Coal seam model dynamic correction method and device based on LSTM, terminal equipment and medium
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHENGZHOU HENGDA INTELLIGENT CONTROL TECHNOLOGY CO LTD
- Filing Date
- 2026-02-13
- Publication Date
- 2026-06-02
AI Technical Summary
Existing technologies lack the accuracy of three-dimensional geological models of coal seams in fully mechanized mining faces, are outdated, and cannot accurately predict coal seam fluctuations, nor can they fully utilize the revealed coal seam trends for forward-looking predictions.
A dynamic correction method for coal seam models based on LSTM is adopted. By constructing an initial three-dimensional coal seam model, rolling prediction and weighted fusion are performed using the LSTM prediction model to dynamically correct the coal seam model. Combined with Kriging interpolation and real-time data updates, the model accuracy and prediction accuracy are improved.
It enables more intelligent and accurate dynamic correction of coal seam models, reduces reliance on human experience, improves the level of intelligent mining, and provides forward-looking predictive capabilities for coal mining machines.
Smart Images
Figure CN122132420A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of coal mining technology, specifically relating to a method, apparatus, terminal equipment, and medium for dynamic correction of coal seam models based on LSTM. Background Technology
[0002] In intelligent fully mechanized mining faces, a high-precision 3D model of the coal seam is crucial for achieving adaptive cutting by the mining machine and the construction of a transparent working face. Initial models are typically established based on sparse borehole data using geostatistical methods such as Kriging interpolation. However, due to the complexity of geological structures, the initial model often deviates significantly from the actual situation. Therefore, it is necessary to dynamically correct the model using new data revealed during the mining process.
[0003] Existing methods for updating coal seam models based on Kalman filtering or sequential Gaussian simulations typically include the following steps: Establish the initial model: Use borehole data to establish an initial three-dimensional model of the coal seam roof and floor.
[0004] Acquiring real-time data: During the mining process, the elevation data of the top and bottom plates of the coal seam at the location of the coal mining machine drum is obtained through tilt sensors, positioning systems, and other devices installed on the coal mining machine, serving as "hard data".
[0005] Assimilation and Update: Using data assimilation algorithms such as Kalman filtering, new measured data are fused with the initial model, and local corrections are made to the nodes in the mined areas of the model to make the model more closely resemble the actual situation that has been revealed.
[0006] The main drawback of this plan is: Methods such as Kalman filtering mainly focus on correcting the state estimate based on the current observations, making it difficult to effectively explore and utilize the long-term spatial dependence and temporal evolution of coal seam undulations.
[0007] These methods are essentially "retrospective," effective in correcting revealed areas but lacking in forward-looking predictive capabilities for unmined areas, and unable to fully utilize the revealed coal seam trends to predict hidden geological changes ahead.
[0008] The variations in coal seam thickness and undulation are often highly nonlinear, and traditional statistical methods are inadequate in handling such complex nonlinear relationships, resulting in limited correction accuracy.
[0009] In summary, the existing three-dimensional geological models of coal seams in fully mechanized mining faces suffer from insufficient accuracy, outdated updates, and an inability to accurately predict coal seam undulations. Summary of the Invention
[0010] The purpose of this invention is to provide a method, device, terminal equipment and medium for dynamic correction of coal seam models based on LSTM, so as to solve the problem of insufficient accuracy of three-dimensional geological models of coal seams in fully mechanized mining faces in the prior art.
[0011] To address the aforementioned problems, the LSTM-based dynamic correction method for coal seam models in fully mechanized mining faces, as described in this invention, employs the following technical solution: Initial data of the fully mechanized mining face were obtained, and an initial three-dimensional coal seam model of the fully mechanized mining face was constructed using the Kriging interpolation method. The training dataset is determined based on relevant data from the mined areas, and the LSTM prediction model is obtained by pre-training the training dataset based on the LSTM model. The initial coal seam data sequence is determined based on the initial data, and the initial coal seam data sequence is input into the LSTM prediction model for prediction to obtain the next coal seam prediction data. As the fully mechanized mining face advances, the actual elevation values of the current coal seam control points are obtained, and the current coal seam data sequence is updated on a rolling basis according to the actual elevation values of the current coal seam control points. The updated current coal seam data sequence is then input into the LSTM prediction model for prediction, and the next coal seam prediction data is obtained by rolling. The initial three-dimensional coal seam prediction data is weighted and fused with the initial three-dimensional coal seam model in real time to update the initial three-dimensional coal seam model. The elevation prediction values of each coal seam control point obtained by rolling prediction are then weighted and fused with the updated initial three-dimensional coal seam model to dynamically correct the initial three-dimensional coal seam model.
[0012] In some embodiments, the process of determining the training dataset based on relevant data from the mined area includes: Taking the advance direction of the fully mechanized mining face as the time sequence, based on multiple advance moments, a continuous trajectory is extracted from the mined area; and the advance moment and coal seam data sequence corresponding to each continuous trajectory are used as training datasets. Each continuous trajectory segment includes the same number of coal seam control points.
[0013] In some embodiments, the process of determining the coal seam data sequence corresponding to each continuous trajectory segment includes: Determine multiple coal seam control points corresponding to each continuous trajectory based on the advancement time corresponding to each continuous trajectory; Obtain the feature vector corresponding to each coal seam control point; Based on all feature vectors corresponding to multiple coal seam control points, a coal seam data sequence corresponding to each continuous trajectory is constructed according to a time-series arrangement.
[0014] In some embodiments, the feature vector includes coal seam elevation, planar coordinates, and auxiliary feature data.
[0015] In some embodiments, the step of using the propulsion time and coal seam data sequence corresponding to each segment of the continuous trajectory as the training dataset includes: Let each of the aforementioned propulsion times t be a time label S t Label each time point S t Coal seam data sequence Q corresponding to the next propagation time t+1 t+1 Form a set of tag sequence groups (S) t Q t+1 ); The training dataset is formed based on all the label sequence groups.
[0016] In some embodiments, during the process of training the training dataset to obtain an LSTM prediction model based on the LSTM network model, the LSTM network model... F The learning objectives are: Where θ represents all the weights and bias parameters of the LSTM.
[0017] To address the aforementioned issues, the LSTM-based dynamic correction device for coal seam models in fully mechanized mining faces, as described in this invention, employs the following technical solution: it includes a geological model module, an LSTM prediction model module, a rolling prediction module, and a dynamic correction module. The LSTM prediction model module is used to determine the training dataset based on relevant data from the mined area, and to train the LSTM prediction model on the training dataset based on the LSTM model. The geological model module is used to obtain the initial data of the fully mechanized mining face and to construct the initial three-dimensional coal seam model of the fully mechanized mining face using the Kriging interpolation method. The rolling prediction module is used to determine the initial coal seam data sequence based on the initial data, and input the initial coal seam data sequence into the LSTM prediction model for prediction to obtain the next coal seam prediction data; and as the fully mechanized mining face advances, it obtains the actual elevation value of the current coal seam control point, and updates the current coal seam data sequence rollingly based on the actual elevation value of the current coal seam control point, and inputs the current coal seam data sequence after each rolling update into the LSTM prediction model for prediction to obtain the next coal seam prediction data. The dynamic correction module is used to perform weighted fusion of the initial obtained next coal seam prediction data with the initial three-dimensional coal seam model in real time to update the initial three-dimensional coal seam model, and to perform weighted fusion of the elevation prediction value of each coal seam control point obtained by rolling prediction with the updated initial three-dimensional coal seam model to dynamically correct the initial three-dimensional coal seam model.
[0018] To address the aforementioned issues, the terminal device involved in this invention includes a processor and a memory. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned dynamic correction method for coal seam models in fully mechanized mining faces based on LSTM.
[0019] To address the aforementioned problems, the present invention relates to a computer-readable storage medium storing a computer program, which, when executed on a processor, implements the aforementioned method for dynamic correction of coal seam models in fully mechanized mining faces based on LSTM.
[0020] The beneficial effects of this invention are as follows: This invention presents a dynamic correction method for coal seam models in fully mechanized mining faces based on the LSTM algorithm. It acquires initial data for the fully mechanized mining face and constructs an initial three-dimensional coal seam model using Kriging interpolation. Based on the initial data, an initial coal seam data sequence is determined and input into a pre-trained LSTM prediction model to predict the next coal seam. As the fully mechanized mining face advances, the current coal seam data sequence is continuously updated based on the actual elevation values of the current coal seam control points. Each updated sequence is then input into the LSTM prediction model for rolling prediction to obtain the next coal seam prediction data. The initial and predicted next coal seam prediction data are then weighted and fused with the initial three-dimensional coal seam model in real time to dynamically correct the model. This invention constructs the LSTM prediction model using a spatiotemporal sequence approach and employs a rolling method to dynamically correct the three-dimensional coal seam model, thereby improving the real-time accuracy and prediction accuracy of the coal seam model. Attached Figure Description
[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below: Figure 1 This is a schematic diagram of the process structure of the dynamic correction method for the coal seam model of a fully mechanized mining face according to an embodiment of the present invention. Detailed Implementation
[0022] To make the technical objectives, technical solutions, and beneficial effects of the present invention clearer, the technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention; that is, the described embodiments are merely some embodiments of the present invention, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0023] In the following text, the terms "comprising," "having," and their cognates, which may be used in various embodiments of this application, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more combinations thereof. Furthermore, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.
[0024] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be construed as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.
[0025] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0026] The purpose of this application is to provide a more intelligent, accurate, and forward-looking method for dynamic correction of coal seam models. Specific objectives include: By leveraging the powerful temporal memory and nonlinear mapping capabilities of LSTM neural networks, we can learn the dynamic changes of coal seams from the sequence data of mined areas.
[0027] This enables accurate prediction of future coal seam trends in unmined areas based on historical sequences, providing a basis for coal mining machines to adjust their cutting trajectories in advance.
[0028] Build an automated, closed-loop model dynamic correction system to reduce reliance on human experience and improve the level of intelligent mining.
[0029] The following describes the LSTM-based dynamic correction method, apparatus, terminal equipment, and medium for coal seam models using specific embodiments.
[0030] Specific embodiments of the dynamic correction method for coal seam models in fully mechanized mining faces based on LSTM involved in this invention are as follows: Figure 1 As shown, the dynamic correction method for coal seam models in fully mechanized mining faces based on LSTM includes: S100: Obtain initial data of the fully mechanized mining face and construct an initial three-dimensional coal seam model of the fully mechanized mining face using the Kriging interpolation method.
[0031] In this embodiment of the invention, the initial data includes surface borehole data, downhole borehole data, 3D seismic exploration data, and adjacent working face data.
[0032] Specifically, the surface borehole data is obtained during the design phase of the fully mechanized mining face. The planar coordinates (x, y) and borehole depth of each of the 15 exploration boreholes constructed on the surface are acquired. The absolute elevation of the coal seam floor, i.e., the coal seam elevation (z-coordinate), is calculated through inclination surveying. The coal seam thickness is also recorded for each borehole.
[0033] Specifically, the underground borehole data consists of eight fan-shaped boreholes drilled from the two roadways into the working face during the tunnel excavation process to supplement local geological information.
[0034] Specifically, 3D seismic exploration data uses seismic wave reflection time data to interpret the structural morphology trend surface of the coal seam floor.
[0035] Specifically, the adjacent working face data consists of geological sketches and actual exposed data of the fully mechanized working faces that have already been mined.
[0036] The acquired multi-source data is processed by coordinate unification, noise reduction, alignment and standardization to form high-quality initial data. Then, the commonly used Kriging interpolation method is used to construct the initial three-dimensional coal seam model of the fully mechanized mining face, which serves as the basis model for subsequent updates.
[0037] S200: The training dataset is determined based on relevant data from the mined area, and the LSTM prediction model is trained on the training dataset based on the LSTM model.
[0038] In this embodiment of the invention, the process of determining the training dataset based on relevant data from the mined area includes: taking the advance direction of the fully mechanized mining face as the time sequence, and based on multiple advance times, extracting a continuous trajectory from the mined area; and using the advance time and coal seam data sequence corresponding to each continuous trajectory as the training dataset; wherein, each continuous trajectory includes the same number of coal seam control points.
[0039] Each segment of the continuous trajectory includes N coal seam control points; the feature vector corresponding to each coal seam control point includes coal seam elevation, plane coordinates, and auxiliary feature data; for example, the auxiliary feature data is support pressure or ground-penetrating radar data, etc.
[0040] It should be noted that a coal seam data sequence refers to a list of discrete data points sampled at certain intervals (the cutting trajectory of the coal mining machine) on a continuous spatial surface, arranged in sequence; it is a discretized and digitized representation of the coal seam spatial sequence. Specifically, the coal seam spatial sequence refers to the actual shape and trend of the coal seam roof or floor surface in three-dimensional space; it is a continuous spatial surface concept. For example, "in the direction of face advancement, the coal seam floor first descends and then rises" describes its spatial sequence shape.
[0041] For example, for a given working face advancement time t, a continuous trajectory is extracted from the exposed area. Assume the data for each coal seam control point i is a feature vector P. i : P i =[z i ,x i ,y i ,f i1 ,...,f ij ,...] Among them, z i It is the coal seam elevation (the core prediction target), (x i ,y i ) is a planar coordinate, f ij These are auxiliary characteristic data such as stent pressure.
[0042] Based on all feature vectors of N coal seam control points of the continuous trajectory, a coal seam data sequence Q corresponding to the advancement time t is constructed based on time series. t : Q t =[P t−N+1 ,P t−N+2 ,...,P t ] This coal seam data sequence is arranged in chronological order of coal cutting by the mining machine.
[0043] This invention first uses each propulsion moment as a time tag S. tAnd the coal seam data sequence Q corresponding to the next propagation time t+1. t+1 Form a set of tag sequence groups (S) t Q t+1 ), where Q t+1 It is a time tag S t The next location is the actual exposed elevation, i.e., the coal seam data sequence. This is then used to form a training dataset from many groups of labeled sequences.
[0044] In this embodiment of the invention, the process of training the training dataset based on the LSTM network model to obtain the LSTM prediction model includes: LSTM network model F The learning objectives are: Where θ represents all the weights and bias parameters of the LSTM.
[0045] By minimizing the mean squared error (MSE) between the predicted and true values through backpropagation and gradient descent (such as the Adam optimizer), model F eventually learns to infer the elevation of the (N+1)th point from the spatial distribution and attributes of the historical N points.
[0046] Grid search optimization process: (1) Define the hyperparameter search space: determine the key hyperparameters to be optimized and their candidate value ranges.
[0047] For example: Number of hidden layer neurons: {64, 128, 256} Number of LSTM layers: {1, 2, 3} Learning rate: {0.001, 0.005, 0.01} Sequence length N: {10, 20, 30} Batch size: {16, 32, 64} (2) Cross-validation: Divide the historical dataset into training set, validation set and test set.
[0048] For each combination of parameters, the training dataset is divided into 5 parts, with 4 parts used as the training set and 1 part as the validation set in rotation.
[0049] (3) Traversal and training: For all combinations in the parameter grid (e.g., 3x3x3x3=81), train an LSTM model using the training set.
[0050] (4) Evaluation and selection: Evaluate the performance of each model using the validation set. Select the hyperparameter combination that performs best on the validation set.
[0051] (5) Final test: Retrain the model on the full training set using the optimal hyperparameters and evaluate its generalization ability using an independent test set.
[0052] Five-fold cross-validation is used. For each parameter combination, the training dataset is divided into five parts. Four parts are used as the training set and one part as the validation set in turn. The average of the five validation errors is calculated as the performance index of the parameter combination. The parameter combination with the smallest average validation error is selected as the optimal parameter.
[0053] Therefore, the training objective of the LSTM network model in this invention is to minimize the mean square error between the predicted elevation value and the actual exposed elevation value. The trained LSTM model can capture the spatiotemporal evolution pattern of coal seam undulations.
[0054] The LSTM prediction model trained by this invention outputs the predicted elevation value of the next coal seam control point (i.e., the unmined area ahead).
[0055] In this embodiment of the invention, the coal seam data sequence uses absolute coordinates, and the coal seam elevation data can also be converted into relative changes (such as the elevation difference between adjacent points) as model input, which may help the model learn the changing trend more quickly.
[0056] S300: Determine the initial coal seam data sequence based on the initial data, and input the initial coal seam data sequence into the LSTM prediction model for prediction to obtain the next coal seam prediction data.
[0057] In this embodiment of the invention, the predicted data for the next coal seam is the predicted elevation value of the next coal seam control point. As another implementation, as the working face advances, the data of the N newly exposed continuous coal seam control points are used as input and fed into the trained LSTM prediction model to predict the elevation values of the M coal seam control points ahead. This method can output a complete coal seam profile for a region, realizing a one-time update of a larger-scale three-dimensional coal seam model, thereby improving efficiency.
[0058] In some embodiments, during initialization, the data of the first N exposed coal seam control points in the fully mechanized mining face constitute the initial sequence S. 0= [P t−N+1 ,...,P0].
[0059] Initialization: At the starting point of mining, the feature vectors of the first N exposed real coal seam control points are used to construct the initial coal seam data sequence Q0.
[0060] Initial prediction: Input the initial coal seam data sequence Q0 into the pre-trained LSTM prediction model to obtain the coal seam prediction data, i.e., the elevation prediction value Z`1 of the next coal seam control point.
[0061] S400: As the fully mechanized mining face advances, the actual elevation values of the current coal seam control points are obtained, and the current coal seam data sequence is updated on a rolling basis according to the actual elevation values of the current coal seam control points. The current coal seam data sequence after each rolling update is input into the LSTM prediction model for prediction, and the next coal seam prediction data is obtained by rolling.
[0062] In some embodiments, the coal mining machine cuts forward, actually revealing the actual elevation value Z1 of the next coal seam control point; and the current coal seam data sequence is updated on a rolling basis according to the plane coordinates of the exploration borehole corresponding to the actually revealed next coal seam control point and auxiliary feature data.
[0063] Specifically, the feature vector of the next coal seam control point is added to the end of the current coal seam data sequence, while the point P at the beginning of the current coal seam data sequence is discarded. t−N+1 This forms a new coal seam data sequence of length N, Q1=[P t−N+2 [,...,P0,P1]. Then, a rolling loop is executed to perform sequence sliding, and the prediction data for the next coal seam is predicted in sequence.
[0064] S500 performs real-time weighted fusion of the initial predicted data of the next coal seam with the initial three-dimensional coal seam model to update the initial three-dimensional coal seam model. It also performs weighted fusion of the elevation prediction values of each coal seam control point obtained from the rolling prediction with the updated initial three-dimensional coal seam model to dynamically correct the initial three-dimensional coal seam model.
[0065] In this embodiment of the invention, the prediction results of LSTM are weighted and fused with the initial three-dimensional coal seam model. During the weighted fusion process, the LSTM prediction results are given higher weights to areas closer to the most recently exposed area, while the weights of the three-dimensional coal seam model gradually increase to areas further away from the most recently exposed area. It is important to emphasize that the weighted fusion only changes the geological model (elevation information) and does not change the material properties.
[0066] The dynamic correction device for coal seam model of fully mechanized mining face based on LSTM involved in this invention includes a geological model module, an LSTM prediction model module, a rolling prediction module, and a dynamic correction module.
[0067] The LSTM prediction model module is used to determine the training dataset based on relevant data from the mined area, and to train the LSTM prediction model on the training dataset based on the LSTM model. The geological model module is used to obtain the initial data of the fully mechanized mining face and to construct the initial three-dimensional coal seam model of the fully mechanized mining face using the Kriging interpolation method. The rolling prediction module is used to determine the initial coal seam data sequence based on the initial data, and input the initial coal seam data sequence into the LSTM prediction model for prediction to obtain the next coal seam prediction data; and as the fully mechanized mining face advances, it obtains the actual elevation value of the current coal seam control point, and updates the current coal seam data sequence rollingly based on the actual elevation value of the current coal seam control point, and inputs the current coal seam data sequence after each rolling update into the LSTM prediction model for prediction to obtain the next coal seam prediction data. The dynamic correction module is used to perform weighted fusion of the initial obtained next coal seam prediction data with the initial three-dimensional coal seam model in real time to update the initial three-dimensional coal seam model, and to perform weighted fusion of the elevation prediction value of each coal seam control point obtained by rolling prediction with the updated initial three-dimensional coal seam model to dynamically correct the initial three-dimensional coal seam model.
[0068] This application also provides a terminal device, exemplary of which includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to enable the terminal device to perform the above-described LSTM-based dynamic correction method for coal seam models.
[0069] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including at least one of a Central Processing Unit (CPU), Graphics Processing Unit (GPU), Network Processor (NP), Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.
[0070] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), and Electrically Erasable Programmable Read-Only Memory (EEPROM). The memory stores computer programs, and the processor, upon receiving execution instructions, can execute the computer programs accordingly.
[0071] This application also provides a computer-readable storage medium for storing the computer program used in the aforementioned terminal device. For example, the computer-readable storage medium may include, but is not limited to, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0072] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, in alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0073] In addition, the functional modules or units in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0074] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a smartphone, personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.
[0075] Finally, it should be noted that the above embodiments are only for illustration and not for limiting the technical solutions of the present invention. Any equivalent substitutions, modifications or partial substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A dynamic correction method for coal seam models in fully mechanized mining faces based on LSTM, characterized in that, The process of determining the training dataset based on relevant data from the mined areas includes: Taking the advance direction of the fully mechanized mining face as the time sequence, based on multiple advance moments, a continuous trajectory is extracted from the mined area; and the advance moment and coal seam data sequence corresponding to each continuous trajectory are used as training datasets. Each continuous trajectory segment includes the same number of coal seam control points.
2. The method for dynamic correction of coal seam models in fully mechanized mining faces based on LSTM according to claim 1, characterized in that, The process of determining the training dataset based on relevant data from the mined areas includes: Taking the advance direction of the fully mechanized mining face as the time sequence, based on multiple advance moments, a continuous trajectory is extracted from the mined area; and the advance moment and coal seam data sequence corresponding to each continuous trajectory are used as training datasets. Each continuous trajectory segment includes the same number of coal seam control points.
3. The method for dynamic correction of coal seam models in fully mechanized mining faces based on LSTM according to claim 2, characterized in that, The process of determining the coal seam data sequence corresponding to each continuous trajectory includes: Determine multiple coal seam control points corresponding to each continuous trajectory based on the advancement time corresponding to each continuous trajectory; Obtain the feature vector corresponding to each coal seam control point; Based on all feature vectors corresponding to multiple coal seam control points, a coal seam data sequence corresponding to each continuous trajectory is constructed according to a time-series arrangement.
4. The method for dynamic correction of coal seam models in fully mechanized mining faces based on LSTM according to claim 3, characterized in that, The feature vector includes coal seam elevation, plane coordinates, and auxiliary feature data.
5. The method for dynamic correction of coal seam models in fully mechanized mining faces based on LSTM according to claim 4, characterized in that, The training dataset, which uses the advancement time and coal seam data sequence corresponding to each segment of the continuous trajectory as the basis, includes: Let each of the aforementioned propulsion times t be a time label S t Label each time point S t Coal seam data sequence Q corresponding to the next propagation time t+1 t+1 Form a set of tag sequence groups (S) t Q t+1 ); The training dataset is formed based on all the label sequence groups.
6. The method for dynamic correction of coal seam models in fully mechanized mining faces based on LSTM according to claim 5, characterized in that, In the process of training the training dataset based on the LSTM network model to obtain the LSTM prediction model, the LSTM network model F The learning objectives are: Where θ represents all the weights and bias parameters of the LSTM.
7. A dynamic correction device for coal seam models in fully mechanized mining faces based on LSTM, characterized in that, It includes a geological model module, an LSTM prediction model module, a rolling prediction module, and a dynamic correction module; The LSTM prediction model module is used to determine the training dataset based on relevant data from the mined area, and to train the LSTM prediction model on the training dataset based on the LSTM model. The geological model module is used to obtain the initial data of the fully mechanized mining face and to construct the initial three-dimensional coal seam model of the fully mechanized mining face using the Kriging interpolation method. The rolling prediction module is used to determine the initial coal seam data sequence based on the initial data, and input the initial coal seam data sequence into the LSTM prediction model for prediction to obtain the next coal seam prediction data. As the fully mechanized mining face advances, the actual elevation values of the current coal seam control points are obtained, and the current coal seam data sequence is updated on a rolling basis according to the actual elevation values of the current coal seam control points. The current coal seam data sequence after each rolling update is input into the LSTM prediction model for prediction, and the next coal seam prediction data is obtained on a rolling basis. The dynamic correction module is used to perform weighted fusion of the initial obtained next coal seam prediction data with the initial three-dimensional coal seam model in real time to update the initial three-dimensional coal seam model, and to perform weighted fusion of the elevation prediction value of each coal seam control point obtained by rolling prediction with the updated initial three-dimensional coal seam model to dynamically correct the initial three-dimensional coal seam model.
8. A terminal device, characterized in that, The terminal device includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the dynamic correction method for coal seam models of fully mechanized mining faces based on LSTM as described in any one of claims 1-6.
9. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed on a processor, implements the dynamic correction method for coal seam models of fully mechanized mining faces based on LSTM according to any one of claims 1-6.