Machine learning combined prediction method for coal seam overlying strata two-zone height

Abstract

The invention belongs to the technical field of mining overlying strata damage intelligent prediction, and particularly relates to a coal seam overlying strata two-zone height machine learning joint prediction method, which specifically comprises the following steps of: acquiring relevant parameters and actually measured data of multiple mining area overlying strata two-zone heights, and storing the relevant parameters and the actually measured data into a table; importing related data, constructing a prediction model of a multi-machine learning algorithm, and evaluating the accuracy of the model; the algorithm with high accuracy is screened, and hyper-parameter optimization is carried out; combining the optimized single models to construct a combined prediction model; according to the method, multiple machine learning algorithms are fused to construct the joint prediction model, the advantages of different algorithms in processing complex data are fully played, compared with a single machine learning algorithm, the complex relation between the height of the two zones of the overlying strata of the coal seam and each influence factor can be more accurately captured, and the prediction accuracy is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent prediction technology for mining-induced overburden damage, specifically a machine learning-based joint prediction method for the height of two zones of coal seam overburden. Background Technology

[0002] With the continuous rise in global energy demand, coal, as a key basic energy source, occupies a pivotal position in the world's energy landscape. Behind the large-scale development and utilization of coal resources, the safety and efficiency of coal mining have become key focuses for the industry's development. Existing research results indicate that, based on the different degrees of damage to the overlying rock mass in the mining area, it can be divided from top to bottom into a flexural subsidence zone, a fracture zone, and a caving zone. The two overlying rock zones are the water-conducting fracture zone and the caving zone. The water-conducting fracture zone is the sum of the caving zone and the fracture zone. Its height serves as a key indicator for measuring the impact of mining activities on the stability of the overlying strata, and is of vital importance to many aspects of coal mining, including roof control, water protection, gas migration, and surface subsidence. Accurate prediction of its height has always been one of the important issues in coal mining.

[0003] To predict the height of the "two zones" of coal seam overburden, many scholars have conducted extensive research using theoretical and empirical analysis, numerical simulation, physical simulation, and field monitoring, achieving numerous results and providing solid theoretical support for coal mine roof stability and water hazard prevention.

[0004] However, the aforementioned research methods often rely on simplified geological model assumptions, leading to significant errors in predictions when faced with complex lithological combinations and different mining methods. In recent years, with the rapid development of computer technology and the explosive growth of data, machine learning, as one of the core technologies in artificial intelligence, has demonstrated powerful data analysis and pattern recognition capabilities across various fields. Applying machine learning methods to the prediction of overburden "two zones" height can effectively handle nonlinear and complex data, analyze its intricate patterns, and provide new approaches and methods for accurately predicting the height of overburden "two zones." Currently, numerous machine learning methods have been applied to "two-zone" height prediction, but current research mainly focuses on predictions based on single machine learning methods, resulting in certain limitations in model accuracy. Summary of the Invention

[0005] The purpose of this invention is to provide a machine learning-based joint prediction method for the height of two zones of coal seam overburden, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a machine learning-based joint prediction method for the height of two zones of coal seam overburden, the method comprising the following specific steps: Step 1: Obtain relevant parameters and measured data on the height of the two overburden zones in multiple mining areas, and save them into a table; Step 2: Import relevant data, build a prediction model using multiple machine learning algorithms, and evaluate the model's accuracy; Step 3: Select the algorithms with higher accuracy and optimize their hyperparameters; Step 4: Combine the optimized individual models to construct a joint prediction model; Step 5: Open the Overburden Height Prediction System, register and log in to enter the model prediction interface of the Overburden Height Prediction System; Step 6: Input the parameters required for predicting the height of the two overburden zones; Step 7: Call the local joint prediction model, execute the prediction function, and obtain the predicted overburden height data for both zones.

[0007] Preferably, in step 1, the two overburden zones are the water-conducting fracture zone and the caving zone, and the influencing factors are six items in total: overburden lithology, mining height, coal seam dip angle, working face length, mining depth, and mining method.

[0008] Preferably, in step 2, the machine learning algorithm selected nine classic machine learning algorithms: linear regression, nearest neighbor algorithm, support vector machine, backpropagation neural network, random forest, gradient boosting, histogram gradient boosting, XGBoost and CatBoost, and their accuracy was evaluated.

[0009] Preferably, in step 4, the hyperparameter optimization of the model includes selecting algorithms with higher accuracy, optimizing their hyperparameters through multiple optimization algorithms, and using a weighted average method and a stacking ensemble method to jointly construct a joint prediction model by combining the optimized single algorithm models.

[0010] Preferably, in step 6, the prediction of the height of the two overburden zones is reflected in the local call to the joint prediction model, which inputs six relevant parameters: overburden lithology, mining height, coal seam dip angle, working face length, mining depth, and mining method, and can predict the height of the two overburden zones.

[0011] Compared with the prior art, the beneficial effects of the present invention are: 1) This invention integrates multiple machine learning algorithms to construct a joint prediction model, giving full play to the advantages of different algorithms in processing complex data. Compared with a single machine learning algorithm, it can more accurately capture the complex relationship between the height of the two overburden zones of coal seam and various influencing factors, and significantly improve the prediction accuracy. 2) By analyzing the importance of features, key features are selected, reducing the data dimensionality, lowering the computational complexity of the model, and improving the training efficiency and generalization ability of the model. 3) The prediction model provided by this invention is data-driven, does not require complex physical modeling, can adapt to coal seams with different geological conditions and mining processes, and has wide applicability. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating the steps of constructing the overburden two-zone height prediction system of the present invention. Figure 2 This is a screenshot of the login interface in this embodiment; Figure 3 Bar chart for evaluating multiple algorithm models for predicting the height of water-conducting fracture zones; Figure 4 Bar chart for evaluating multiple algorithm models for landslide zone height prediction; Figure 5 This is a diagram showing the parameter prediction interface of the overburden two-zone height prediction system; Figure 6 The image shows the predicted height of the two overlying zones after the prediction was performed. Detailed Implementation

[0013] 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.

[0014] Example: Please see Figure 1-6 The present invention provides a technical solution: A machine learning-based joint prediction method for the height of two zones in coal seam overburden, comprising the following specific steps: Step 1: Obtain relevant parameters and measured data on the height of the two overburden zones in multiple mining areas and store them in a table; this will provide a dataset when building a prediction model. The two overburden zones are the water-conducting fracture zone and the caving zone. The influencing factors are six items: overburden lithology, mining height, coal seam dip angle, working face slope length, mining depth, and mining method, as well as the height of the water-conducting fracture zone and the height of the caving zone, for a total of eight items. Step 2: Import relevant data, construct a prediction model using multiple machine learning algorithms, and evaluate the model's accuracy. Nine classic machine learning algorithms were selected: linear regression, nearest neighbor algorithm, support vector machine, backpropagation neural network, random forest, gradient boosting, histogram gradient boosting, XGBoost, and CatBoost. Their accuracy was then evaluated. Step 3: Select the algorithms with higher accuracy and perform hyperparameter optimization; use three hyperparameter optimization methods, grid search, random search, and Bayesian optimization, to optimize the hyperparameters of the selected algorithms, build a single algorithm optimization model, and compare the model performance before and after optimization. The optimized model has higher accuracy. Step 4: Combine the optimized single models to construct a joint prediction model; the hyperparameter optimization of the model includes selecting the algorithms with higher accuracy, optimizing the hyperparameters of the models through multiple optimization algorithms, and using the weighted average method and the Stacking ensemble method to combine the optimized single algorithm models to construct a joint prediction model. The Stacking ensemble method performs better in predicting the height of the collapse zone, while the weighted average method performs better in predicting the height of the water-conducting fracture zone.

[0015] Step 5: Open the Overburden Height Prediction System, register and log in to enter the model prediction interface of the Overburden Height Prediction System; Step 6: Input the parameters required for predicting the height of the two overburden zones; The prediction of the height of the two overburden zones is reflected in the local call of the joint prediction model. Input six relevant parameters, namely overburden lithology, mining height, coal seam dip angle, working face length, mining depth and mining method, and the height of the two overburden zones can be predicted. Step 7: Call the local joint prediction model, execute the prediction function, and obtain the predicted height data of the two overburden zones. Click the "Execute Prediction" button to predict the height of the water-conducting fracture zone and the caving zone, and obtain the predicted height of the two coal seam overburden zones.

[0016] For example, the table below shows some of the relevant characteristics and height data of the two overlying zones. After processing and analysis, the data is summarized into a .csv file for later use. Table 1. Height and Related Characteristics of the Two Overburden Zones (Partial) A machine learning model was built using a data table as the dataset. Nine classic machine learning algorithms were selected, including linear regression, nearest neighbor algorithm, support vector machine, backpropagation neural network, random forest, gradient boosting, histogram gradient boosting, XGBoost, and CatBoost. The coefficient of determination was used to evaluate the model, and the model performance is as follows: Figure 3 and Figure 4 The selected algorithms were optimized using three hyperparameter optimization methods: grid search, random search, and Bayesian optimization, to construct single-algorithm optimization models. A joint prediction model was then constructed from the optimized single models using weighted average and stacking ensemble methods, respectively. This invention enables the prediction of the height of the "two zones" of coal seam overburden. To use the overburden "two zones" height prediction system, open the system, register and log in, and enter the system's model prediction interface. Figure 5Within the software, select the overburden lithology, mining method, and input data such as mining height, coal seam dip angle, working face length, and mining depth; call the local joint prediction model, execute the prediction function, and obtain the predicted height data of the two overburden zones; click the "Execute Prediction" button to predict the height of the water-conducting fracture zone and caving zone, obtaining the predicted height of the two overburden zones of the coal seam, such as... Figure 6 .

[0017] This invention utilizes machine learning methods and the construction technology of a joint prediction model and software system based on PyCharm software to predict the height of the two overburden zones of coal seams. It has certain novelty, and the invention is simple to operate and easy to apply in practice, providing a new method and idea for accurate prediction of the height of the two overburden zones of coal seams.

[0018] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or basic characteristics. Therefore, the embodiments should be considered exemplary and non-limiting in all respects. The scope of the invention is defined by the appended claims rather than the foregoing description. Therefore, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0019] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A machine learning-based joint prediction method for the height of two zones of coal seam overburden, characterized in that, The method includes the following specific steps: Step 1: Obtain relevant parameters and measured data on the height of the two overburden zones in multiple mining areas, and save them into a table; Step 2: Import relevant data, build a prediction model using multiple machine learning algorithms, and evaluate the model's accuracy; Step 3: Select the algorithms with higher accuracy and optimize their hyperparameters; Step 4: Combine the optimized individual models to construct a joint prediction model; Step 5: Open the Overburden Height Prediction System, register and log in to enter the model prediction interface of the Overburden Height Prediction System; Step 6: Input the parameters required for predicting the height of the two overburden zones; Step 7: Call the local joint prediction model, execute the prediction function, and obtain the predicted overburden height data for both zones.

2. The machine learning joint prediction method for the height of two zones of coal seam overburden according to claim 1, characterized in that: In step 1, the two overburden zones are the water-conducting fracture zone and the caving zone, and the influencing factors are six items in total: overburden lithology, mining height, coal seam dip angle, working face length, mining depth, and mining method.

3. The machine learning joint prediction method for the height of two zones of coal seam overburden according to claim 1, characterized in that: In step 2, nine classic machine learning algorithms were selected: linear regression, nearest neighbor algorithm, support vector machine, backpropagation neural network, random forest, gradient boosting, histogram gradient boosting, XGBoost and CatBoost, and their accuracy was evaluated.

4. The machine learning joint prediction method for the height of two zones of coal seam overburden according to claim 1, characterized in that: In step 4, the hyperparameter optimization of the model includes selecting algorithms with higher accuracy, optimizing the hyperparameters of the model through multiple optimization algorithms, and using the weighted average method and the stacking ensemble method to jointly construct a joint prediction model by combining the optimized single algorithm models.

5. The machine learning joint prediction method for the height of two zones of coal seam overburden according to claim 1, characterized in that: In step 6, the prediction of the height of the two overburden zones is reflected in the local call to the joint prediction model. By inputting six relevant parameters, namely overburden lithology, mining height, coal seam dip angle, working face length, mining depth and mining method, the height of the two overburden zones can be predicted.

Images