its accuracy can meet the requirements for predicting the development height of water-conducting fracture zones in actual modern mine engineering.

[0130] This embodiment combines the SOA algorithm to realize the automatic optimization of the key parameters of the KELM model, avoiding the performance defects caused by the artificial pre-setting of the key parameters of the KELM model, effectively improving the model prediction accuracy, and verifying the stability and superiority of the SOA-KELM-based coal seam roof water-conducting fracture zone height prediction model when compared with the traditional KELM model and the SVR model.

[0131] Example 2

[0132] Based on the same inventive concept, this embodiment further provides a coal mine coal seam roof water-conducting fracture zone height prediction system based on SOA-KELM, which is used to implement the method described in Example 1, including:

[0133] The data acquisition module is used to determine the main controlling factors affecting the height of the water-conducting fracture zone in the coal seam roof and collect the corresponding measured data;

[0134] The data processing module is used to divide the collected measured data into training set samples and test set samples, and standardize the data;

[0135] The parameter setting module is used to initialize the parameters of the Seagull Optimization Algorithm (SOA) and set the fitness function to the root mean square error between the expected output and the actual output of the training sample of the kernel extreme learning machine (KELM) model.

[0136] Parameter optimization module, which is used to optimize the regularization coefficient and kernel function parameters of the KELM model based on the SOA algorithm to obtain the optimal parameter combination;

[0137] A model building module, used to establish a SOA-KELM prediction model based on the optimal parameter combination;

[0138] The result prediction module is used to evaluate and verify the performance of the SOA-KELM prediction model using test set samples and output the height of the water-conducting fracture zone in the coal seam roof of the coal mine.

[0139] The coal mine coal seam roof water-conducting fracture zone height prediction system based on SOA-KELM provided in this embodiment has all the advantages of the coal mine coal seam roof water-conducting fracture zone height prediction method based on SOA-KELM provided in Example 1.

[0140] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for predicting the height of water-conducting fracture zones in coal seam roofs of coal mines based on SOA-KELM, characterized in that: The following steps are involved: Determine the main controlling factors affecting the height of the water-conducting fracture zone in the coal seam roof and collect the corresponding measured data; Divide the collected measured data into training set samples and test set samples, and standardize the data; Initialize the parameters of the Seagull Optimization Algorithm (SOA) and set the fitness function to the root mean square error between the expected output and the actual output of the training sample of the kernel extreme learning machine (KELM) model. Based on the SOA algorithm, the regularization coefficient and kernel function parameters of the KELM model are optimized to obtain the optimal parameter combination; Establishing a SOA-KELM prediction model based on the optimal parameter combination; The test set samples were used to evaluate and verify the performance of the SOA-KELM prediction model, and the height of the water-conducting fracture zone in the coal seam roof of the coal mine was output.

2. The method according to claim 1, characterized in that The main control factors include the selection of coal seam roof type, mining depth, coal seam inclination, mining thickness, and working face inclined length.

3. The method according to claim 1, characterized in that The parameters of the initialization seagull optimization algorithm SOA include setting the number of seagull populations, the maximum number of iterations, the optimization parameter dimensions, and the upper and lower bounds of the parameters.

4. The method according to claim 1, wherein The formula of the fitness value function is: Where: y i is the actual value; is the predicted value.

5. The method according to claim 1, characterized in that The process of optimizing the regularization coefficient and kernel function parameters of the KELM model includes: Calculate the initial seagull fitness value and sort them; Update the seagull position and recalculate the fitness value; Determine whether the maximum number of iterations of the algorithm has been reached or whether the optimal fitness has been reached. If so, output the current optimal parameter combination. If not, continue to update the seagull position and recalculate the fitness value.

6. The method according to claim 5, characterized in that The updating of the seagull position includes updating the position based on the seagull's migration behavior and foraging behavior.

7. The method according to claim 6, characterized in that When updating the position based on the seagull's migration behavior, the seagull is moved to the current optimal position by controlling the variables.

8. The method according to claim 6, characterized in that When updating the position based on the seagull's foraging behavior, the seagull continuously attacks the prey in a spiral descending arrangement and updates the position according to the corresponding motion trajectory formula.

9. The method according to claim 1, characterized in that The process of building the SOA-KELM prediction model includes: The optimal parameter combination is substituted into the KELM model to construct a model for predicting the height of the water-conducting fracture zone in the coal seam roof.

10. A coal mine coal seam roof water-conducting fracture zone height prediction system based on SOA-KELM, characterized in that: The method for implementing any one of claims 1 to 9 comprises: The data acquisition module is used to determine the main controlling factors affecting the height of the water-conducting fracture zone in the coal seam roof and collect the corresponding measured data; The data processing module is used to divide the collected measured data into training set samples and test set samples, and standardize the data; The parameter setting module is used to initialize the parameters of the Seagull Optimization Algorithm (SOA) and set the fitness function to the root mean square error between the expected output and the actual output of the training sample of the kernel extreme learning machine (KELM) model. Parameter optimization module, which is used to optimize the regularization coefficient and kernel function parameters of the KELM model based on the SOA algorithm to obtain the optimal parameter combination; A model building module, used to establish a SOA-KELM prediction model based on the optimal parameter combination; The result prediction module is used to evaluate and verify the performance of the SOA-KELM prediction model using test set samples and output the height of the water-conducting fracture zone in the coal seam roof of the coal mine.

Images