A coal production optimization method and system based on data mining

Abstract

The application discloses a coal production optimization method and system based on data mining, and relates to the technical field of data mining.The method comprises the following steps: collecting multi-source safety monitoring data and historical accident data of a target coal production scene, and constructing a dynamic graph neural network model according to the historical accident data; iteratively predicting accidents according to the multi-source safety monitoring data, the dynamic graph neural network model and a production control scheme randomly selected in combination with a design working condition domain of the target scene, and performing safety boundary analysis based on a fitting method to obtain a real-time production control safety boundary; and controlling production parameters of the target coal production scene by taking the real-time production control safety boundary as a constraint. The application solves the technical problem of poor safety of coal production parameter control in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data mining, and particularly relates to a coal production optimization method and system based on data mining. BACKGROUND

[0002] The field of coal production safety monitoring has long relied on static early warning systems based on fixed thresholds and manual experience decision-making. Existing technologies usually collect single-dimensional data in real time by deploying sensor networks and set static safety thresholds to trigger alarms. However, such methods have significant drawbacks: first, static thresholds cannot adapt to the complex dynamic environment underground, resulting in high false alarm rates or missing critical risks; second, isolated analysis of multi-source data leads to inaccurate analysis results. In addition, existing technologies cannot dynamically adapt to changes in working conditions, making it difficult to balance production efficiency and safety assurance. SUMMARY

[0003] The present application provides a coal production optimization method and system based on data mining, which is used to solve the technical problem of poor safety of coal production parameter control in the prior art.

[0004] In view of the above problems, the present application provides a coal production optimization method and system based on data mining.

[0005] In the first aspect, the present application provides a coal production optimization method based on data mining, which comprises:

[0006] Collecting multi-source safety monitoring data and historical accident data of a target coal production scene, and constructing a dynamic graph neural network model according to the historical accident data;

[0007] Combining the multi-source safety monitoring data, iteratively predicting and analyzing a production control scheme through the dynamic graph neural network model to obtain an iterative prediction and analysis result, wherein the production control scheme is determined based on the design working condition domain of the target production scene;

[0008] According to the iterative prediction and analysis result and a preset safety boundary constraint, filtering a production control scheme that meets the safety boundary constraint as a set of safe production control schemes, wherein the safety boundary constraint is determined based on the working face gas emission, and the working face gas emission includes absolute emission and relative emission;

[0009] Performing fitting analysis on the set of safe production control schemes to generate a real-time production control safety boundary;

[0010] Controlling the production parameters of the target coal production scene with the real-time production control safety boundary as a constraint.

[0011] In a second aspect, the application provides a coal production optimization system based on data mining, comprising:

[0012] a model construction module configured to collect multi-source safety monitoring data and historical accident data of a target coal production scene, and construct a dynamic graph neural network model according to the historical accident data;

[0013] an iterative prediction module configured to perform iterative prediction analysis on a production control scheme based on the multi-source safety monitoring data and the dynamic graph neural network model, and obtain an iterative prediction analysis result, wherein the production control scheme is determined based on a design working condition domain of the target production scene;

[0014] a scheme screening module configured to screen a production control scheme that meets a preset safety boundary constraint as a set of safe production control schemes according to the iterative prediction analysis result and the safety boundary constraint, wherein the safety boundary constraint is determined based on a working face gas emission amount, and the working face gas emission amount includes an absolute emission amount and a relative emission amount;

[0015] a fitting analysis module configured to perform fitting analysis on the set of safe production control schemes, and generate a real-time production control safety boundary;

[0016] a production parameter control module configured to control production parameters of the target coal production scene with the real-time production control safety boundary as a constraint.

[0017] One or more technical solutions provided in the application have at least the following technical effects or advantages:

[0018] The application provides a coal production optimization method and system based on data mining, which significantly improves the dynamic adaptability and safety of production control parameters by fusing dynamic graph neural network modeling and multi-source safety data collaborative analysis. Compared with traditional methods, the technical solution provided in the application significantly breaks through the limitations of static threshold regulation, realizes multi-dimensional safety index spatiotemporal correlation mining and real-time collaborative optimization of production control parameters, and achieves the technical effect of improving the safety of coal production control parameters. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort.

[0020] Figure 1 A flowchart of a coal production optimization method based on data mining provided in the embodiments of the application is shown.

[0021] Figure 2 A structural schematic diagram of a coal production optimization system based on data mining provided by an embodiment of the present application.

[0022] In the drawings, the components represented by the respective reference numerals are described as follows:

[0023] The model construction module 100, the iterative prediction module 200, the scheme screening module 300, the fitting analysis module 400, and the production parameter control module 500. DETAILED DESCRIPTION

[0024] The present application provides a coal production optimization method and system based on data mining, which is used to solve the technical problem of poor safety of coal production parameter control in the prior art.

[0025] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0026] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or modules that are not clearly listed or inherent to the process, method, product or device.

[0027] Embodiment one, as shown in the present application provides a coal production optimization method based on data mining, wherein the method comprises: Figure 1

[0028] S10: Collecting multi-source safety monitoring data and historical accident data of a target coal production scene, and constructing a dynamic graph neural network model according to the historical accident data.

[0029] In coal production safety monitoring, the traditional method relies on a static model or independent analysis of single-source data, and cannot effectively contact the dynamic coupling relationship of multiple factors underground. The data collected by the prior art is often stored in isolation, and the historical accident records are only used for post-tracing, which leads to the fact that the correlation between monitoring points is ignored, the accident transmission mechanism such as the equipment failure chain and the stress migration chain cannot be expressed, and the risk prediction lags behind the actual production dynamics.

[0030] The step S10 in the method provided by the embodiment of the present application comprises:

[0031] ​Real-time collection of multiple types of safety monitoring data in the coal production process, generating the multiple-source safety monitoring data;

[0032] Synchronously acquiring the accident type, accident location and accident associated monitoring data recorded in the historical production cycle to constitute the historical accident data;

[0033] The multiple types of safety monitoring data at least include gas concentration data, dust concentration data, surrounding rock stress data, equipment operating state data and geological environment monitoring data.

[0034] The dynamic graph neural network model is constructed according to the historical accident data, including:

[0035] Mapping the safety monitoring points of the target coal production scene to nodes in the graph structure;

[0036] Based on the physical dependence relationship, geological distribution relationship and historical accident correlation relationship between the safety monitoring points, a directed connection edge between nodes is constructed and a corresponding edge weight matrix is generated to form the graph structure;

[0037] The dynamic graph neural network model is obtained by combining the graph structure with the time series monitoring data and historical production control data in the historical accident data for model training;

[0038] The dynamic graph neural network model is constructed according to the historical accident data, and before that, including:

[0039] The historical accident data is subjected to data enhancement processing, including data expansion, data smoothing and abnormal value correction;

[0040] Based on the historical accident data after data enhancement processing, an adversarial generative network model for accident data expansion is constructed, and synthetic accident data is generated through the adversarial generative network model;

[0041] The synthetic accident data is added to the historical accident data for training of the dynamic graph neural network model.

[0042] In the embodiments of the present application, multiple types of safety monitoring data in the coal production process are collected in real time, such as generating multiple-source safety monitoring data. The multiple types of safety monitoring data at least include: gas concentration data, such as oxygen concentration data collected using a gas sensor, in %; dust concentration data, such as dust particulate matter concentration collected using a dust sensor, in mg / m³; surrounding rock stress data, such as stress data of surrounding rock collected using a vibrating wire stress meter, in MPa; equipment operating state data, such as current parameters of motor equipment, in A; and geological environment monitoring data, such as rock thickness, in cm.

[0043] According to the production log in the coal production process, the recorded accident types in the historical production period are synchronously obtained, such as roof collapse accident; the accident position is obtained by collecting the specific accident coordinate latitude and longitude through GPS; the accident related monitoring data is obtained by collecting multiple types of safety monitoring data within 1 hour before the accident; the accident type, accident position and accident related monitoring data are integrated to form historical accident data.

[0044] The historical accident data is subjected to data enhancement processing, which includes data expansion, data smoothing and abnormal value correction. The data expansion can be subjected to linear difference method for data expansion and supplement, so that there is a data point at each time point; the data smoothing can be subjected to moving average method, and the sliding window size is set to 5 data points to obtain the average value, and the sensor noise is removed; the abnormal value correction can adopt the quartile range method to correct the outliers and eliminate the abnormal values caused by sensor failure and other problems.

[0045] Based on the historical accident data subjected to data enhancement processing, an adversarial generative network model for accident data expansion is constructed to solve the problem of insufficient accident data. The adversarial generative network model generates synthetic accident data, for example, the input layer of the generator is used to receive historical accident data, the full connection layer adopts 64 nodes and uses ReLU function for activation, and the output layer outputs synthetic accident data. The input layer of the discriminator is used to receive synthetic accident data, the full connection layer adopts 64 nodes and uses LeakyReLU function for activation, and the output layer uses Sigmoid function to output the accuracy of the synthetic accident data. The adversarial generative network model is trained until convergence, for example, the accuracy of the output synthetic accident data is more than 90%, that is, the generative adversarial network training is completed.

[0046] The synthetic accident data is added to the historical accident data for training of the dynamic graph neural network model.

[0047] According to the historical accident data, a dynamic graph neural network model is constructed.

[0048] Specifically, the safety monitoring points of the target coal production scene are mapped to the nodes in the graph structure;

[0049] Based on the physical dependence relationship, geological distribution relationship and historical accident correlation relationship between the safety monitoring points, the directed connection edges between the nodes are constructed and the corresponding edge weight matrix is generated to form the graph structure. For example, the safety monitoring points in the connected roadway establish bidirectional edges, the safety monitoring points within 50 meters in the same rock layer establish undirected edges, and the safety monitoring points that have triggered the same accident alarm at the same time establish directed edges.

[0050] The model is trained by combining the graph structure with the time series monitoring data and historical production control data in the historical accident data. The time series monitoring data in the historical accident data is time series monitoring data integrated according to time series of accident data occurring in a historical time. For example, the historical time can be set to the past 360 days, and multiple safety monitoring records of each safety monitoring point are recorded. The historical production control data refers to control data in coal mine production in a historical time, such as the power of mining machinery, which is in W; and the ventilation volume of the coal mine channel, which is in m³ / s. A three-layer graph convolution network structure is used to build a dynamic graph neural network model. The first layer is used to receive the time series monitoring data in the historical accident data and the historical production control data, the neighbor aggregation mode is average, and the activation function uses the ReLU function. The second layer uses 128 nodes, the neighbor aggregation mode is average, and the activation function uses the ReLU function. The third layer outputs the risk probability and the potential accident propagation path, and the activation function uses the Softmax. The model is repeatedly trained until the convergence condition is reached, for example, the input graph structure node, the time series monitoring data in the historical accident data, and the historical production control data. The output risk probability corresponding to the graph structure node is within ±5%, and the accuracy of the potential accident propagation path is more than 90%. The dynamic graph neural network is constructed.

[0051] By mapping the safety monitoring point to the graph node and dynamically constructing the edge weight matrix based on the physical dependence, geological distribution and accident correlation, the deep coupling of multi-source data is realized, the limitation of single-point data is broken, the multi-dimensional data chain conduction rule is accurately captured, the dynamic graph neural network is trained through the time series monitoring data, and the prediction sensitivity of the model to the compound accident is significantly improved, providing a high-precision spatio-temporal evolution basis for subsequent safety boundary analysis.

[0052] S20: Combine the multi-source safety monitoring data, and perform iterative prediction analysis on the production control scheme through the dynamic graph neural network model to obtain an iterative prediction analysis result, wherein the production control scheme is determined based on a design working condition domain of a target production scene.

[0053] The existing production control strategy has limitations. The safety boundary based on fixed rules often ignores real-time working condition changes, resulting in conservative strategies reducing production capacity or aggressive strategies triggering accidents. There is a lack of quantitative mapping mechanism between multi-dimensional safety indicators and production control parameters, which causes the safety control and production optimization to be disconnected, and the manual experience adjustment has response lag and subjective bias risk.

[0054] Further comprising:

[0055] Collecting the safety boundary analysis result to generate a safety boundary analysis result set;

[0056] According to the safety boundary analysis result set, a historical multi-source safety monitoring data set is extracted and constructed, and a reinforced sample data set is constructed in combination of the safety boundary analysis result set and the historical multi-source safety monitoring data set;

[0057] A lightweight safety boundary rapid prediction model is constructed, and supervised training of the safety boundary rapid prediction model is performed based on the reinforced sample data set;

[0058] The analysis computing power index and the analysis timeliness index of the tunnel anchor equipment management node of the multiple production operation faces in the target coal production scene are evaluated, and the safety boundary rapid prediction model is deployed and verified according to the analysis computing power index and the analysis timeliness index;

[0059] If the deployment verification result shows that the analysis timeliness index is not satisfied, the safety boundary rapid prediction model is iteratively compressed;

[0060] If the deployment verification result shows that the analysis timeliness index is satisfied, the safety boundary rapid prediction model is deployed to the target coal production scene to perform safety boundary analysis.

[0061] In the embodiment of the application, the dynamic graph neural network model is combined with multi-source safety monitoring data to iteratively predict and analyze the production control scheme, and an iteratively predicted and analyzed result is obtained, wherein the production control scheme is determined based on the design working condition domain of the target production scene. For example, the production control scheme includes multiple production control parameters, such as the power of the mining machinery, in units of W, and the ventilation volume of the coal mine channel, in units of m³ / s. The multiple production control schemes and the multi-source safety monitoring data are input into the dynamic graph neural network model, and the risk probability and the potential accident propagation path of the multiple production control schemes are output.

[0062] S30: According to the iteratively predicted and analyzed result and a preset safety boundary constraint, a production control scheme that satisfies the safety boundary constraint is screened as a safety production control scheme set, wherein the safety boundary constraint is determined based on the working face gas emission amount, and the working face gas emission amount includes an absolute emission amount and a relative emission amount;

[0063] According to the iteratively predicted and analyzed result and a preset safety boundary constraint, a production control scheme that satisfies the safety boundary constraint is screened as a safety production control scheme set. The safety boundary constraint can be a risk probability constraint, for example, when the risk probability of the production control scheme is greater than or equal to 30% or the potential accident propagation path exceeds 4 nodes, the safety boundary constraint is not satisfied. Multiple production control schemes that satisfy the safety boundary production constraint are screened as the safety production control scheme set.

[0064] Optionally, the safety boundary constraint can also be determined based on the working face gas emission, wherein the working face gas emission is the gas volume released from the coal seam or rock stratum to the working space in the underground engineering such as coal mine, and is divided into absolute emission (the gas volume released from the entire working face per unit time, with the unit of m³ / min) and relative emission (the gas volume released per unit coal, with the unit of m³ / t);

[0065] S40: performing fitting analysis on the safety production control scheme set to generate a real-time production control safety boundary;

[0066] The fitting analysis is performed on the safety production control scheme set to generate a real-time production control safety boundary. Specifically, the union of each safety production control parameter in the safety production control scheme is obtained, such as the union of the mining and excavation mechanical power, to obtain the maximum and minimum values of the mining and excavation mechanical power as the real-time production control safety boundary of the mining and excavation mechanical power.

[0067] The method provided by the embodiment of the application further includes:

[0068] The safety boundary analysis result set is collected to generate a safety boundary analysis result set.

[0069] According to the safety boundary analysis result set, a historical multi-source safety monitoring data set is extracted and constructed, and the safety boundary analysis result set and the historical multi-source safety monitoring data set are combined to construct a reinforcement sample data set.

[0070] A lightweight safety boundary rapid prediction model is constructed, for example, a full connection neural network is constructed, the input layer is used to receive the multi-source safety monitoring data set, the hidden layer adopts 32 nodes and uses the ReLU function for activation, and the output layer is used to output the predicted safety boundary. Based on the reinforcement sample data set, the safety boundary rapid prediction model is supervised and trained until the model converges, for example, the input multi-source safety monitoring data set, and the error range of the output predicted safety boundary is within ±5%, that is, the safety boundary rapid prediction model is trained.

[0071] The analysis computing power index and the analysis timeliness index of the mining and anchoring equipment management node of the multiple production working faces in the target coal production scene are evaluated, and the safety boundary rapid prediction model is deployed and verified according to the analysis computing power index and the analysis timeliness index. For example, according to the analysis computing power of the on-board edge computing node (used for safety management) corresponding to the current mining and anchoring integrated machine in the target coal production scene, the safety boundary rapid prediction model is deployed. The analysis timeliness index refers to how long the target coal production scene needs to complete the analysis, with the unit of s.

[0072] When the analysis completion time of the safety boundary rapid prediction model is greater than the analysis completion time target required by the target coal scene, that is, when the current safety boundary rapid prediction model does not meet the analysis timeliness index of the target coal production scene.

[0073] If the deployment verification result shows that the analysis timeliness index is not met, the safety boundary rapid prediction model is iteratively compressed, for example, the safety boundary rapid prediction model can be iteratively compressed by a model compression method based on knowledge distillation.

[0074] If the deployment verification result shows that the analysis timeliness index is met, the safety boundary rapid prediction model is deployed to the target coal production scene to perform safety boundary analysis.

[0075] Based on the cooperative analysis of the accident prediction result based on the dynamic graph neural network and the multi-dimensional safety index, the real-time quantitative generation of the safety boundary is realized. By iteratively predicting the accident probability under different production control schemes, the parameter combination meeting all safety index constraints is identified, the experience threshold limitation is broken through, and the abstract prediction value is converted into a specific boundary such as the specific range of mining power to directly guide production decision-making. At the same time, according to the corresponding timeliness, the prediction model is iteratively compressed and optimized, so that the prediction model can better adapt to the demand for limited computing power and fast response in the actual production scene.

[0076] S50: Perform production parameter control of the target coal production scene with the real-time production control safety boundary as a constraint.

[0077] The current coal production parameter control has the defect of passive response. When the sensor alarms, the system usually emergency shutdowns or calls a fixed emergency plan, causing production interruption or over-treatment.

[0078] The step S50 in the method provided in the embodiments of the present application comprises:

[0079] Extracting a real-time production control scheme of the target coal production scene;

[0080] When any production control parameter of the real-time production control scheme exceeds the real-time production control safety boundary, generating an emergency treatment scheme based on a historical case library and an expert knowledge base;

[0081] According to the emergency treatment scheme, performing adaptive production control degradation combined with a preset degradation step;

[0082] Generating a production risk warning signal according to the adaptive production control degradation result, and feeding back the production risk warning signal to a monitoring terminal of the target coal production scene.

[0083] In the embodiments of the present application, a plurality of production control parameters of the target coal scene are collected, and a real-time production control scheme of the target coal production scene is extracted.

[0084] When any production control parameter of the real-time production control scheme exceeds the real-time production control safety boundary, an emergency treatment scheme is generated based on the historical case library and the expert knowledge base, for example, the stress treatment scheme can include reducing the power of the mining machinery, increasing the ventilation volume, etc.

[0085] In combination with a preset degradation step, adaptive production control degradation is performed according to the emergency treatment scheme, wherein the preset degradation step refers to a parameter for degrading the production control when there is a risk, for example, reducing the power of the mining equipment by 20%.

[0086] According to the adaptive production control degradation result, a production risk warning signal is generated and fed back to the monitoring terminal of the target coal production scene, for example, a text message of "there is a safety risk in coal mine production, please confirm in time" is generated and sent to the monitoring terminal of the target coal production scene for warning.

[0087] With the real-time safety boundary as a hard constraint, a dual protection of adaptive control and risk emergency is constructed, first, the production parameters are continuously compared with the dynamic boundary, and the case library driven emergency treatment scheme is triggered before the boundary is crossed to avoid unnecessary shutdown; second, the parameters are adjusted in stages through the preset degradation step to prevent other risks caused by sudden changes in control instructions, and the production continuity is maximized under the premise of safety. Finally, precise and safe control of coal mine production parameters is realized.

[0088] Embodiment two, as shown in Figure 2 Based on the same inventive concept as the coal production optimization method based on data mining provided in embodiment one, the present embodiment also provides a coal production optimization system based on data mining, comprising:

[0089] A model construction module 100 is configured to collect multi-source safety monitoring data and historical accident data of a target coal production scene, and construct a dynamic graph neural network model according to the historical accident data;

[0090] An iterative prediction module 200 is configured to perform iterative prediction analysis on a production control scheme through the dynamic graph neural network model in combination with the multi-source safety monitoring data, and obtain an iterative prediction analysis result, wherein the production control scheme is determined based on a design working condition domain of a target production scene;

[0091] A scheme screening module 300 is configured to screen a production control scheme that meets a safety boundary constraint as a set of safe production control schemes according to the iterative prediction analysis result and the preset safety boundary constraint, wherein the safety boundary constraint is determined based on a working face gas emission amount, and the working face gas emission amount includes an absolute emission amount and a relative emission amount;

[0092] The fitting analysis module 400 is configured to perform fitting analysis on the set of safety production control schemes to generate a real-time production control safety boundary.

[0093] The production parameter control module 500 is configured to perform production parameter control of a target coal production scene with the real-time production control safety boundary as a constraint.

[0094] In one embodiment, the model construction module 100 is further configured to:

[0095] collecting multiple types of safety monitoring data in a coal production process in real time to generate the multi-source safety monitoring data;

[0096] synchronously acquiring accident types, accident locations, and accident-related monitoring data recorded in a historical production cycle to constitute the historical accident data;

[0097] The multiple types of safety monitoring data at least include gas concentration data, dust concentration data, surrounding rock stress data, equipment operating state data, and geological environment monitoring data.

[0098] The dynamic graph neural network model is constructed according to the historical accident data, and the construction includes:

[0099] mapping safety monitoring points of a target coal production scene to nodes in a graph structure;

[0100] constructing directed connection edges between nodes and generating a corresponding edge weight matrix based on physical dependency relationships, geological distribution relationships, and historical accident correlation relationships between the safety monitoring points to form the graph structure;

[0101] training the model in combination with the graph structure and time series monitoring data and historical production control data in the historical accident data to obtain the dynamic graph neural network model;

[0102] The dynamic graph neural network model is constructed according to the historical accident data, and before the construction, the method further includes:

[0103] performing data enhancement processing on the historical accident data, and the data enhancement processing includes data expansion, data smoothing, and abnormal value correction;

[0104] constructing an adversarial generative network model for accident data expansion based on the historical accident data after data enhancement processing, and generating synthetic accident data through the adversarial generative network model;

[0105] adding the synthetic accident data to the historical accident data for training of the dynamic graph neural network model.

[0106] In one embodiment, the method further includes:

[0107] collecting the safety boundary analysis results to generate a safety boundary analysis result set;

[0108] According to the safety boundary analysis result set, a historical multi-source safety monitoring data set is extracted and constructed, and a reinforced sample data set is constructed in combination of the safety boundary analysis result set and the historical multi-source safety monitoring data set;

[0109] A lightweight safety boundary rapid prediction model is constructed, and supervised training of the safety boundary rapid prediction model is performed based on the reinforced sample data set;

[0110] The analysis computing power index and the analysis timeliness index of the excavation anchor equipment management node of the multiple production operation faces in the target coal production scene are evaluated, and the safety boundary rapid prediction model is deployed and verified according to the analysis computing power index and the analysis timeliness index;

[0111] If the deployment verification result shows that the analysis timeliness index is not satisfied, the safety boundary rapid prediction model is iteratively compressed;

[0112] If the deployment verification result shows that the analysis timeliness index is satisfied, the safety boundary rapid prediction model is deployed to the target coal production scene to perform safety boundary analysis.

[0113] In one embodiment, the production parameter control module 500 is further configured to:

[0114] extract a real-time production control scheme of the target coal production scene;

[0115] When any production control parameter of the real-time production control scheme exceeds the real-time production control safety boundary, an emergency disposal scheme is generated based on a historical case library and an expert knowledge base;

[0116] According to the emergency disposal scheme, adaptive production control degradation is performed in combination with a preset degradation step;

[0117] A production risk early warning signal is generated according to the adaptive production control degradation result, and the production risk early warning signal is fed back to a monitoring terminal of the target coal production scene.

[0118] In summary, the embodiments of the present application have at least the following technical effects:

[0119] ​The application provides a coal production optimization method and system based on data mining. By fusing dynamic graph neural network modeling and multi-source safety data collaborative analysis, the dynamic adaptability and safety of production control parameters are significantly improved. First, by mapping safety monitoring points to dynamic graph nodes and constructing edge weight matrices of physical dependence, geological distribution and accident correlation, the nonlinear coupling relationship between heterogeneous data such as gas concentration and surrounding rock stress is effectively captured, solving the problem of missing accident evolution rules caused by traditional single-dimensional analysis. Second, based on the safety boundary of production parameters derived from accident prediction results, the control strategies such as mining machinery power and ventilation volume are always adjusted within the dynamic safety threshold, avoiding the conservative defects or blindness risks of manual experience decision. Third, by using the lightweight model deployment mechanism and reinforced sample training, the safety boundary can still respond quickly in the limited computing power environment of the underground, ensuring the timeliness of analysis and the real-time closed-loop linkage of production control. Finally, through the adaptive degradation mechanism of emergency disposal scheme and the fusion of historical case library, the risk of systematic collapse caused by parameter out-of-bound is greatly reduced. Compared with the traditional method, the technical scheme provided by the application significantly breaks through the limitations of static threshold regulation, realizes the spatio-temporal correlation mining of multi-dimensional safety indicators and the real-time collaborative optimization of production control parameters, and achieves the technical effect of improving the safety of coal production control parameters.

[0120] It should be noted that the above sequence of the embodiments of the application is only for description, and does not represent the advantages and disadvantages of the embodiments. Moreover, the above describes specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or can be advantageous.

[0121] The above only describes the preferred embodiments of the application and does not limit the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

[0122] The present application and the drawings are only exemplary descriptions of the application, and any and all modifications, changes, combinations or equivalents within the scope of the application are considered to be covered by the present application. Obviously, those skilled in the art can make various modifications and changes to the application without departing from the scope of the application. Thus, if these modifications and changes of the application belong to the scope of the application and its equivalent technology, the application intends to include these modifications and changes.

Claims

1. A coal production optimization method based on data mining, characterized by, The method comprises the following steps: Collecting multi-source safety monitoring data and historical accident data of a target coal production scene, and constructing a dynamic graph neural network model according to the historical accident data; Combining the multi-source safety monitoring data, iteratively predicting and analyzing a production control scheme through the dynamic graph neural network model to obtain an iterative prediction and analysis result, wherein the production control scheme is determined based on a design working condition domain of the target production scene; According to the iterative prediction and analysis result and a preset safety boundary constraint, filtering a production control scheme that meets the safety boundary constraint as a set of safe production control schemes, wherein the safety boundary constraint is determined based on a working face gas emission amount, and the working face gas emission amount includes an absolute emission amount and a relative emission amount; Performing fitting analysis on the set of safe production control schemes to generate a real-time production control safety boundary; Controlling production parameters of the target coal production scene with the real-time production control safety boundary as a constraint; Wherein generating a real-time production control safety boundary further comprises: Collecting safety boundary analysis results to generate a set of safety boundary analysis results; According to the set of safety boundary analysis results, extracting and constructing a historical multi-source safety monitoring data set, and combining the set of safety boundary analysis results and the historical multi-source safety monitoring data set to construct a reinforced sample data set; Constructing a lightweight safety boundary rapid prediction model, and performing supervised training of the safety boundary rapid prediction model based on the reinforced sample data set; Evaluating the analysis computing power index and the analysis timeliness index of the excavation and anchoring equipment management nodes of multiple production working faces in the target coal production scene, and deploying and verifying the safety boundary rapid prediction model according to the analysis computing power index and the analysis timeliness index; If the deployment verification result shows that the analysis timeliness index is not met, iteratively compressing the safety boundary rapid prediction model; If the deployment verification result shows that the analysis timeliness index is met, deploying the safety boundary rapid prediction model to the target coal production scene for safety boundary analysis.

2. The data mining based coal production optimization method of claim 1, wherein, Collecting multi-source safety monitoring data and historical accident data of a target coal production scene, comprising: Real-time collection of multi-type safety monitoring data in the coal production process to generate the multi-source safety monitoring data; Synchronous acquisition of accident types, accident locations and accident associated monitoring data recorded in the historical production period to form the historical accident data; Wherein the multi-type safety monitoring data at least includes gas concentration data, dust concentration data, surrounding rock stress data, equipment running state data and geological environment monitoring data.

3. The data mining based coal production optimization method of claim 2, wherein, According to the historical accident data, a dynamic graph neural network model is constructed, comprising: Mapping the safety monitoring points of the target coal production scene to nodes in a graph structure; Based on the physical dependence relationship, geological distribution relationship and historical accident correlation relationship between the safety monitoring points, a directed connection edge between the nodes is constructed and a corresponding edge weight matrix is generated to form the graph structure; Combining the graph structure and the time series monitoring data, historical production control data in the historical accident data to perform model training to obtain the dynamic graph neural network model.

4. The data mining based coal production optimization method of claim 1, wherein, According to the historical accident data, a dynamic graph neural network model is constructed, and before that, the historical accident data is processed by data enhancement, including data expansion, data smoothing and abnormal value correction. Based on the historical accident data after data enhancement, an adversarial generative network model for accident data expansion is constructed, and synthetic accident data is generated through the adversarial generative network model. The synthetic accident data is added to the historical accident data for training of the dynamic graph neural network model. The real-time production control scheme of the target coal production scene is extracted.

5. The data mining based coal production optimization method of claim 1, wherein, When any production control parameter of the real-time production control scheme exceeds the real-time production control safety boundary, an emergency disposal scheme is generated based on the historical case library and the expert knowledge base. According to the adaptive production control degradation result, a production risk warning signal is generated, and the production risk warning signal is fed back to the monitoring terminal of the target coal production scene. A coal production optimization method based on data mining for implementing any one of claims 1-5, the system comprising: A model construction module is used to collect multi-source safety monitoring data and historical accident data of a target coal production scene, and a dynamic graph neural network model is constructed according to the historical accident data. An iterative prediction module is used to combine the multi-source safety monitoring data, and the production control scheme is iteratively predicted and analyzed by the dynamic graph neural network model to obtain an iterative prediction analysis result, wherein the production control scheme is determined based on the design working condition domain of the target production scene.

6. A coal production optimization system based on data mining, characterized by, A scheme screening module is used to screen production control schemes that meet the safety boundary constraint as a set of safe production control schemes according to the iterative prediction analysis result and the preset safety boundary constraint, wherein the safety boundary constraint is determined based on the working face gas emission, and the working face gas emission includes absolute emission and relative emission. A fitting analysis module is used to perform fitting analysis on the set of safe production control schemes to generate a real-time production control safety boundary. A production parameter control module is used to perform production parameter control of the target coal production scene with the real-time production control safety boundary as a constraint. ​ ​ ​

Images