Coal and gas outburst strength prediction method based on machine learning

Abstract

The invention provides a coal and gas outburst strength prediction method based on machine learning, and belongs to the technical field of mine coal and gas exploitation, and the method comprises the following steps: analyzing collected data, introducing a grey correlation analysis method to carry out dimension reduction to obtain main control factors influencing outburst, and carrying out extreme dimensionless method processing on the characteristic data to obtain a coal and gas outburst strength prediction result; and obtaining a required data set, and distributing a training set and a verification set according to the proportion. And configuring a learning process, training the model by using a training set, optimizing an SVM by using IPSO, and automatically iterating input sample data to obtain optimal parameters of machine learning so as to determine a prediction model of machine learning. And verifying the model, performing strength prediction on the coal and gas outburst sample on the verification set, comparing with an actual result, and determining the prediction precision and the prediction accuracy of the model. According to the method, the optimal parameter self-adaption problem of machine learning can be solved, so that the coal and gas outburst strength can be quickly and accurately predicted, and safe production of a mine is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of coal and gas mining in mines, and in particular to a method for predicting coal and gas outburst intensity based on machine learning. Background Art

[0002] As a mine accident with obvious dynamic effects and extremely destructive power, coal and gas outbursts pose a huge threat to coal mine safety production. They have always seriously threatened coal mine safety production and stable coal supply, and are a stumbling block in my country's coal mining process. In recent years, with the continuous increase in the intensity and depth of coal mining, the severity of coal and gas outbursts has become increasingly prominent. The coal seams in my country's coal mines are buried deep and have high gas content. Under such conditions, accurately predicting the danger of outbursts and taking corresponding gas control measures in a timely manner are a strong guarantee for mine safety production. Although researchers have conducted a large number of experiments and studies in the field of mine gas prevention and control, the dynamic principles and prediction and prevention methods of gas outburst accidents have not been thoroughly mastered. Therefore, in this context, finding an effective outburst intensity prediction method is of extremely important practical significance for mine outburst prevention and elimination work.

[0003] The nonlinear dynamic mechanism of outburst phenomenon shows that the relationship between the basic factors affecting outburst and outburst danger is a complex nonlinear mapping relationship that is difficult to describe by simple functions. Therefore, this paper takes a mining area as the research object, takes the comprehensive effect hypothesis of coal and gas outburst as the theoretical guidance, and obtains the processed data set on the basis of fully extracting the dominant factors affecting outburst. Using support vector machine and a particle swarm optimization algorithm with a nonlinear decreasing inertia weight strategy, a coal and gas outburst intensity prediction model is established, and a coal and gas outburst prediction method based on machine learning is proposed. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] Therefore, in response to the problems existing in the prior art, the purpose of the present invention is to provide a coal and gas outburst intensity prediction method based on machine learning, which can accurately predict the level, has stronger expression ability, and better mapping effect, thereby improving the accuracy of outburst intensity prediction and the practicality of the model.

[0006] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:

[0007] A method for predicting coal and gas outburst intensity based on machine learning, comprising the following steps:

[0008] Step 1: Data collection and preprocessing, using the grey correlation analysis method to select the dominant factors of multiple factors, obtain the prediction indicators of coal and gas outburst, define the data set and allocate the training set and validation set according to the ratio, and standardize the data;

[0009] Step 2: Feature extraction: define the training set and test set according to the data set, map the input to the target, and determine the level characteristics of the highlight intensity index;

[0010] Step 3: Configure the learning process, train the model with the training set, select IPSO to optimize SVM and the indicators that need to be monitored, iterate the input sample data, automatically optimize and select the parameter combination, and obtain the best parameters for machine learning, thereby determining the prediction model of machine learning;

[0011] Step 4: Train the model, input samples and call the model's nonlinear inertia weight reduction strategy method to iterate on the training data to train and optimize the model;

[0012] Step 5: Verify the model, predict the strength of coal and gas outburst samples on the verification set, and compare with the actual results to determine the prediction accuracy of the model;

[0013] As a method for predicting the intensity of coal and gas outburst based on machine learning described in the present invention, the content includes: using a method combining grey correlation analysis method and entropy weight method to select the main factors affecting gas outburst and establish an intensity classification prediction model indicator system.

[0014] As a method for predicting coal and gas outburst intensity based on machine learning described in the present invention, the content includes: using an improved particle swarm optimization algorithm as the optimizer of the model to optimize the support vector machine to accelerate the learning speed of the algorithm and realize automatic optimization and selection of parameters, improve the global search capability and enhance the ability to find the optimal solution locally.

[0015] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0016] In an embodiment of the present invention, it is proposed to use an optimized support vector machine SVM to establish a coal and gas outburst intensity prediction model, and a coal and gas outburst prediction method based on machine learning is proposed. By introducing a gray correlation analysis method in parameter extraction from a large amount of historical data, the dimension is optimized, and the influencing factors that dominate the entire outburst prediction model system are found. Specific data of a mine are selected, a data set is defined, and a training set and a validation set are allocated according to a ratio. A model is established, and a particle swarm optimization algorithm (IPSO) with a nonlinear decreasing inertia weight strategy is proposed to optimize the support vector machine to obtain an IPSO-SVM prediction model. Compared with traditional coal and gas outburst prediction methods and shallow neural network prediction methods, the learning speed of the algorithm is accelerated, the expression ability is stronger, and the mapping effect is better, thereby being able to improve the accuracy of outburst intensity prediction and ensure safe and reliable operation of mine mining. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0018] Figure 1 It is a roadmap of the steps of the present invention;

[0019] Figure 2 A flow chart of selecting prediction indicators for the present invention;

[0020] Figure 3 Improved particle swarm optimization algorithm flow chart for the present invention;

[0021] Figure 4 A prediction model structure diagram is constructed for the present invention. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0023] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the present invention is described in detail with reference to the schematic diagram. When describing the embodiments of the present invention in detail, program codes representing algorithms and related formulas will be shown for ease of explanation.

[0025] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0026] Example 1

[0027] The outburst data of typical outburst mines are selected as samples for outburst intensity prediction, and the prediction of outburst intensity is mainly achieved through the following steps.

[0028] (1) Select indicators and quantify them;

[0029] According to the comprehensive hypothesis theory of gas outburst, combined with gas geology and gas outburst prediction, 12 parameters were selected from three aspects: geological structure, coal body structure and gas as influencing factor indicators of outburst accidents (Table 1).

[0030] Table 1 Classification results of coal and gas outburst accidents

[0031]

[0032] According to the data in Table 1, the factors affecting outbursts are divided into 6 primary factors and 12 secondary factors, which are the classification results of more than 100 outburst accidents. It can be seen that among the primary factors, geological structure, coal seam occurrence and mining are ranked in the top three, while among the secondary factors, faults, folds and coal mining technology are ranked in the top three.

[0033] (2) The measured data were subjected to grey correlation analysis, extreme value processing, and arranged in order of correlation degree. It can be concluded that the coal and gas outburst prediction indicators are: coal solidity coefficient, initial velocity of gas outburst in the borehole, initial velocity of gas release, gas desorption index of drill cuttings, and maximum amount of drill cuttings.

[0034] (3) Numerical processing of indicators. Before the prediction model is input, the input data is vectorized and classified into four categories according to convention: 1 indicates no danger of a sudden explosion and the intensity of the sudden explosion is none; 2 indicates general danger of a sudden explosion and the intensity of the sudden explosion is small; 3 indicates medium danger of a sudden explosion and the intensity of the sudden explosion is medium; 4 indicates that the danger of a sudden explosion is severe and the intensity of the sudden explosion is large.

[0035] Table 2 Training samples

[0036]

[0037]

[0038] (4) The particle swarm algorithm is used to optimize the support vector machine. The optimization is mainly used to measure the error between the predicted value and the target value. Considering the multi-classification characteristics of the model, a nonlinear decreasing inertia weight strategy is introduced to train and predict the salient, improve the global search capability and enhance the ability to find the optimal solution locally, so as to speed up the learning speed of the algorithm.

[0039] (5) Train the prediction model.

[0040] The training set data is processed by feature standardization and then input into the model for training. After automatic iteration, the accuracy of the model for the training set reaches 1, and the prediction model meets the accuracy requirements for coal and gas outburst prediction.

[0041] (6) Prediction results.

[0042] In order to verify the effectiveness of the prediction model, 30 examples of outstanding accidents were selected from the scene to test the model. The first 20 groups were used as training samples, and the last 10 groups were used as test samples. The quantitative results of the prediction parameters of the outstanding examples are shown in Table 3.

[0043] Table 3 Verification examples

[0044]

[0045] (7) Example verification.

[0046] After comparing with the validation set, the accuracy rate was 100%, which met the requirements and verified the accuracy and feasibility of the model.

[0047] Although the present invention has been described above with reference to the implementation steps, various improvements can be made to it and the data therein can be replaced with different sets of quantities without departing from the scope of the present invention. In particular, the stability of the model was tested, and the test results showed that it was highly stable. Therefore, the present invention has strong robustness and generalization, proving the feasibility of the model and providing technical support for subsequent safety predictions in mines.

Claims

1. A method for predicting coal and gas outburst intensity based on machine learning, characterized in that: The steps include: Step 1: Data collection and preprocessing, screening out the dominant factors that have an impact on multiple factors, obtaining the prediction indicators of coal and gas outburst, defining the data set and allocating the training set and validation set according to the ratio, and standardizing the data; Step 2: Feature extraction: define the training set and test set according to the data set, map the input to the target, and determine the level characteristics of the highlight intensity index; Step 3: Configure the learning process, train the model with the training set, select IPSO to optimize SVM and the indicators that need to be monitored, iterate the input sample data, automatically optimize and select the parameter combination, and obtain the best parameters for machine learning, thereby determining the prediction model of machine learning; Step 4: Train the model, input samples and call the model's nonlinear inertia weight reduction strategy method to iterate on the training data to train and optimize the model; Step 5: Verify the model, make strength predictions for coal and gas outburst samples on the validation set, and compare them with the actual results to determine the prediction accuracy of the model.

2. The method for predicting coal and gas outburst intensity based on machine learning according to claim 1, characterized in that: In step one, a combination of grey correlation analysis and entropy weight method is used to sort by correlation degree, and the factors that contribute most to coal and gas outburst are selected, that is, the factors ranked at the top, which are the main factors affecting gas outburst, to obtain the intensity classification prediction model index system.

3. The method for predicting coal and gas outburst intensity based on machine learning according to claim 1, characterized in that: In step three, an improved particle swarm optimization algorithm is used as the optimizer of the model to optimize the support vector machine, speed up the learning speed of the algorithm, realize automatic optimization and selection of parameters, improve the global search capability, and enhance the local search for the optimal solution.

4. The method for predicting coal and gas outburst intensity based on machine learning according to claim 1, characterized in that: In step 4, the nonlinear inertia weight reduction strategy method is selected to act on the input data in the training model for self-iteration.

Images