Coal mine underground mine pressure comprehensive monitoring method and system based on Internet of Things

Abstract

The invention relates to the technical field of mine pressure monitoring, in particular to a coal mine underground mine pressure comprehensive monitoring method and system based on the Internet of Things, and the method comprises the steps: collecting coal mine underground multi-source heterogeneous mine pressure data through a distributed sensor network; wherein the sensor network adopts a wireless communication protocol with low power consumption and high bandwidth; constructing a mine pressure data set based on the multi-source heterogeneous mine pressure data; key features of the mine pressure data set are extracted through a principal component analysis method, and a feature vector set is obtained; based on the feature vector set, constructing a mine pressure prediction model based on machine learning; and predicting the mine pressure data in real time by using the mine pressure prediction model, and judging whether the current mine pressure state is abnormal or not. Through a series of data processing, model prediction and early warning mechanisms, the mine safety management efficiency and accuracy can be greatly improved, and powerful guarantee is provided for the life safety of miners.

Description

Technical Field

[0001] The present invention relates to the technical field of mine pressure monitoring, and in particular to a method and system for comprehensive monitoring of mine pressure in underground coal mines based on the Internet of Things. Background Art

[0002] When conducting comprehensive monitoring of mine pressure in coal mines, data transmission and processing face some technical difficulties. First, the underground environment is harsh, with high temperature, high humidity, and a lot of dust, which places high demands on the performance and stability of sensors. Secondly, the underground space is small, wiring is difficult, and traditional wired transmission methods are not easy to implement. Furthermore, the mine pressure data is collected at a high frequency and with a large amount of data, which also poses challenges to the bandwidth and power consumption of wireless transmission.

[0003] In terms of data processing and analysis, mine pressure data has the characteristics of multi-source heterogeneity. The data formats and semantics collected by different types of sensors are different. How to effectively fuse data is a major problem. In addition, mine pressure data often contains a lot of redundancy and noise. How to clean the data and extract features is also a thorny issue. The geological environment reflected by the mine pressure data is complex and changeable. The change of mine pressure is affected by many factors. Therefore, when establishing a mine pressure prediction model, it is necessary to consider a variety of influencing factors and comprehensively analyze the law of mine pressure changes to improve the accuracy of the prediction. Finally, mine safety monitoring has high real-time requirements. How to quickly detect abnormal situations in massive data and issue timely warnings is also a technical problem that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to propose a method and system for comprehensive monitoring of underground coal mine pressure based on the Internet of Things to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A comprehensive monitoring method for underground coal mine pressure based on the Internet of Things, comprising:

[0007] Collect multi-source heterogeneous mine pressure data in coal mines through distributed sensor networks; wherein the sensor networks use low-power and high-bandwidth wireless communication protocols;

[0008] Based on the multi-source heterogeneous mine pressure data, construct a mine pressure data set;

[0009] For the mine pressure data set, a principal component analysis method is used to extract key features and obtain a feature vector set;

[0010] Based on the feature vector set, construct a mine pressure prediction model based on machine learning;

[0011] The mine pressure prediction model is used to make real-time predictions on the mine pressure data to determine whether the current mine pressure status is abnormal.

[0012] Optionally, based on the multi-source heterogeneous mine pressure data, constructing a mine pressure data set includes:

[0013] Using a data fusion algorithm, the formats of the mine pressure data from different sources and in different formats in the multi-source heterogeneous mine pressure data are unified and semantically aligned;

[0014] Through data cleaning and data conversion methods, the semantically aligned mine pressure data is denoised and missing value filled;

[0015] For the filled mine pressure data, semantic mapping and association between different data sources are performed;

[0016] The associated data are fused using a data fusion algorithm to generate a comprehensive mine pressure data set.

[0017] Optionally, for the mine pressure data set, a principal component analysis method is used to extract key features, and obtaining a feature vector set includes:

[0018] For the mine pressure data set, the principal component analysis method is used to extract the main characteristic components of the data set through eigenvalue decomposition and eigenvector calculation;

[0019] According to the results of principal component analysis, the first k principal components with eigenvalues ​​greater than the preset threshold are selected as the key features of the data set to obtain the mine pressure feature subset after dimensionality reduction;

[0020] Normalize the mine pressure feature subset, scale the value range of each feature to the interval [0, 1], and obtain the normalized mine pressure feature vector;

[0021] The K-means clustering algorithm is used to divide the normalized mine pressure feature vector into several clusters, and the feature vector in each cluster has a feature attribute with a preset similarity;

[0022] For each cluster, the center point is calculated as the representative feature vector of the cluster to obtain a set of feature vectors of the mine pressure after dimension reduction, and each vector represents a typical combination of mine pressure features.

[0023] Optionally, based on the feature vector set, constructing a mine pressure prediction model based on machine learning includes:

[0024] Build an integrated model combining decision tree and neural network models;

[0025] The integrated model is trained using the feature vector set to obtain the mine pressure prediction model.

[0026] Optionally, using the feature vector set to train the integrated model includes:

[0027] Determine the target variables for the mine pressure prediction task;

[0028] According to the relationship between the feature vector and the target variable, the branch structure of the decision tree is constructed, the feature vector and the target variable are used to train the decision tree, and after the training is completed, the pruning operation is performed;

[0029] The neural network model is trained using feature vectors and target variables. The output of the model is calculated through forward propagation, and the weights and biases of the network model are updated through the back propagation algorithm. In each iteration, the prediction error of the network model is calculated according to the loss function, and the network model parameters are adjusted according to the error.

[0030] A weighted average method based on the performance of the validation set is used to assign different weights to the decision tree and the neural network, and the trained decision tree and neural network models are fused to obtain the mine pressure prediction model.

[0031] A comprehensive monitoring system for underground coal mine pressure based on the Internet of Things, the system comprising: a data acquisition module, a data set construction module, a feature extraction module, a model construction module and a mine pressure prediction module;

[0032] The data acquisition module is used to collect multi-source heterogeneous mine pressure data in coal mines through a distributed sensor network; wherein the sensor network adopts a low-power and high-bandwidth wireless communication protocol;

[0033] The data set construction module is used to construct a mine pressure data set based on the multi-source heterogeneous mine pressure data;

[0034] The feature extraction module is used to extract key features from the mine pressure data set using a principal component analysis method to obtain a feature vector set;

[0035] The model building module is used to build a mine pressure prediction model based on machine learning based on the feature vector set;

[0036] The mine pressure prediction module is used to use the mine pressure prediction model to make real-time predictions on the mine pressure data and determine whether the current mine pressure state is abnormal.

[0037] Optionally, the data set construction module constructs a mine pressure data set based on the multi-source heterogeneous mine pressure data, including:

[0038] Using a data fusion algorithm, the formats of the mine pressure data from different sources and in different formats in the multi-source heterogeneous mine pressure data are unified and semantically aligned;

[0039] Through data cleaning and data conversion methods, the semantically aligned mine pressure data is denoised and missing value filled;

[0040] For the filled mine pressure data, semantic mapping and association between different data sources are performed;

[0041] The associated data are fused using a data fusion algorithm to generate a comprehensive mine pressure data set.

[0042] Optionally, the feature extraction module extracts key features from the mine pressure data set using a principal component analysis method, and obtaining a feature vector set includes:

[0043] For the mine pressure data set, the principal component analysis method is used to extract the main characteristic components of the data set through eigenvalue decomposition and eigenvector calculation;

[0044] According to the results of principal component analysis, the first k principal components with eigenvalues ​​greater than the preset threshold are selected as the key features of the data set to obtain the mine pressure feature subset after dimensionality reduction;

[0045] Normalize the mine pressure feature subset, scale the value range of each feature to the interval [0, 1], and obtain the normalized mine pressure feature vector;

[0046] The K-means clustering algorithm is used to divide the normalized mine pressure feature vector into several clusters, and the feature vector in each cluster has a feature attribute with a preset similarity;

[0047] For each cluster, the center point is calculated as the representative feature vector of the cluster to obtain a set of feature vectors of the mine pressure after dimension reduction, and each vector represents a typical combination of mine pressure features.

[0048] Optionally, the model building module builds a mine pressure prediction model based on machine learning based on the feature vector set, including:

[0049] Construct an integrated model combining decision tree and neural network models

[0050] The integrated model is trained using the feature vector set to obtain the mine pressure prediction model.

[0051] The beneficial effects of the present invention are:

[0052] The present invention discloses a method and system for comprehensive monitoring of underground coal mine pressure based on the Internet of Things. The method collects underground mine pressure data in real time through a distributed high-performance sensor network, solves the wiring difficulty problem by using a low-power and high-bandwidth wireless communication protocol, and efficiently transmits the data to a ground monitoring center. For multi-source heterogeneous data, a fusion algorithm is used for unified processing, and key features are extracted through principal component analysis. Based on this, a machine learning prediction model is constructed to achieve accurate prediction of mine pressure changes. The present invention continuously acquires mine pressure data and inputs it into a model for prediction, and triggers an early warning mechanism when the prediction result exceeds a threshold. This method significantly improves the accuracy and real-time performance of mine pressure monitoring, and provides a strong guarantee for safe production in mines. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. 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 paying creative labor.

[0054] Figure 1 The present invention is a flowchart of a method for comprehensive monitoring of underground coal mine pressure based on the Internet of Things. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] like Figure 1 As shown, this embodiment proposes a comprehensive monitoring method for underground coal mine pressure based on the Internet of Things, including:

[0058] Collect multi-source and heterogeneous mine pressure data in coal mines;

[0059] Construct a mine pressure dataset based on multi-source heterogeneous mine pressure data;

[0060] For the mine pressure data set, the principal component analysis method is used to extract key features and obtain the feature vector set;

[0061] Based on the feature vector set, a mine pressure prediction model based on machine learning is constructed;

[0062] The mine pressure prediction model is used to make real-time predictions on the mine pressure data to determine whether the current mine pressure status is abnormal.

[0063] This embodiment collects underground mine pressure data in real time through a distributed high-performance sensor network, uses a low-power and high-bandwidth wireless communication protocol to solve the wiring difficulty problem, and efficiently transmits data to the ground monitoring center. For multi-source heterogeneous data, a fusion algorithm is used for unified processing, and key features are extracted through principal component analysis. Based on this, a machine learning prediction model is constructed to achieve accurate prediction of mine pressure changes. The present invention will continuously acquire data and input it into the model for prediction, and trigger an early warning mechanism when the prediction result exceeds the threshold. This method significantly improves the accuracy and real-time performance of mine pressure monitoring, providing a strong guarantee for mine safety production.

[0064] Specifically, in this embodiment, the specific method of collecting multi-source heterogeneous mine pressure data in the coal mine is as follows:

[0065] The selection of mining sensors is crucial for underground environmental monitoring. According to the environmental characteristics of key areas underground, mining sensors with high temperature resistance, dust and moisture resistance are selected to ensure that the sensors can work stably in harsh environments. Distributed deployment is adopted to reasonably arrange sensor nodes in key areas underground to form a comprehensive sensor network to achieve all-round monitoring of mine pressure data. Sensor nodes transmit data through wireless communication technology and upload the collected mine pressure data to the ground monitoring center in real time to ensure the timeliness and reliability of the data. A data processing platform is established in the ground monitoring center to pre-process the received mine pressure data, including data cleaning, denoising and normalization operations to improve data quality. High temperature resistant sensors can work stably in environments up to 200°C, such as thermocouple temperature sensors. Dustproof sensors are designed with a sealed shell with a protection level of up to IP67, which effectively blocks dust intrusion. Moisture-proof sensors are treated with special coatings and can still operate normally at a relative humidity of 100%. Distributed deployment can achieve full coverage. In key areas such as coal mining faces and tunnel intersections, a sensor node is arranged every 50 meters to form a grid-like monitoring network. Wireless communication technologies such as ZigBee have low power consumption and self-organizing network characteristics, which are suitable for complex underground environments. Sensor nodes transmit data to aggregation nodes through multi-hop routing, and then transmit to the ground via optical fiber networks to ensure real-time performance.

[0066] Furthermore, based on multi-source heterogeneous mine pressure data, the mine pressure dataset is constructed including:

[0067] Adopt data fusion algorithm to unify the format and semantically align the mine pressure data from different sources and formats in multi-source heterogeneous mine pressure data;

[0068] Through data cleaning and data conversion methods, the semantically aligned mine pressure data is denoised and missing value filled;

[0069] For the filled mine pressure data, semantic mapping and association between different data sources are performed;

[0070] The associated data are fused using a data fusion algorithm to generate a comprehensive mine pressure data set.

[0071] Specifically, in this embodiment, the processing of multi-source heterogeneous mine pressure data is a key link in mine safety monitoring. Data fusion algorithms play an important role in the preprocessing stage. For example, weighted average fusion can integrate the mine pressure data collected by different sensors to improve data reliability. For example, a mine is equipped with stress sensors and displacement sensors, and a more accurate mine pressure estimate can be obtained through weighted average fusion. Data cleaning is a necessary step to improve data quality. For outliers in mine pressure data, median filtering and other methods can be used for denoising. Missing value filling can use interpolation methods such as linear interpolation or spline interpolation to ensure data continuity. These processes can significantly improve the accuracy of subsequent analysis. Semantic alignment is the key to achieving multi-source data integration. By establishing an ontology model, mine pressure data from different sources can be mapped to a unified semantic framework. For example, different expressions such as "roof subsidence" and "bottom plate uplift" are unified under the concept of "surrounding rock deformation" to facilitate comprehensive analysis. Feature extraction and selection are the basis of data analysis. For mine pressure data, time domain features such as mean and variance, and frequency domain features such as main frequency components can be extracted.

[0072] Furthermore, the principal component analysis method is used to extract key features from the mine pressure data set, and the feature vector set obtained includes:

[0073] For the mine pressure data set, the principal component analysis method is used to extract the main characteristic components of the data set through eigenvalue decomposition and eigenvector calculation;

[0074] According to the results of principal component analysis, the first k principal components with eigenvalues ​​greater than the preset threshold are selected as the key features of the data set to obtain the mine pressure feature subset after dimensionality reduction;

[0075] Normalize the mine pressure feature subset, scale the value range of each feature to the interval [0, 1], and obtain the normalized mine pressure feature vector;

[0076] The K-means clustering algorithm is used to divide the normalized mine pressure feature vector into several clusters, and the feature vector in each cluster has a feature attribute with a preset similarity;

[0077] For each cluster, the center point is calculated as the representative feature vector of the cluster to obtain a set of feature vectors of the mine pressure after dimension reduction, and each vector represents a typical combination of mine pressure features.

[0078] Specifically, in this embodiment, principal component analysis is a commonly used dimensionality reduction method. By calculating the covariance matrix of the data set, eigenvalue decomposition is performed to obtain eigenvalues ​​and corresponding eigenvectors. The size of the eigenvalue reflects the importance of the feature in the data set. For example, by performing principal component analysis on the multidimensional mine pressure data of a mining area, it may be found that the eigenvalues ​​of the first three principal components are 5.2, 3.8 and 2.1, respectively, and the cumulative variance contribution rate reaches 85%. Selecting these three principal components as key features not only retains the main information of the data, but also greatly reduces the data dimension. Normalization is an important means to eliminate the influence of different feature dimensions. Taking mine pressure, temperature and humidity as examples, their numerical ranges vary greatly. By mapping each eigenvalue to the [0,1] interval, different features have the same weight in subsequent analysis. For example, the original mine pressure value of a certain measuring point is 20MPa, the maximum value is 30MPa, the minimum value is 10MPa, and the normalized value is (20-10) / (30-10)=0.5. K-means clustering algorithm is an unsupervised learning method used to discover the intrinsic structure in data. Assume that the normalized mine pressure feature vector is clustered and the number of clusters k is set to 3. The algorithm randomly selects 3 initial center points and then iterates the sample allocation and center point update. In the end, three typical combinations of mine pressure features may be obtained: low-pressure stable area, medium-pressure fluctuation area, and high-pressure risk area. The center point of each cluster represents the average state of the mine pressure feature of this type. The set of mine pressure feature vectors after dimensionality reduction is stored in the database.

[0079] Furthermore, based on the feature vector set, a mine pressure prediction model based on machine learning is constructed, including:

[0080] Build an integrated model combining decision tree and neural network models;

[0081] The integrated model is trained using the feature vector set to obtain the mine pressure prediction model.

[0082] Furthermore, using the feature vector set, training the integrated model includes:

[0083] Determine the target variables for the mine pressure prediction task;

[0084] According to the relationship between the feature vector and the target variable, the branch structure of the decision tree is constructed, the feature vector and the target variable are used to train the decision tree, and after the training is completed, the pruning operation is performed;

[0085] The neural network model is trained using feature vectors and target variables. The output of the model is calculated through forward propagation, and the weights and biases of the network model are updated through the back propagation algorithm. In each iteration, the prediction error of the network model is calculated according to the loss function, and the network model parameters are adjusted according to the error.

[0086] The weighted average method based on the performance of the validation set is used to assign different weights to the decision tree and neural network. The trained decision tree and neural network models are fused to obtain the mine pressure prediction model.

[0087] Specifically, in this embodiment, it is found that the decision tree performs better in dealing with sudden changes in mine pressure, while the neural network is more accurate in predicting long-term trends. Therefore, this embodiment considers combining the advantages of both to build an integrated model. The decision tree model and the neural network model are integrated using a weight distribution method.

[0088] Training the integrated model specifically includes:

[0089] Divide the set of mine pressure feature vectors extracted and normalized by principal component analysis into training set, validation set and test set. Usually, it can be divided according to a certain ratio, such as 70% for training set, 15% for validation set and 15% for test set. Make sure the data distribution in the training set, validation set and test set is as consistent as possible so that the model can learn and generalize better. Clarify the target variable of the mine pressure prediction task, such as the abnormal state of mine pressure (normal or abnormal) or the specific mine pressure value. According to the type of target variable (classification or regression), select the corresponding model output form.

[0090] The decision tree model is trained with the training set data, and the branch structure of the decision tree is constructed according to the relationship between the feature vector and the target variable. The model will find the optimal features and thresholds to maximize information gain or minimize Gini impurity by recursively splitting nodes. During the training process, pruning operations can be used to prevent the decision tree from overfitting. Pruning methods include pre-pruning and post-pruning. Pre-pruning is to stop splitting in advance during the tree growth process, and post-pruning is to cut off some unnecessary branches after the tree is fully grown.

[0091] The neural network model is trained using the training set data, the output of the model is calculated through forward propagation, and then the weights and biases of the network are updated through the backpropagation algorithm. In each iteration, the model calculates the prediction error based on the loss function (such as mean square error, cross entropy loss), and adjusts the network parameters based on the error. During the training process, optimization methods such as batch gradient descent, stochastic gradient descent, or mini-batch gradient descent can be used. At the same time, regularization techniques (such as L1 regularization and L2 regularization) can be used to prevent the model from overfitting.

[0092] Fusion of the prediction results of the decision tree model and the neural network model. The model weight fusion method can be used to assign different weights to the decision tree and the neural network according to the performance of the model on the validation set. For example, if the decision tree performs better in dealing with sudden changes in mine pressure, a higher weight can be assigned to the decision tree; and if the neural network is more accurate in predicting long-term trends, a higher weight can be assigned to the neural network.

[0093] Specifically, in this embodiment, the prediction results of the decision tree model and the neural network model are fused, and a weighted average method based on the performance of the validation set is adopted;

[0094] First, the model validation set is evaluated;

[0095] Calculate the performance indicators (such as accuracy, AUC, RMSE, etc.) of the decision tree (DT) and neural network (NN) on the validation set respectively. Let the indicator of DT be M DT , the index of NN is M NN ;

[0096] Secondly, weights are dynamically allocated based on the performance indicator ratio;

[0097]

[0098]

[0099] in, is the decision tree model weight, is the neural network model weight, is the performance indicator of the decision tree in the validation set, is the performance indicator of the neural network in the validation set;

[0100] Example: If the AUC of DT is 0.86 and the AUC of NN is 0.88, then the weight is =0.494, =0.506.

[0101] Finally, the final prediction result is a weighted average;

[0102]

[0103] in, is the final prediction result after fusion. is the prediction result of the decision tree model. is the prediction result of the neural network model.

[0104] The prediction results of the decision tree and the neural network are used as input features to train a fusion model (such as linear regression, logistic regression, etc.) to combine the prediction results of the two models. The fusion model will learn how to output the final mine pressure prediction results based on the input features and weights. During the fusion model training process, the performance of the model can be further improved by adjusting the model parameters and optimizing the algorithm. For example, cross-validation can be used to evaluate the performance of the model under different parameter settings and select the optimal parameter combination.

[0105] This embodiment also determines whether abnormal mine pressure is likely to occur based on the predicted trend of mine pressure changes. If the predicted mine pressure exceeds the preset warning threshold, the early warning mechanism is triggered to notify relevant personnel to take measures in advance. Continuously collect mine pressure data, regularly retrain and optimize the mine pressure prediction model, improve the model's adaptability to unknown data, and ensure the accuracy and reliability of the prediction results. Comprehensively analyze the mine pressure prediction results and actual mine pressure changes, evaluate the performance of the prediction model, and mark data with large prediction deviations for iterative optimization of subsequent models. Establish a visualization display system for mine pressure prediction results, and intuitively display the historical changes and future trends of mine pressure in the form of curve graphs, bar charts, etc., so that managers can grasp the dynamics of mine pressure in a timely manner and make correct decisions.

[0106] This embodiment also proposes a comprehensive monitoring system for underground coal mine pressure based on the Internet of Things, including: a data acquisition module, a data set construction module, a feature extraction module, a model construction module and a mine pressure prediction module;

[0107] Data acquisition module, used to collect multi-source heterogeneous mine pressure data in coal mines;

[0108] A data set construction module is used to construct a mine pressure data set based on multi-source heterogeneous mine pressure data;

[0109] The feature extraction module is used to extract key features of the mine pressure data set using the principal component analysis method to obtain a feature vector set;

[0110] A model building module is used to build a mine pressure prediction model based on machine learning based on a feature vector set;

[0111] The mine pressure prediction module is used to use the mine pressure prediction model to make real-time predictions on the mine pressure data and determine whether the current mine pressure status is abnormal.

[0112] Furthermore, the dataset construction module constructs the mine pressure dataset based on multi-source heterogeneous mine pressure data, including:

[0113] Adopt data fusion algorithm to unify the format and semantically align the mine pressure data from different sources and formats in multi-source heterogeneous mine pressure data;

[0114] Through data cleaning and data conversion methods, the semantically aligned mine pressure data is denoised and missing value filled;

[0115] For the filled mine pressure data, semantic mapping and association between different data sources are performed;

[0116] The associated data are fused using a data fusion algorithm to generate a comprehensive mine pressure data set.

[0117] Furthermore, the feature extraction module uses the principal component analysis method to extract key features from the mine pressure data set, and obtains a feature vector set including:

[0118] For the mine pressure data set, the principal component analysis method is used to extract the main characteristic components of the data set through eigenvalue decomposition and eigenvector calculation;

[0119] According to the results of principal component analysis, the first k principal components with eigenvalues ​​greater than the preset threshold are selected as the key features of the data set to obtain the mine pressure feature subset after dimensionality reduction;

[0120] Normalize the mine pressure feature subset, scale the value range of each feature to the interval [0, 1], and obtain the normalized mine pressure feature vector;

[0121] The K-means clustering algorithm is used to divide the normalized mine pressure feature vector into several clusters, and the feature vector in each cluster has a feature attribute with a preset similarity;

[0122] For each cluster, the center point is calculated as the representative feature vector of the cluster to obtain a set of feature vectors of the mine pressure after dimension reduction, and each vector represents a typical combination of mine pressure features.

[0123] Furthermore, the model building module builds a mine pressure prediction model based on machine learning based on the feature vector set, including:

[0124] Build an integrated model combining decision tree and neural network models;

[0125] The integrated model is trained using the feature vector set to obtain the mine pressure prediction model.

[0126] Specifically, the implementation structure of the comprehensive underground coal mine pressure monitoring system of this embodiment is as follows:

[0127] Sensor deployment: a sensor node is arranged every 50 meters on the coal mining face to form a grid-like monitoring network;

[0128] Data acquisition module: sensor nodes transmit data to aggregation nodes through ZigBee protocol, and then transmit data to the ground monitoring center via optical fiber network;

[0129] Data construction module: Use weighted average fusion algorithm to fuse multi-source heterogeneous data, remove noise through median filtering, and fill missing values ​​through linear interpolation;

[0130] Feature extraction module: extract key features through principal component analysis, select the first three principal components, perform normalization, and divide them into three clusters using the K-means clustering algorithm;

[0131] Model building module: build an integrated model of decision trees and neural networks, use training set data for training, evaluate model performance through validation sets, and assign weights;

[0132] Mine pressure prediction module: Use the trained model to make real-time predictions on mine pressure data. When the prediction result exceeds the preset threshold, the early warning mechanism is triggered to notify relevant personnel.

[0133] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A comprehensive monitoring method for underground coal mine pressure based on the Internet of Things, characterized in that: include: Collect multi-source heterogeneous mine pressure data in coal mines through distributed sensor networks; wherein the sensor networks use low-power and high-bandwidth wireless communication protocols; Based on the multi-source heterogeneous mine pressure data, construct a mine pressure data set; For the mine pressure data set, a principal component analysis method is used to extract key features and obtain a feature vector set; Based on the feature vector set, construct a mine pressure prediction model based on machine learning; The mine pressure prediction model is used to make real-time predictions on the mine pressure data to determine whether the current mine pressure status is abnormal.

2. The method for comprehensive monitoring of underground coal mine pressure based on the Internet of Things according to claim 1 is characterized in that: Based on the multi-source heterogeneous mine pressure data, constructing a mine pressure data set includes: Using a data fusion algorithm, the formats of the mine pressure data from different sources and in different formats in the multi-source heterogeneous mine pressure data are unified and semantically aligned; Through data cleaning and data conversion methods, the semantically aligned mine pressure data is denoised and missing value filled; For the filled mine pressure data, semantic mapping and association between different data sources are performed; The associated data are fused using a data fusion algorithm to generate a comprehensive mine pressure data set.

3. The method for comprehensive monitoring of underground coal mine pressure based on the Internet of Things according to claim 1 is characterized in that: For the mine pressure data set, the principal component analysis method is used to extract key features, and the feature vector set obtained includes: For the mine pressure data set, the principal component analysis method is used to extract the main characteristic components of the data set through eigenvalue decomposition and eigenvector calculation; According to the results of principal component analysis, the first k principal components with eigenvalues ​​greater than the preset threshold are selected as the key features of the data set to obtain the mine pressure feature subset after dimensionality reduction; Normalize the mine pressure feature subset, scale the value range of each feature to the interval [0, 1], and obtain the normalized mine pressure feature vector; The K-means clustering algorithm is used to divide the normalized mine pressure feature vector into several clusters, and the feature vector in each cluster has a feature attribute with a preset similarity; For each cluster, the center point is calculated as the representative feature vector of the cluster to obtain a set of feature vectors of the mine pressure after dimension reduction, and each vector represents a typical combination of mine pressure features.

4. The method for comprehensive monitoring of underground coal mine pressure based on the Internet of Things according to claim 1 is characterized in that: Based on the feature vector set, constructing a mine pressure prediction model based on machine learning includes: Build an integrated model combining decision tree and neural network models; The integrated model is trained using the feature vector set to obtain the mine pressure prediction model.

5. The method for comprehensive monitoring of underground coal mine pressure based on the Internet of Things according to claim 4 is characterized in that: Using the feature vector set, training the integrated model includes: Determine the target variables for the mine pressure prediction task; According to the relationship between the feature vector and the target variable, the branch structure of the decision tree is constructed, the feature vector and the target variable are used to train the decision tree, and after the training is completed, the pruning operation is performed; The neural network model is trained using feature vectors and target variables. The output of the model is calculated through forward propagation, and the weights and biases of the network model are updated through the back propagation algorithm. In each iteration, the prediction error of the network model is calculated according to the loss function, and the network model parameters are adjusted according to the error. A weighted average method based on the performance of the validation set is used to assign different weights to the decision tree and the neural network, and the trained decision tree and neural network models are fused to obtain the mine pressure prediction model.

6. A comprehensive monitoring system for underground coal mine pressure based on the Internet of Things, characterized in that: Used to implement the method according to any one of claims 1 to 5, the system comprises: a data acquisition module, a data set construction module, a feature extraction module, a model construction module and a mine pressure prediction module; The data acquisition module is used to collect multi-source heterogeneous mine pressure data in coal mines through a distributed sensor network; wherein the sensor network adopts a low-power and high-bandwidth wireless communication protocol; The data set construction module is used to construct a mine pressure data set based on the multi-source heterogeneous mine pressure data; The feature extraction module is used to extract key features from the mine pressure data set using a principal component analysis method to obtain a feature vector set; The model building module is used to build a mine pressure prediction model based on machine learning based on the feature vector set; The mine pressure prediction module is used to use the mine pressure prediction model to make real-time predictions on the mine pressure data and determine whether the current mine pressure state is abnormal.

7. The comprehensive monitoring system for underground coal mine pressure based on the Internet of Things according to claim 6 is characterized in that: The data set construction module constructs a mine pressure data set based on the multi-source heterogeneous mine pressure data, including: Using a data fusion algorithm, the formats of the mine pressure data from different sources and in different formats in the multi-source heterogeneous mine pressure data are unified and semantically aligned; Through data cleaning and data conversion methods, the semantically aligned mine pressure data is denoised and missing value filled; For the filled mine pressure data, semantic mapping and association between different data sources are performed; The associated data are fused using a data fusion algorithm to generate a comprehensive mine pressure data set.

8. The comprehensive monitoring system for underground coal mine pressure based on the Internet of Things according to claim 6 is characterized in that: The feature extraction module extracts key features from the mine pressure data set using a principal component analysis method to obtain a feature vector set including: For the mine pressure data set, the principal component analysis method is used to extract the main characteristic components of the data set through eigenvalue decomposition and eigenvector calculation; According to the results of principal component analysis, the first k principal components with eigenvalues ​​greater than the preset threshold are selected as the key features of the data set to obtain the mine pressure feature subset after dimensionality reduction; Normalize the mine pressure feature subset, scale the value range of each feature to the interval [0, 1], and obtain the normalized mine pressure feature vector; The K-means clustering algorithm is used to divide the normalized mine pressure feature vector into several clusters, and the feature vector in each cluster has a feature attribute with a preset similarity; For each cluster, the center point is calculated as the representative feature vector of the cluster to obtain a set of feature vectors of the mine pressure after dimension reduction, and each vector represents a typical combination of mine pressure features.

9. The comprehensive monitoring system for underground coal mine pressure based on the Internet of Things according to claim 6 is characterized in that: The model building module builds a mine pressure prediction model based on machine learning based on the feature vector set, including: Build an integrated model combining decision tree and neural network models; The integrated model is trained using the feature vector set to obtain the mine pressure prediction model.

Images