A Prediction Method for Grouting Effect in Underground Water-Sealed Caverns Based on an Ensemble Learning Model

Abstract

This invention relates to a method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model. The method includes: acquiring environmental and construction parameters of the underground water-sealed cavern; inputting these parameters into an ensemble learning-based grouting effect prediction model to obtain grouting effect prediction results; training the grouting effect prediction model using a training set and simultaneously optimizing the trained model parameters using a simulated annealing algorithm; the training set includes geological parameters, hydrological parameters, and historical construction parameters; wherein the grouting effect prediction model includes: a random forest model, a gradient boosting tree model, an AdaBoost model, and an ANN model. This invention combines machine learning techniques, especially ensemble learning and simulated annealing algorithm for hyperparameter optimization, to provide a scientific basis for rock mass grouting seepage control construction decisions in underground water-sealed cavern projects, thereby improving the overall construction quality and energy efficiency of grouting projects.

Description

Technical Field

[0001] This invention relates to the field of rock mass grouting and seepage control construction technology in underground water-sealed cavern engineering, and in particular to a method for predicting the grouting effect of underground water-sealed caverns based on an integrated learning model. Background Technology

[0002] Underground water-sealed caverns, as an important component of the national energy reserve system, are used not only for storing crude oil but also for storing other petroleum products, such as refined oil and LPG. It is a method of altering and utilizing natural rock formations through artificial intervention, offering advantages such as safety, environmental friendliness, high concealment, low construction and operating costs, and minimal occupation of arable land. The construction and operation of underground water-sealed caverns bring significant benefits to the national economy, society, and environment, playing a crucial role in enhancing national energy security and ensuring the stability of energy supply.

[0003] During the construction of underground water-sealed caverns, the rock mass is affected by a combination of factors such as blasting and excavation, resulting in various fissures on the surface and inside of the cavern's surrounding rock. This allows groundwater to seep through, which reduces the stability of the surrounding rock and increases the pumping and drainage costs of the cavern's operation. Therefore, it is essential to carry out seepage control treatment for the cavern.

[0004] Currently, grouting is the primary method for controlling seepage fields in caverns. By injecting grout into joints, fissures, alteration zones, and other adverse geological structures, the integrity of the cavern's rock mass structure is improved. After solidification, the grout significantly enhances the strength and integrity of the rock mass, effectively preventing leakage. However, due to the concealed nature of rock grouting, it is difficult to visually evaluate the distribution and solidification of the grout, making it challenging to accurately assess the construction quality and effectiveness. Therefore, if the grouting quality of each borehole sequence could be conveniently and effectively predicted in advance during the grouting process, it would be of great significance for controlling the overall grouting construction quality and improving grouting efficiency.

[0005] With the development of computer technology, numerical simulation analysis methods such as the finite element method can be used to simulate the grouting process. However, because grouting is a complex and hidden system engineering project, many factors affect the grouting effect, making it difficult to play a significant role in evaluating grouting results. Moreover, numerical simulation methods often suffer from drawbacks such as complex modeling processes and high modeling time costs, failing to meet the needs of actual on-site construction. Therefore, it is necessary to establish a data-driven, fast, and accurate intelligent model to predict grouting effects and provide assistance for grouting construction decisions.

[0006] With the development of artificial intelligence technology, machine learning methods have become feasible for predicting grouting effects. This approach can effectively address the shortcomings in evaluating grouting effects caused by the concealment of rock mass grouting through data-driven methods. Early predictions of rock mass grouting effects were mostly based on simple weak learners, such as decision trees. These prediction models often lead to unsatisfactory prediction results when dealing with insufficient sample data or too many input factors. Summary of the Invention

[0007] To address the problems of existing technologies in predicting the effect of rock grouting, such as high evaluation difficulty, limited simulation methods, high prediction model complexity, difficulty in hyperparameter optimization, insufficient generalization ability, and lack of real-time prediction and decision support, this invention provides a method for predicting the grouting effect of underground water-sealed caverns based on an integrated learning model. This method can provide reliable methodological support for estimating the grouting volume in the area to be grouted, has significant engineering application value, and due to its high requirements for controlling rock mass seepage, the model is not only applicable to specific projects but also has the potential to be extended to other underground grouting projects.

[0008] To achieve the above objectives, the present invention provides the following solution:

[0009] A method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model includes:

[0010] Environmental and construction parameters of the underground water-sealed cavern are obtained, and the environmental and construction parameters are input into a grouting effect prediction model based on ensemble learning to obtain grouting effect prediction results. The grouting effect prediction model is trained using a training set, and the parameters of the trained model are simultaneously optimized using a simulated annealing algorithm. The training set includes geological parameters, hydrological parameters, and historical construction parameters.

[0011] The grouting effect prediction model includes: random forest model, gradient boosting tree model, AdaBoost model and ANN model; the random forest model, gradient boosting tree model and AdaBoost model are all connected to the ANN model.

[0012] Optionally, the geological parameters include: surrounding rock grade, joint surface density, and alteration zone density;

[0013] The hydrological parameters include: pre-irrigation permeability;

[0014] The construction parameters include: grouting pressure, hole sequence, and elevation parameters.

[0015] Optionally, obtaining the grouting effect prediction result includes:

[0016] The environmental parameters and construction parameters are input into a random forest model, a gradient boosting tree model, and an AdaBoost model to obtain a first prediction result; the random forest model is a random forest model that integrates the decision tree as the base learner using the Bagging method, and the gradient boosting tree model is a gradient boosting tree model that integrates the decision tree as the base learner using the Boosting method.

[0017] The first prediction result is used as input data to input the ANN model to obtain the grouting effect prediction result; the ANN model is an ANN model that integrates the base learners using the Stacking method.

[0018] Optionally, using simulated annealing to simultaneously optimize the trained model parameters includes:

[0019] Step 1: Use the model parameters as the initial solution and set the initial temperature and cooling coefficient;

[0020] Step 2: Update the initial temperature according to the cooling plan, and randomly select a new solution within the neighborhood of the initial solution;

[0021] Step 3: Calculate the fitness difference of the new solution, and judge the fitness difference. If the fitness difference is less than a preset value, the new solution is accepted. If the fitness difference is greater than the preset value, the new solution is accepted according to the Metropolis criterion.

[0022] Step 4: Take the new solution as the current solution and repeat steps 1-3 until the updated temperature is lower than the target threshold or the preset number of iterations is reached to obtain the optimal solution.

[0023] Optionally, the model parameters include: the learning rate and number of weak learners for the AdaBoost model, the learning rate, maximum tree depth, number of trees, and minimum number of samples per leaf node for the gradient boosting tree model, the number of trees, maximum tree depth, and minimum number of samples per leaf node for the random forest model, and the neuron weights, number of hidden layers, and activation function for the ANN model.

[0024] Optionally, calculating the fitness difference of the new solution includes:

[0025] Δf = f(i') - f(i);

[0026] Where Δf is the fitness difference, f(i') is the fitness of the new solution, and f(i) is the fitness difference of the current solution.

[0027] Optionally, accepting the new solution according to the Metropolis criterion includes:

[0028]

[0029] Where P(Δf) is the new solution accepted according to the Metropolis criterion, T is the temperature, and e is the base of the natural logarithm.

[0030] Optionally, the method further includes:

[0031] The synchronously optimized model was validated using the coefficient of determination, root mean square error, mean square error, mean absolute error, and mean absolute percentage error.

[0032] The beneficial effects of this invention are as follows:

[0033] The advantages and positive effects of this invention are as follows: This invention employs an ensemble learning-based prediction model, combining the diversity and stability of base learners such as random forests, gradient boosting trees, and AdaBoost, with the nonlinear relationship processing capabilities of ANN neural network meta-learners, effectively improving the accuracy and stability of grouting volume prediction. This ensemble method fully utilizes the advantages of different learners. Random forests, due to the randomness of their construction process, possess excellent robustness and are less prone to overfitting; gradient boosting trees, through iterative optimization, can adapt to various loss functions and achieve high accuracy; AdaBoost improves model performance by focusing on misclassified samples. Simultaneously, ANN neural networks, as meta-learners, effectively handle the uncertainty of prediction results and prevent overfitting due to their excellent numerical approximation capabilities. Furthermore, this invention employs parameter optimization techniques such as simulated annealing to further enhance model performance. The simulated annealing algorithm is chosen because it maintains population diversity during the search process and avoids getting trapped in local optima.

[0034] This invention addresses the following technical problems of existing machine learning models for predicting grouting effects: insufficient consideration of hydrogeological parameters, excessively high time costs for numerical simulation, insufficient sample size leading to a lack of persuasiveness, and insufficient accuracy despite sufficient sample size. Compared to weak learner models, this invention significantly improves the performance of predicting grouting effects, enabling rapid and accurate guidance for on-site construction decisions and ensuring the economic benefits and engineering stability of grouting. In actual grouting projects, this invention provides reliable methodological support for estimating the grouting volume in the area to be grouted, possessing significant engineering application value. Furthermore, due to its high requirements for controlling rock mass seepage, the model is not only applicable to specific projects but also has the potential to be extended to other underground grouting projects. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a flowchart of a method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model, according to an embodiment of the present invention.

[0037] Figure 2 This is a schematic diagram of the process for optimizing the grouting effect of underground water-sealed caverns based on an ensemble learning model, according to an embodiment of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] This embodiment uses Bagging, Boosting, and Stacking as three main ensemble learning methods in machine learning. They improve predictive performance by combining multiple models, and each method has its unique characteristics and applicable scenarios. By innovatively fusing these three ensemble learning methods, a powerful predictive model is constructed that can not only capture complex patterns and nonlinear relationships in the data but also improve the accuracy and robustness of predictions.

[0041] Furthermore, the parameter configuration of each model has a crucial impact on the prediction results. Precise parameter settings can significantly improve the accuracy of model predictions. Therefore, choosing an efficient optimization algorithm to tune the model's hyperparameters is particularly important. Simulated Annealing, as a probabilistic global optimization method, mimics the stochastic search mechanism in the annealing process of materials, effectively avoiding local optima and thus exploring the global optimum, providing a powerful strategy for model hyperparameter optimization.

[0042] In summary, this embodiment, by combining machine learning technology, especially the hyperparameter optimization of ensemble learning and simulated annealing algorithms, can provide a fast and accurate intelligent model for predicting the effect of rock grouting, and provide a scientific basis for decision-making on seepage control construction of rock grouting in underground water-sealed cavern projects, thereby improving the overall construction quality and energy efficiency of grouting projects.

[0043] like Figure 1 As shown in the figure, this embodiment discloses a method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model. The method includes: acquiring environmental and construction parameters of the underground water-sealed cavern; inputting the environmental and construction parameters into an ensemble learning-based grouting effect prediction model to obtain grouting effect prediction results; training the grouting effect prediction model using a training set and simultaneously optimizing the trained model parameters using a simulated annealing algorithm; the training set includes geological parameters, hydrological parameters, and historical construction parameters; the grouting effect prediction model includes a random forest model, a gradient boosting tree model, an AdaBoost model, and an ANN model; the random forest model, gradient boosting tree model, and AdaBoost model are all connected to the ANN model.

[0044] Furthermore, the geological parameters include: surrounding rock grade, joint surface density, and alteration zone density; the hydrological parameters include: pre-grouting permeability; and the construction parameters include: grouting pressure, borehole sequence, and elevation parameters.

[0045] Furthermore, obtaining the grouting effect prediction results includes: inputting environmental parameters and construction parameters into a random forest model, a gradient boosting tree model, and an AdaBoost model to obtain the first prediction result; the random forest model is a random forest model ensembled using decision trees as base learners and the gradient boosting tree model is a gradient boosting tree model ensembled using decision trees as base learners and the boosting method; the first prediction result is input into an ANN model as input data to obtain the grouting effect prediction result; the ANN model is an ANN model ensembled using base learners and the stacking method.

[0046] Specifically, the grouting effect prediction models based on ensemble learning include a random forest model that uses decision trees as base learners and employs the Bagging method, a gradient boosting tree model and an AdaBoost model that use decision trees as base learners and employ the Boosting method, and an ANN model that uses the aforementioned base learners and employs the Stacking method. The dataset is randomly divided into a training set and a validation set according to a certain ratio, and the final result is predicted using the validation set.

[0047] The method for constructing the dataset involves collecting historical data on geological parameters, hydrological parameters, and actual construction parameters. Geological parameters include the surrounding rock grade, which quantifies the stability and complexity of geological conditions; joint density, which affects the permeability and stability of the rock mass; and alteration zone density, which reflects the spatial distribution of the degree of rock alteration. Hydrological parameters include the pre-grouting permeability rate, which measures the permeability of the rock and soil before grouting. Actual construction parameters include the hole sequence and grouting pressure for each grouting hole, with unit ash consumption used as an indicator for evaluating the grouting effect.

[0048] Furthermore, the simulated annealing algorithm is used to synchronously optimize the trained model parameters, including: Step 1, using the model parameters as the initial solution and setting the initial temperature and cooling coefficient; Step 2, updating the initial temperature according to the cooling plan and randomly selecting a new solution in the neighborhood of the initial solution; Step 3, calculating the fitness difference of the new solution, judging the fitness difference, if the fitness difference is less than the preset value, accepting the new solution, if the fitness difference is greater than the preset value, accepting the new solution according to the Metropolis criterion; Step 4, using the new solution as the current solution, repeating steps 1-3 until the updated temperature is lower than the target threshold or the preset number of iterations is reached, to obtain the optimal solution.

[0049] Furthermore, the model parameters include: the learning rate and number of weak learners for the AdaBoost model, the learning rate, maximum tree depth, number of trees, and minimum number of samples per leaf node for the gradient boosting tree model, the number of trees, maximum tree depth, and minimum number of samples per leaf node for the random forest model, and the neuron weights, number of hidden layers, and activation function for the ANN model.

[0050] Specifically:

[0051] like Figure 2 As shown, the method for synchronously optimizing model parameters using the simulated annealing algorithm includes the following steps:

[0052] Initialization: Select an initial solution i and set the initial temperature T and cooling coefficient α (where 0 < α < 1);

[0053] Temperature update formula: The temperature decrease typically follows a predetermined cooling schedule, such as linear or exponential decrease. T k+1 =α·T k T k It is the temperature of the k-th iteration, T k+1 It is the temperature of the (k+1)th iteration;

[0054] Neighborhood solution generation: Randomly select a new solution i' within the neighborhood of the current solution i;

[0055] Fitness calculation: Calculate the fitness difference Δf between the new solution i' and the current solution i, which is f(i') - f(i).

[0056] Acceptance criteria: If Δf < 0, then accept the new solution i'; if Δf > 0, then accept the new solution according to the Metropolis criterion.

[0057]

[0058] Where P(Δf) is the new solution accepted according to the Metropolis criterion, and T is the temperature. In the simulated annealing algorithm, temperature is a control parameter used to determine the probability of accepting a new solution; e is the base of the natural logarithm, approximately equal to 2.71828. In the Metropolis criterion, e is used to calculate the probability formula for accepting a new solution.

[0059] Iterative update: Set i = i', that is, update the current solution with the new solution;

[0060] Termination condition: If T k The algorithm terminates if the number of iterations falls below a certain threshold or reaches the maximum number of iterations.

[0061] Output: Returns the current solution i as the approximate optimal solution.

[0062] The model parameters optimized using the simulated annealing optimization algorithm include: the learning rate and number of weak learners for the AdaBoost model; the learning rate, maximum tree depth, number of trees, and minimum number of samples per leaf node for the gradient boosting tree model; the number of trees, maximum tree depth, and minimum number of samples per leaf node for the random forest model; and the neuron weights, number of hidden layers, and activation function for the ANN model.

[0063] Figure 2 Chinese: X 1-n Let Y1 be the set of input variables, and X be the set of output variables. 1-n Train Y1 is the input variable in the training set. Train For the output in the training set, X 1-n Test For the input variables in the validation set, Y1 Test To validate the outputs in the set, Bagging is a type of ensemble learning, Boosting is a type of ensemble learning, Stacking is a type of ensemble learning, SA is a type of heuristic search algorithm, and M... 1-n Several models with different parameters, M best: Optimal model parameters, decision tree is a basic regression method, AdaBoost is a type of boosting algorithm, random forest is a type of bagging algorithm, gradient boosting tree is a type of boosting algorithm, ANN is a type of neural network mathematical model algorithm, P1 is the output prediction value obtained by the AdaBoost model based on the training set or validation set, P2 is the output prediction value obtained by the random forest model based on the training set or validation set, and P3 is the output prediction value obtained by the gradient boosting tree model based on the training set or validation set.

[0064] Among them, Bagging is a type of ensemble learning that generates multiple subsets of the original dataset through multiple random samplings with replacement, trains a base learner on each subset, and finally averages or votes the predictions of these base learners. Boosting is another type of ensemble learning that sequentially trains multiple base learners, each focusing on samples misclassified by the previous learner, and finally combines the results of multiple base learners in a weighted manner. Stacking is a meta-learning method that trains multiple base learners and uses a meta-learner to combine their predictions. Simulated Annealing is a heuristic search algorithm used to find the global optimum in a large search space. Decision Trees are a basic regression method that predicts the value of a target variable by learning simple decision rules. AdaBoost is a type of boosting algorithm that iteratively trains decision trees and adjusts sample weights to focus on misclassified samples, finally combining the predictions of these decision trees in a weighted manner into a strong learner. Random Forest: A type of Bagging algorithm that improves the model's prediction accuracy by constructing multiple decision trees and summing their predictions. Gradient Boosting Tree: A type of Boosting algorithm that iteratively trains decision trees, with each tree predicting the residual of the previous prediction, and then adding the predictions of the new tree to the model's final prediction, thereby gradually reducing the value of the loss function. ANN: A type of neural network mathematical model algorithm that learns and predicts complex patterns in data through forward propagation, the application of non-linear activation functions, and backpropagation to adjust weights.

[0065] Furthermore, the method also includes validating the synchronously optimized model using the coefficient of determination, root mean square error, mean square error, mean absolute error, and mean absolute percentage error.

[0066] Specifically for

[0067] The performance of the optimized ensemble learning model in predicting grouting volume was analyzed and compared with the unoptimized model to verify the effectiveness of the simulated annealing optimization algorithm. The prediction accuracy of different models, including the coefficient of determination (R²), was compared. 2 The optimal model is selected based on the following: root mean square error (RMSE), mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAE).

[0068]

[0069] Among them, R 2 The coefficient of determination is RMSE, MSE is mean square error, MAE is mean absolute error, and MAE is mean absolute percentage error.

[0070] This embodiment discloses a method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model, including:

[0071] The grouting effect prediction model based on ensemble learning consists of two layers: a random forest model using decision trees as base learners and employing the Bagging method; a gradient boosting tree model and an AdaBoost model using decision trees as base learners and employing the Boosting method; and an ANN model using the aforementioned base learners and employing the Stacking method. The first layer includes three ensemble base learners for training and validation; the second layer includes a meta-learner, the ANN artificial neural network. The dataset for the ANN model is derived from the prediction outputs of the three ensemble base learners.

[0072] Furthermore, the method for constructing the dataset includes: based on the seepage control characteristics of underground water-sealed caverns, two construction processes, namely pre-grouting and post-grouting, are constructed separately. Corresponding parameters such as surrounding rock grade, joint surface density, alteration zone density, pre-grouting permeability, grouting pressure, borehole sequence, and elevation are obtained through manual compilation and summarization of geological sketches and grouting construction records. The output parameter is the unit ash consumption. Specifically, for the pre-grouting construction process, the parameters selected are surrounding rock grade, joint surface density, alteration zone density, pre-grouting permeability, grouting pressure, borehole sequence, and elevation; for the post-grouting construction process, the parameters selected are surrounding rock grade, joint surface density, alteration zone density, pre-grouting permeability, grouting pressure, and borehole sequence. The two datasets for grouting construction are randomly divided into training and validation sets according to a determined ratio.

[0073] Furthermore, the performance of the optimized ensemble learning model in predicting grouting volume was analyzed and compared with the unoptimized model to verify the effectiveness of the simulated annealing optimization algorithm. The prediction accuracy of different models, including the coefficient of determination (R²), was compared. 2The optimal model is selected based on the following: root mean square error (RMSE), mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAE).

[0074]

[0075] Among them, R 2 The coefficient of determination is RMSE, MSE is mean square error, MAE is mean absolute error, and MAE is mean absolute percentage error.

[0076] Furthermore, MSE is adopted as the target loss function of the simulated annealing algorithm because it can directly measure the average squared error between the predicted and actual values, which is convenient for optimization and can effectively reflect the accuracy of model prediction. It is a commonly used loss function in regression problems.

[0077] Furthermore, the specific implementation steps of the method for synchronously optimizing model parameters using the simulated annealing algorithm include:

[0078] Step 1: Select an initial solution i and set the initial temperature T and cooling coefficient α (where 0 < α < 1);

[0079] Step 2: Develop a cooling plan and set the temperature update formula T. k+1 =α·T k T k It is the temperature of the k-th iteration, T k+1 It is the temperature of the (k+1)th iteration;

[0080] Step 3: Randomly select a new solution i' within the neighborhood of the current solution i;

[0081] Step 4: Calculate the MSE values ​​of the current solution i and the new solution i', and calculate the fitness difference Δf = f(i') - f(i) between the new solution i' and the current solution i;

[0082] Step 5: Compare the fitness of the new solution and the current solution. If the new solution is better, accept it as the current solution. If the new solution is worse, there is still a certain probability of accepting it; this probability decreases as the "temperature" decreases. The acceptance principle is as follows: if Δf < 0, accept the new solution i'; if Δf > 0, accept the new solution according to the Metropolis criterion.

[0083]

[0084] Step 6: Perform iterative updates, setting i = i', that is, updating the current solution with the new solution;

[0085] Step 7, set the termination condition, if T k The algorithm terminates if the number of iterations falls below a certain threshold or reaches the maximum number of iterations.

[0086] Step 8: Output the result and return the current solution i as the approximate optimal solution.

[0087] Furthermore, the model parameters optimized using the simulated annealing optimization algorithm include: the learning rate l1 and the number of weak learners n1 for the AdaBoost model, and the learning rate l2 and the maximum tree depth D for the gradient boosting tree model. max1 The number of trees n², and the minimum number of samples J for each leaf node. min1 The number of trees n3 and the maximum depth D of the random forest model max2 Minimum number of leaf node samples J min2 And the neuron weights w and the number of hidden layers n in the ANN model hid The activation function f. The optimization process based on the simulated annealing algorithm can be transformed into the following formula:

[0088]

[0089] In the formula, y i and f(x) i θ(φ) and φ) represent the actual unit ash consumption value and the final predicted unit ash consumption value, respectively; φ represents the set of parameters to be optimized; and H is the range of values ​​for the model parameters to be optimized. Thus, an ensemble learning-based grouting effect prediction model is constructed.

[0090] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model, characterized in that, include: The environmental and construction parameters of the underground water-sealed cavern are obtained, and the environmental and construction parameters are input into the grouting effect prediction model based on ensemble learning to obtain the grouting effect prediction results. The grouting effect prediction model is trained using a training set, and the parameters of the trained model are simultaneously optimized using a simulated annealing algorithm; the training set includes: geological parameters, hydrological parameters, and historical construction parameters; Obtaining the grouting effect prediction results includes: The environmental parameters and construction parameters are input into a random forest model, a gradient boosting tree model, and an AdaBoost model to obtain a first prediction result; the random forest model is a random forest model that integrates the decision tree as the base learner using the Bagging method, and the gradient boosting tree model is a gradient boosting tree model that integrates the decision tree as the base learner using the Boosting method. The first prediction result is used as input data to input the ANN model to obtain the grouting effect prediction result; the ANN model is an ANN model that integrates the base learners using the Stacking method. Simultaneous optimization of trained model parameters using simulated annealing includes: Step 1: Use the model parameters as the initial solution and set the initial temperature and cooling coefficient; Step 2: Update the initial temperature according to the cooling plan, and randomly select a new solution within the neighborhood of the initial solution; Step 3: Calculate the fitness difference of the new solution, and judge the fitness difference. If the fitness difference is less than a preset value, the new solution is accepted. If the fitness difference is greater than the preset value, the new solution is accepted according to the Metropolis criterion. Step 4: Take the new solution as the current solution and repeat steps 1-3 until the updated temperature is lower than the target threshold or the preset number of iterations is reached to obtain the optimal solution; The grouting effect prediction model includes: random forest model, gradient boosting tree model, AdaBoost model and ANN model; the random forest model, gradient boosting tree model and AdaBoost model are all connected to the ANN model.

2. The method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model according to claim 1, characterized in that, The geological parameters include: surrounding rock grade, joint surface density, and alteration zone density; The hydrological parameters include: pre-irrigation permeability; The construction parameters include: grouting pressure, hole sequence, and elevation parameters.

3. The method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model according to claim 1, characterized in that, The model parameters include: the learning rate and number of weak learners for the AdaBoost model; the learning rate, maximum tree depth, number of trees, and minimum number of samples per leaf node for the gradient boosting tree model; the number of trees, maximum tree depth, and minimum number of samples per leaf node for the random forest model; and the neuron weights, number of hidden layers, and activation function for the ANN model.

4. The method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model according to claim 1, characterized in that, Calculating the fitness difference of the new solution includes: Δf = f(i') - f(i); Where Δf is the fitness difference, f(i') is the fitness of the new solution, and f(i) is the fitness difference of the current solution.

5. The method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model according to claim 4, characterized in that, Accepting the new solution according to the Metropolis criterion includes: Where P(Δf) is the new solution accepted according to the Metropolis criterion, T is the temperature, and e is the base of the natural logarithm.

6. The method for predicting the grouting effect of underground water-sealed caverns based on an ensemble learning model according to claim 1, characterized in that, The method also includes: The synchronously optimized model was validated using the coefficient of determination, root mean square error, mean square error, mean absolute error, and mean absolute percentage error.

Images