Coal mining water flowing fractured zone development height prediction method based on SSA-RF model

Abstract

The invention relates to the technical field of mining, in particular to a coal mining water flowing fractured zone development height prediction method based on an SSA-RF model. Acquiring sample data influencing the development height of the water-flowing fractured zone and preprocessing the sample data; constructing a feature vector sample according to the preprocessed sample data; constructing a plurality of regression decision tree prediction models, and inputting the feature vector samples into the regression decision tree models for training; optimizing the trained regression decision tree prediction model by adopting a sparrow search algorithm to obtain an optimized prediction model; and predicting the development height of the water-flowing fractured zone of the to-be-mined area by using the optimized prediction model to obtain a prediction result of the development height of the water-flowing fractured zone, so that multiple influence factors can be effectively integrated to improve the prediction precision, thereby helping to formulate corresponding treatment measures according to the prediction result of the development height of the water-flowing fractured zone. The method has important significance for guiding and promoting safe production.

Description

Technical Field

[0001] The present invention relates to the technical field of mining, and particularly relates to a method for predicting the development height of water-conducting fissure zones in coal mining based on the SSA-RF model. Background Art

[0002] With the expansion of coal mining, the threat of the water-bearing sand layer at the bottom of the Neogene during coal seam mining is increasing. Once the water-conducting fissure zone caused by mining touches the water-bearing sand layer at the bottom of the Neogene, even if the bottom of the loose layer is a weak aquifer, water and sand may suddenly gush into the mining face, triggering water inrush and sand bursting safety accidents. How to accurately predict the development height of the water-conducting fissure zone under thick loose layers and thin bedrocks is the key to realizing safe and efficient coal mining under the Neogene loose layer in deep wells.

[0003] Currently, the main methods for predicting the development height of water-conducting fissure zones include empirical formula method, theoretical calculation method, similar material simulation, numerical simulation, in-situ measurement, etc. The empirical formula in the "Three-Under" regulations cannot reflect the comprehensive influence of various factors; the theoretical calculation method model has a large deviation from the complex geological occurrence conditions; the similar material simulation has high requirements for the accuracy of material ratio and is difficult to implement for the simulation of complex geological conditions; the accuracy of numerical simulation depends too much on the simplification of geological conditions and parameter acquisition; the in-situ measurement has high accuracy but is time-consuming and laborious. With the development of decision theory and artificial intelligence algorithms, some scholars have applied methods such as analytic hierarchy process-fuzzy clustering method, entropy weight method, principal component analysis method, fuzzy mathematics method, grey clustering evaluation, neural network method, support vector machine method, random forest algorithm, genetic algorithm, etc. to the prediction research of the development height of water-conducting fissure zones. However, these technical methods focus too much on mining thickness, mining depth, working face dip length, coal seam dip angle, overlying rock structure, mining method, roof compressive strength, mudstone ratio, etc., and there are few reports on factors such as the thickness of the bottom aquifer of the Neogene, water pressure, water-richness, loose base ratio, and loose depth ratio in the coal-bearing strata with thick loose layers and thin bedrocks. Although many prediction methods consider multiple influencing factors, they have the deficiencies of being prone to local optimal solutions and low model generalization ability, and it is difficult to balance prediction accuracy and robustness. Summary of the Invention

[0004] To solve the problems existing in the prior art, the present invention provides a method for predicting the development height of water-conducting fissure zones in coal mining based on the SSA-RF model. This method obtains sample data affecting the development height of water-conducting fissure zones and preprocesses it; constructs a feature vector sample according to the preprocessed sample data; constructs multiple regression decision tree prediction models, and inputs the feature vector sample into the regression decision tree model for training; uses the sparrow search algorithm to optimize the trained regression decision tree prediction model to obtain an optimized prediction model; and uses the optimized prediction model to predict the development height of the water-conducting fissure zone in the area to be mined, obtaining the prediction result of the development height of the water-conducting fissure zone, which can effectively integrate various influencing factors to improve the prediction accuracy, thereby helping to formulate corresponding treatment measures according to the prediction result of the development height of the water-conducting fissure zone, and having important significance for guiding and promoting safe production.

[0005] The present invention adopts the following technical solution. A method for predicting the development height of water-conducting fissure zones in coal mining based on the SSA-RF model includes:

[0006] Obtain sample data affecting the development height of water-conducting fissure zones and preprocess it;

[0007] Construct a feature vector sample according to the preprocessed sample data;

[0008] Construct multiple regression decision tree prediction models, and input the feature vector sample into the regression decision tree model for training;

[0009] Use the sparrow search algorithm to optimize the trained regression decision tree prediction model to obtain an optimized prediction model;

[0010] Use the optimized prediction model to predict the development height of the water-conducting fissure zone in the area to be mined.

[0011] Further, the sample data affecting the development height of water-conducting fissure zones includes: geological parameters and mining condition parameters;

[0012] The geological parameters include: the thickness of the bottom aquifer, the water-richness of the bottom aquifer, the water pressure of the bottom aquifer, the thickness of the bottom clay layer, the thickness of the bedrock, the coal seam dip angle, the coal seam burial depth, the loose base ratio, and the loose depth ratio;

[0013] The mining condition parameters include: the mining height, the working face length, and the mining method.

[0014] Further, the method for obtaining sample data affecting the development height of water-conducting fissure zones and preprocessing it is:

[0015] Divide the sample data into quantitative factors and qualitative factors, and perform fuzzy quantification processing on the qualitative factors; where:

[0016] The quantitative factors include: the thickness of the bottom aquifer, the water pressure of the bottom aquifer, the thickness of the bottom clay layer, the thickness of the bedrock, the dip angle of the coal seam, the burial depth of the coal seam, the loose base ratio, the loose depth ratio, the mining height, and the working face length;

[0017] The qualitative factors include: the mining method and the water-richness of the bottom aquifer.

[0018] Furthermore, a feature vector sample is constructed based on the preprocessed sample data, specifically:

[0019] The sample data is transformed into a feature vector sample through a standardization method, and the expression is:

[0020]

[0021] In the formula, X scaled is the standardized feature vector, and X represents the sample data value.

[0022] Furthermore, the feature vector sample is input into the regression decision tree model for training, including:

[0023] The feature vector sample is divided into a training set and a test set;

[0024] Using the feature vector sample in the training set as the input of the regression tree decision model and the development height of the water-conducting fissure zone as the output, a mapping relationship between the input and the output is established, and the expression is:

[0025] S = TR(Z)

[0026] Among them, S is the development height of the water-conducting fissure zone, TR is the expression form of the established mapping relationship, and Z is the input feature vector sample.

[0027] Furthermore, the sparrow search algorithm is used to optimize the trained regression decision tree prediction model, including:

[0028] Define the parameter search range and set the hyperparameters of the regression decision tree prediction model; the hyperparameters include: the number of prediction models, the maximum tree depth, the minimum number of samples for splitting, and the minimum number of samples for leaf nodes;

[0029] Set the initial population size in the sparrow search algorithm and use the hyperparameters to represent the position of each individual in the population;

[0030] Use the out-of-bag error as the fitness function to calculate the model performance corresponding to each individual in the population;

[0031] Use the update strategy of the sparrow search algorithm to update the individual positions and calculate the model performance after each update;

[0032] Iterate sequentially until the population converges to the optimal solution or the iteration reaches the maximum number of times, and then stop. Obtain the hyperparameters corresponding to each individual at this time as the optimal parameters.

[0033] The beneficial effects of the present invention are as follows: The prediction method proposed by the present invention can effectively integrate various influencing factors and improve the prediction accuracy, realizing the rapid and relatively accurate acquisition of the development height of the water-conducting fractured zone in the unmined area of the coal mine, which has important significance for guiding and promoting safe production. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0035] Figure 1 It is a schematic flow chart of a method for predicting the development height of the water-conducting fractured zone in coal mine mining based on the SSA-RF model according to an embodiment of the present invention;

[0036] Figure 2 It is a schematic diagram of a method for establishing a feature vector sample according to an embodiment of the present invention;

[0037] Figure 3 It is a schematic diagram of the structure of a prediction model according to an embodiment of the present invention;

[0038] Figure 4 It is a histogram of the comparison results between the prediction results and the survey results according to an embodiment of the present invention;

[0039] Figure 5 It is a schematic diagram of the comparison curve between the prediction results and the survey results according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0041] A schematic flow chart of a method for predicting the development height of the water-conducting fractured zone in coal mine mining based on the SSA-RF model according to an embodiment of the present invention is as Figure 1 shown and includes:

[0042] Step 1: Collect the key hydrogeological parameters and mining condition parameters that affect the development height of the water-conducting fractured zone;

[0043] In the embodiments of the present invention, the hydrogeological key parameters and the mining condition parameters together constitute the sample data affecting the development height of the water-conducting fissure zone. Among them, the hydrogeological key parameters include: the thickness of the bottom aquifer, the water-richness of the bottom aquifer, the water pressure of the bottom aquifer, the thickness of the bottom clay layer, the thickness of the bedrock, the coal seam dip angle, the burial depth of the coal seam, the loose base ratio, and the loose depth ratio. The mining condition parameters include: the mining height, the working face length, and the mining method.

[0044] Step 2: Preprocess the prediction sample data;

[0045] In the embodiments of the present invention, the method for preprocessing the sample data is as Figure 2 shown. First, the sample data is divided into quantitative factors and qualitative factors. Among them, the parameters included in the quantitative factors are: the thickness of the bottom aquifer, the water pressure of the bottom aquifer, the thickness of the bottom clay layer, the thickness of the bedrock, the coal seam dip angle, the burial depth of the coal seam, the loose base ratio, the loose depth ratio, the mining height, and the working face length. The parameters included in the qualitative factors are: the mining method and the water-richness of the bottom aquifer.

[0046] Furthermore, the following fuzzy quantification processing needs to be performed on the qualitative factors:

[0047] Let q be the water-richness of the bottom aquifer and d be the mining method. Then the operation of the fuzzy quantification processing of the water-richness of the bottom aquifer is expressed as:

[0048]

[0049] The operation of the fuzzy quantification processing of the mining method is expressed as:

[0050]

[0051] In summary, the specific parameters after the fuzzy quantification of the water-richness of the bottom aquifer and the mining method in the qualitative factors are shown in Table 1.

[0052] Table 1 Schematic table of fuzzy quantification parameters for qualitative factors

[0053]

[0054] Step 3: Construct the characteristic vector sample data affecting the development height of the water-conducting fissure zone;

[0055] In the present invention, by performing standardization processing on the preprocessed sample data, the characteristic vector sample for model training is obtained. The expression for the characteristic vector standardization operation is:

[0056]

[0057] In the formula, X scaled is the standardized characteristic vector.

[0058] Step 4: Based on the feature vector samples, train and construct multiple regression decision tree prediction models, and establish a non-linear mapping relationship between the input parameters and the output parameters through the multiple prediction models;

[0059] In the embodiment of the present invention, the obtained feature vector samples are divided into a training set and a test set according to a set ratio for model training and verification. Twelve feature vector samples are used as the input parameters of the model, and the output parameter is the development height of the water-conducting fissure zone.

[0060] The embodiment of the present invention further constructs multiple regression decision tree prediction models, and establishes a non-linear mapping relationship between the input parameters and the output parameters through the models. The training process is as Figure 3 shown, which can be expressed as: S = TR(Z); where S is the development height of the water-conducting fissure zone, TR is the expression form of the established mapping relationship, Z is the input feature vector sample, Z = [z1, z2, z3, z4, z5, z6, z7, z8, z9, z10, z11, z12], z1 is the thickness of the bottom aquifer, z2 is the water-richness of the bottom aquifer, z3 is the water pressure of the bottom aquifer, z4 is the thickness of the bottom clay layer, z5 is the thickness of the bedrock, z6 is the dip angle of the coal seam; z7 is the burial depth of the coal seam, z8 is the loose base ratio, z9 is the loose depth ratio, z10 is the mining height, z11 is the length of the working face, and z12 is the mining method.

[0061] Step 5: Use the Sparrow Search Algorithm (SSA) to obtain the optimal values of the parameter prediction model quantity Q, the maximum tree depth D, the minimum splitting sample number S, and the minimum leaf node sample number L of the regression decision tree prediction model;

[0062] In the embodiment of the present invention, the obtaining of the optimal values includes the following steps:

[0063] (1) Initialize the population position: First, define the parameter search range and set the hyperparameters of the regression decision tree prediction model: the prediction model quantity Q, the maximum tree depth D, the minimum splitting sample number S, and the minimum leaf node sample number L; Set the initial population size to n, and the position of each individual is represented as W i =(Q i ,D i ,S i ,L i ).

[0064] (2) Define the fitness function: Use the Out-of-Bag Error (OOB Error) as the fitness function to calculate the model performance corresponding to each individual in the current population. The formula is:

[0065]

[0066] In the formula, yi is the development height of the true water-conducting fissure zone, is the model prediction value, and n is the number of samples.

[0067] (3) Population update strategy: Adjust the individual positions using two update strategies of the Sparrow Search Algorithm (SSA), including:

[0068] Discovery update: Conduct extensive searches in areas with a large search range, and update the position through the following formula:

[0069] W i t+1 = W i t ·e -V + Q best ·V

[0070] In the formula: W i t+1 is the position of the individual in the t-th generation, Q best is the position of the current optimal individual, and V is a random number with a normal distribution.

[0071] Joiner update: Adjust through local exploration based on the current global optimal position Q best and the global worst position Q worst :

[0072] W i t+1 = W i t + r·(Q best- Q worst )

[0073] In the formula: r is a random number, and its value range is [-1, 1].

[0074] Scouts: Approximately 10% - 20% of the total population, generally sparrows that discover danger, and their position update is expressed as follows:

[0075]

[0076] In the formula, f i is the fitness value of the sparrow individual, f g and f w are the current global maximum and minimum fitness values respectively; β and K ∈ [-1, 1] are random numbers with a normal distribution; and are the current global optimal position and the worst position respectively; ε is a constant; when f i > f g it means that the sparrow is currently at the periphery of the population; when f i = f gWhen it is, it means that the sparrows in the middle of the population have sensed danger and tend to approach each other to reduce the risk of being captured.

[0077] (4) When the population as a whole converges to the optimal solution or reaches the maximum number of iterations, stop the optimization, record the current optimal parameter combination, and thus complete the parameter optimization of the model.

[0078] Step 6: Assign the optimal prediction model quantity Q, maximum tree depth D, minimum splitting sample number S, and minimum leaf node sample number L obtained by optimizing through the Sparrow Search Algorithm (SSA) to the aforementioned prediction model, train the established feature vector samples, and obtain the optimized prediction model.

[0079] In the embodiment of the present invention, the optimized parameter combination is further used to retrain the regression decision tree prediction model for the established feature vector samples to construct the final prediction model.

[0080] Generate Q distinctive training sets from the established feature vector samples by sampling with replacement, establish Q distinctive prediction models with a maximum tree depth of D, a minimum splitting sample number of S, and a minimum leaf node sample number of L, use the weighted average method to realize the prediction of a single optimized prediction model, obtain the average value of the prediction results of the finally formed multiple optimized prediction models, and finally establish the mapping relationship between the key parameters and the development height of the mining-induced water-conducting fissure zone.

[0081] Step 7: Input the measured values of the key parameters in the unmined area into the optimized prediction model for prediction, obtain the prediction result of the development height of the water-conducting fissure zone, and formulate corresponding disposal measures according to the prediction result of the development height of the water-conducting fissure zone.

[0082] In the embodiment of the present invention, the key parameters of the unmined area of the coal mine, such as the water pressure of the bottom aquifer, the thickness of the bottom clay layer, the thickness of the bedrock, the coal seam dip angle, the coal seam burial depth, the loose base ratio, the loose depth ratio, the mining height, the working face length, and the mining method, are surveyed in advance, and the above-mentioned surveyed parameters are input into the optimized prediction model. Through the optimized prediction model, the predicted value of the development height of the mining-induced water-conducting fissure zone in the unmined area of the coal mine is output. According to the prediction result of the development height of the mining-induced water-conducting fissure zone in the coal mine, corresponding disposal measures are formulated in advance to ensure the smooth progress of safe production.

[0083] The present invention can realize the rapid and relatively accurate acquisition of the development height of the mining-induced water-conducting fissure zone in the unmined area of the coal mine, which has important significance for promoting safe production.

[0084] In another experimental embodiment of the present invention:

[0085] In the embodiment of the present invention, taking the working face parameter data of 23 mines under the condition of thick loose layer and thin bedrock in the actual situation as an example, the specific data is shown in Table 2:

[0086] Table 2 Schematic table of working face parameter data of different mines under the conditions of thick loose layer and thin bedrock

[0087]

[0088] Predict the development height of the water-conducting fissure zone. The prediction results and the actual development height results of the water-conducting fissure zone after actual tunneling in the later stage are as Figure 4 and Figure 5 shown. The comparison between the prediction results and the actual results is represented by a histogram and a curve respectively. It can be seen that the prediction results of the development height of the water-conducting fissure zone in the unmined area are relatively close to the acquisition results of the development height of the water-conducting fissure zone after actual excavation, verifying the feasibility and accuracy of this method. Through this prediction method, it can effectively guide and promote safe production, and effectively avoid the occurrence of water inrush disasters in coal mining.

[0089] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for predicting the development height of water-conducting fracture zones in coal mining based on the SSA-RF model, characterized in that: include: Obtain sample data that affects the height of water-conducting fracture zones and perform preprocessing; Construct feature vector samples based on preprocessed sample data; Constructing multiple regression decision tree prediction models, and inputting the feature vector samples into the regression decision tree models for training; Using a sparrow search algorithm to optimize the trained regression decision tree prediction model to obtain an optimized prediction model; The optimization prediction model is used to predict the development height of the water-conducting fracture zone in the area to be mined.

2. According to claim 1, a method for predicting the height of water-conducting fracture zones in coal mining based on the SSA-RF model, characterized in that: The sample data affecting the height of the water-conducting fracture zone development include: geological parameters and mining condition parameters; The geological parameters include: bottom aquifer thickness, bottom aquifer water richness, bottom aquifer water pressure, bottom clay layer thickness, bedrock thickness, coal seam inclination, coal seam burial depth, loose-to-base ratio and loose-to-depth ratio; The mining condition parameters include: mining height, working face length and mining method.

3. The method for predicting the height of water-conducting fracture zones in coal mining based on the SSA-RF model according to claim 1 is characterized in that: The method for obtaining sample data that affects the height of the water-conducting fracture zone and performing preprocessing is: The sample data is divided into quantitative factors and qualitative factors, and the qualitative factors are subjected to fuzzy quantification processing; wherein: The quantitative factors include: bottom aquifer thickness, bottom aquifer water pressure, bottom clay layer thickness, bedrock thickness, coal seam inclination, coal seam burial depth, loose-to-base ratio, loose-to-depth ratio, mining height, and working face length; Qualitative factors include: extraction method and water richness of the underlying aquifer.

4. The method for predicting the height of water-conducting fracture zones in coal mining based on the SSA-RF model according to claim 1, characterized in that: Construct feature vector samples based on the preprocessed sample data, specifically: The sample data is converted into feature vector samples through standardization method, and the expression is: Where, X sca l e d is the standardized feature vector, and X represents the sample data value.

5. The method for predicting the height of water-conducting fracture zones in coal mining based on the SSA-RF model according to claim 1, characterized in that: Inputting the feature vector sample into the regression decision tree model for training, including: Dividing the feature vector samples into a training set and a test set; The feature vector samples in the training set are used as the input of the regression tree decision model, and the height of the water-conducting fracture zone is used as the output. The mapping relationship between the input and the output is established, and the expression is: S=TR(Z) Among them, S is the development height of the water-conducting fracture zone, TR is the expression form of the established mapping relationship, and Z is the input feature vector sample.

6. The method for predicting the height of water-conducting fracture zones in coal mining based on the SSA-RF model according to claim 1, characterized in that: The trained regression decision tree prediction model is optimized using a sparrow search algorithm, including: Define a parameter search range and set hyperparameters of the regression decision tree prediction model; the hyperparameters include: the number of prediction models, the maximum tree depth, the minimum number of split samples, and the minimum number of leaf node samples; Setting the initial population size in the sparrow search algorithm and using the hyperparameters to represent the position of each individual in the population; Using the out-of-bag error as the fitness function, the model performance corresponding to each individual in the population is calculated; Using the update strategy of the sparrow search algorithm to update the individual positions, and calculating the model performance after each update; Iterate in sequence until the population converges to the optimal solution or the iteration reaches the maximum number of times, and then obtain the hyperparameters corresponding to each individual at this time as the optimal parameters.

Images