Mamba-lstm fusion model-based spatiotemporal prediction method for mine pressure

Abstract

This application proposes a spatiotemporal prediction method for mine pressure based on a Mamba-LSTM fusion model. The method includes: acquiring pressure monitoring data of hydraulic supports operating at the longwall mining face and adjacent hydraulic supports; acquiring external data related to mine pressure at the longwall mining face; constructing multi-dimensional features based on the pressure monitoring data of the hydraulic supports, adjacent hydraulic supports, and external data, including temporal, spatial, and external features; fusing the temporal, spatial, and external features to obtain fused features; and inputting the fused features into a pre-trained mine pressure prediction model to obtain the mine pressure prediction results for future periods. The mine pressure prediction model is a Mamba-LSTM fusion model. This application can provide early warning of roof pressure at longwall mining faces, ensuring safe mine production and improving mining efficiency.

Description

Technical Field

[0001] This application relates to technologies such as deep learning in the fields of coal mining technology and artificial intelligence, and in particular to a spatiotemporal prediction method for mine pressure based on the Mamba-LSTM fusion model. Background Technology

[0002] Monitoring data on the working resistance of hydraulic supports can accurately reflect the movement state of the overburden in the fully mechanized mining face and the support effect, which is of great significance for timely early warning of periodic roof pressure and prevention of roof collapse accidents. However, mine pressure monitoring data is affected by a variety of factors such as geological conditions, mining procedures, support layout and equipment operating status, exhibiting strong time dependence and spatial heterogeneity, and is accompanied by a large amount of noise and nonlinear fluctuations.

[0003] With the acquisition of massive amounts of mine pressure monitoring data by the electro-hydraulic control system of the longwall mining face support, data-driven mine pressure prediction methods have become possible. However, the mine pressure prediction methods in related technologies have significant limitations: traditional physical models rely on simplified assumptions and are difficult to fully capture the spatiotemporal dependence and nonlinear fluctuations of mine pressure affected by geological conditions, mining processes, and equipment status; although data-driven methods can utilize massive amounts of monitoring data, their accuracy decays significantly in long-term predictions, and they are difficult to effectively integrate multi-source heterogeneous information such as time, space, and external working conditions, resulting in insufficient characterization of the complex evolution of mine pressure and limited generalization ability. Summary of the Invention

[0004] This application provides a method for spatiotemporal prediction of mineral pressure based on a Mamba-LSTM fusion model, a model training method, an apparatus, and an electronic device.

[0005] In a first aspect, embodiments of this application provide a method for spatiotemporal prediction of mineral pressure based on a Mamba-LSTM fusion model, including: Acquire pressure monitoring data of the hydraulic supports operating in the fully mechanized coal mining face and adjacent hydraulic supports; Acquire external data associated with the mine pressure at the fully mechanized coal mining face, wherein the external data includes at least one of the following: the advance speed, position, and direction of the coal mining machine; Feature extraction is performed based on the pressure monitoring data of the hydraulic support to obtain time-domain features, and feature extraction is performed based on the pressure monitoring data of the adjacent hydraulic support to obtain spatial features, and feature extraction is performed based on the external data to obtain external features. The temporal features, spatial features, and external features are fused to obtain fused features; The fused features are input into a pre-trained mine pressure prediction model to obtain the mine pressure prediction results for future periods output by the mine pressure prediction model; wherein, the mine pressure prediction model is a Mamba-LSTM fusion model, and the mine pressure prediction model has learned the mapping relationship between the fused features and the roof pressure.

[0006] Secondly, embodiments of this application provide a method for training a mine pressure prediction model, including: Acquire historical pressure monitoring data for hydraulic supports operating in the fully mechanized coal mining face and for adjacent hydraulic supports. Acquire external data associated with the mine pressure at the fully mechanized coal mining face, wherein the external data includes at least one of the following: the advance speed, position, and direction of the coal mining machine; Feature extraction is performed based on the historical pressure monitoring data of the hydraulic support to obtain time-domain features, and feature extraction is performed based on the historical pressure monitoring data of the adjacent hydraulic support to obtain spatial features, and feature extraction is performed based on the external data to obtain external features. The temporal features, spatial features, and external features are fused to obtain fused features; The fused features are input into the mining pressure prediction model to be trained to obtain the mining pressure prediction results for future periods output by the mining pressure prediction model; wherein, the mining pressure prediction model is a Mamba-LSTM fusion model; Based on the predicted mine pressure results and the actual monitored mine pressure values, a loss value is generated, and the mine pressure prediction model is trained based on the loss value.

[0007] Thirdly, embodiments of this application provide a spatiotemporal prediction device for mineral pressure based on a Mamba-LSTM fusion model, comprising: The first acquisition module is used to acquire the pressure monitoring data of the hydraulic supports working in the fully mechanized coal mining face and the adjacent hydraulic supports. The second acquisition module is used to acquire external data related to the mine pressure of the fully mechanized coal mining face, wherein the external data includes at least one of the advancing speed, position, and direction of the coal mining machine; The feature extraction module is used to extract features based on the pressure monitoring data of the hydraulic support to obtain time-domain features, extract features based on the pressure monitoring data of the adjacent hydraulic support to obtain spatial features, and extract features based on the external data to obtain external features. The feature fusion module is used to fuse the temporal features, the spatial features, and the external features to obtain fused features; The prediction module is used to input the fused features into a pre-trained mine pressure prediction model to obtain the mine pressure prediction results for future periods output by the mine pressure prediction model; wherein, the mine pressure prediction model is a Mamba-LSTM fusion model, and the mine pressure prediction model has learned the mapping relationship between the fused features and the roof pressure.

[0008] Fourthly, embodiments of this application provide a training device for a mine pressure prediction model, comprising: The first acquisition module is used to acquire historical pressure monitoring data of the hydraulic supports working in the fully mechanized coal mining face and the adjacent hydraulic supports. The second acquisition module is used to acquire external data related to the mine pressure of the fully mechanized coal mining face, wherein the external data includes at least one of the advancing speed, position, and direction of the coal mining machine; The feature extraction module is used to extract features based on the historical pressure monitoring data of the hydraulic support to obtain time-domain features, extract features based on the historical pressure monitoring data of the adjacent hydraulic support to obtain spatial-domain features, and extract features based on the external data to obtain external features. The feature fusion module is used to fuse the temporal features, the spatial features, and the external features to obtain fused features; The prediction module is used to input the fused features into the mining pressure prediction model to be trained, and obtain the mining pressure prediction results for future periods output by the mining pressure prediction model; wherein, the mining pressure prediction model is a Mamba-LSTM fusion model; The training module is used to generate a loss value based on the predicted mine pressure results and the actual monitored mine pressure values, and to train the mine pressure prediction model based on the loss value.

[0009] Fifthly, embodiments of this application provide an electronic device, including: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first or second aspect above.

[0010] Sixthly, embodiments of this application provide a storage medium storing instructions that, when executed on an electronic device, cause the electronic device to perform the method described in the first or second aspect above.

[0011] In a seventh aspect, embodiments of this application provide a program product, including at least one of a program and instructions, wherein when the program and instructions are executed by a processor, they implement the steps of the method described in the first or second aspect above.

[0012] According to the technical solution of this application, a multi-dimensional feature system is constructed by fusing multi-source information such as time, space, and external features to comprehensively capture the dynamic changes and spatial coupling extrema of support pressure. Based on this, a mine pressure prediction model is adopted, integrating the advantages of the Mamba and LSTM modules to perform optimized calculations for both short-term and long-term predictions. The entire process is efficient and stable, solving the problems of insufficient accuracy and high noise interference in long-term time-series predictions using traditional methods. It provides a reliable and accurate mine pressure prediction solution, effectively mitigating the accuracy decay problem caused by increasing prediction step size, while balancing the accuracy of short-term predictions with the stability of long-term predictions. Furthermore, compared to existing methods, this application not only significantly improves the accuracy of mine pressure prediction, especially in long-term time-series prediction capabilities, but also effectively reduces the negative impact of noise interference and sequence cumulative errors on the prediction results, enhancing the model's generalization ability and stability. This is beneficial for early warning of periodic roof pressure in mines, reducing the risk of mine safety accidents, improving mine production safety and mining efficiency, and has significant economic and social benefits.

[0013] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0014] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating the spatiotemporal prediction method for mineral pressure based on the Mamba-LSTM fusion model provided in the embodiments of this application. Figure 2 A flowchart illustrating the training method for the mine pressure prediction model provided in this application embodiment; Figure 3 A block diagram of a mine pressure spatiotemporal prediction device based on a Mamba-LSTM fusion model provided in this application embodiment; Figure 4 A block diagram of the mine pressure prediction model training device provided in the embodiments of this application; Figure 5 A block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0015] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0016] The embodiments described in this application are not exhaustive, but merely illustrative of some embodiments, and are not intended to limit the scope of protection of this application. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined with each other. For example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0017] In each embodiment of this application, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of the embodiments are consistent and can be referenced by each other. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0018] The terminology used in the embodiments of this application is for the purpose of describing specific embodiments only and is not intended to limit the scope of this application.

[0019] The following description, with reference to the accompanying drawings, describes a method, model training method, apparatus, and electronic equipment for spatiotemporal prediction of mineral pressure based on a Mamba-LSTM fusion model, as well as embodiments thereof.

[0020] It should be noted that the execution entity of the mine pressure spatiotemporal prediction method in this application embodiment can be a mine pressure spatiotemporal prediction device, which can be implemented by software and / or hardware and can be configured in an electronic device. For example, the electronic device can include, but is not limited to, a terminal, a server, etc.

[0021] Figure 1 This is a flowchart illustrating the spatiotemporal prediction method for mineral pressure based on the Mamba-LSTM fusion model provided in an embodiment of this application. Figure 1 As shown, the spatiotemporal prediction method for mine pressure may include, but is not limited to, the following steps.

[0022] In step 101, pressure monitoring data of the hydraulic supports operating in the fully mechanized coal mining face and adjacent hydraulic supports are obtained.

[0023] In the embodiments of this application, in a fully mechanized coal mining face, pressure monitoring data of the hydraulic support and adjacent hydraulic supports can be collected by sensors, thereby obtaining the pressure monitoring data P of the hydraulic support and adjacent hydraulic supports from the data collected by the sensors.

[0024] Optionally, in some embodiments, the raw pressure monitoring data from different hydraulic supports undergoes preliminary cleaning. For missing data, a linear interpolation formula can be used to complete it. For example, the linear interpolation formula is expressed as follows:

[0025] in, i Indicates the number of the hydraulic support; For time; This is the pressure value. For obviously erroneous data records, set a reasonable threshold range. ,like These are then discarded. After processing, the data is aggregated to form time series data, laying the foundation for subsequent spatiotemporal alignment.

[0026] In step 102, external data related to the mine pressure of the fully mechanized coal mining face is acquired. This external data may include at least one of the following: the advance speed, position, and direction of the coal mining machine.

[0027] It should be noted that the coal mining machine in the embodiments of this application can also be called a coal cutter. For example, in the embodiments of this application, the external data may include working status-related data such as the advancing speed, position, and direction of the coal mining machine, but is not limited to this. For example, the external data may also include other external factors that affect mine pressure, such as mine environment temperature and humidity and / or ambient air pressure.

[0028] Optionally, missing data in the external data can be filled in using a linear interpolation formula. The implementation method is similar to that of missing data filling for pressure mentioned above, and will not be repeated here.

[0029] It should be noted that, in some embodiments, the execution order of steps 102 and 101 can be interchanged, or they can be executed simultaneously.

[0030] In step 103, feature extraction is performed based on the pressure monitoring data of the hydraulic support to obtain time-domain features, feature extraction is performed based on the pressure monitoring data of the adjacent hydraulic support to obtain spatial features, and feature extraction is performed based on external data to obtain external features.

[0031] In the embodiments of this application, multidimensional features can be constructed based on the pressure monitoring data of hydraulic supports, the pressure monitoring data of adjacent hydraulic supports, and external data. These multidimensional features may include time-domain features, spatial-domain features, and external features.

[0032] In some embodiments, a sliding window is used to slide the pressure monitoring data of the hydraulic support, and the mean, variance, maximum, and minimum values ​​of the pressure monitoring data within the window are calculated. These mean, variance, maximum, and minimum values ​​are then used as time-domain features. For example, a preset sliding window size can be used to slide the pressure monitoring data of the hydraulic support, and the mean, variance, maximum, and minimum values ​​of the pressure monitoring data within the window can be calculated. These mean, variance, maximum, and minimum values ​​are then determined as time-domain features (also called temporal features). As an example, this time-domain feature... It can be represented as follows:

[0033] in, Indicates from time arrive The pressure value sequence, which utilizes a window size of The sequence of window pressure values ​​obtained by sliding a window through the pressure monitoring data of the hydraulic support, including data from time... arrive All pressure values; Represents the maximum value function; Describes the minimum value function; Represents the mean function; This represents the variance function.

[0034] It is worth noting that spatial features can be used to characterize the interaction mechanism between hydraulic supports in a mining environment. Based on this, embodiments of this application can construct spatial features based on pressure monitoring data of adjacent hydraulic supports. Optionally, in some embodiments, the average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support can be calculated based on the pressure monitoring data. Spatial features are then constructed based on these average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support, where the average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support can be used to quantify the spatial correlation between hydraulic supports. As an example, this spatial feature... It can be represented as follows:

[0035] in, For the first The pressure value of a nearby hydraulic support. This represents the number of adjacent hydraulic supports; express The maximum value among the pressures of adjacent hydraulic supports; express The minimum value among the pressures of adjacent hydraulic supports; express The average pressure of a neighboring hydraulic support; express The variance of the pressure of a neighboring hydraulic support.

[0036] To facilitate subsequent spatial feature analysis by taking into account the relative positional changes of the coal mining machine among the hydraulic supports during the mining process, optionally, in some embodiments, the spatial distribution of the hydraulic supports and adjacent hydraulic supports can be analyzed. The positional information of each hydraulic support and adjacent hydraulic supports can be normalized to ensure that the spatial feature analysis considers the relative positional changes of the coal mining machine among the hydraulic supports during the mining process, thus consistently evaluating the impact of different hydraulic supports on mine pressure changes. The formula used for this normalization is as follows:

[0037] in, This represents the original absolute position value of the hydraulic support in space; This represents the left endpoint of the interval containing the original position value of the hydraulic support; This represents the right endpoint of the interval containing the original position value of the hydraulic support; The final value is [0,1].

[0038] In embodiments of this application, feature extraction can be performed on external data to obtain external features. For example, external features are variables related to the working state of the coal mining machine, such as speed, position, and direction, to enhance the model's ability to perceive the mining state and more accurately grasp the motion state of the coal mining machine and its impact on changes in mine pressure. These external features may include, but are not limited to, features such as speed, position, and direction constructed from variables related to the coal mining machine's operating state. As an example, these external features... It can be represented as follows:

[0039] in, Indicates the location of the coal mining machine; This indicates the speed of the coal mining machine, relative to time. speed The calculation formula is as follows: , l The location of the coal mining machine; Indicates the direction of motion of the coal mining machine with respect to time. direction of movement The calculation formula is as follows: Through feature sets The model can comprehensively capture the dynamic changes of mining pressure, its spatial dependence, and external influences.

[0040] In step 104, the temporal features, spatial features, and external features are fused to obtain fused features.

[0041] To further improve the effectiveness of features and the accuracy of model predictions, multidimensional features (such as temporal, spatial, and external features) can be spatiotemporally aligned before fusion processing. In some embodiments, the physical position of hydraulic supports operating on a fully mechanized coal mining face can be normalized to convert the spatial coordinates of the hydraulic supports to a uniform scale. The movement of the coal mining machine between hydraulic supports is defined by a step size, each step consisting of multiple time steps. Based on the coal mining machine's workflow, a corresponding spatial position is assigned to each time step. Through the mapping relationship between time steps and spatial step sizes, the temporal features and spatial position information in pressure monitoring data and external data are spatiotemporally aligned.

[0042] For example, the position of the hydraulic support and the movement step of the coal mining machine can be processed. Through a mapping function between the time step and the spatial step, precise alignment between the time series and the spatial position can be achieved, eliminating data source inconsistencies. The physical position of the hydraulic support can be normalized, converting the spatial coordinates to a uniform scale. As an example, the formula used for this normalization operation is as follows: ] in, This represents the minimum coordinate value of the hydraulic support in the current fully mechanized coal mining face. This represents the maximum coordinate value of the hydraulic support in the current fully mechanized coal mining face. For the first i The spatial coordinates of a hydraulic support.

[0043] The movement of the coal mining machine between hydraulic supports is defined as a step, and each step consists of multiple time steps. The workflow of the coal mining machine is described, with each time step... Assign corresponding spatial locations to ensure the alignment of time and space. Each step size contains multiple time steps. Each movement of the coal mining machine corresponds to a new step size. Define the coal mining machine movement step size s as corresponding to m time points. The mapping relationship between time steps and spatial steps is as follows:

[0044] Among them, the function M Time step Mapping to spatial step size This allows for the synchronization of mining progress with the time sequence. Each moment is associated with the support position or section where the coal mining machine is located, achieving a one-to-one correspondence between time information and spatial location, thus preparing for the construction of spatiotemporal feature vectors.

[0045] In the embodiments of this application, temporal and spatial features are combined, and corresponding temporal features are combined with spatial location information to form a spatiotemporal feature vector, which is then aligned with the temporal features. Spatial features splicing together to form a joint representation: ,in, This indicates a feature concatenation operation. d = k Ensure dimensionality balance. The concatenated spatiotemporal feature vector contains both temporal and spatial information. The aligned temporal features (also called temporal domain features), spatial features (also called spatial domain features), and external features are concatenated into a joint feature vector, which is the fused feature, and used as input to the mine pressure prediction model. For example, the fused feature... It is expressed as follows: .

[0046] In other words, this application constructs a multi-dimensional feature system encompassing time, space, and external factors, and uses a spatiotemporal alignment strategy to unify the processing of time and space information related to mine pressure at the same scale. This eliminates inconsistencies caused by differences in sampling frequencies and measurement scales of different data sources, resulting in a consistent and reliable spatiotemporal feature representation. Information such as historical pressure trends of supports, interactions between supports, and mining conditions are fully integrated to generate a joint feature vector (i.e., the aforementioned fused features) for model prediction.

[0047] In step 105, the fused features are input into the pre-trained mine pressure prediction model to obtain the mine pressure prediction results for future periods output by the mine pressure prediction model.

[0048] In some embodiments of this application, the mine pressure prediction model can be a Mamba-LSTM fusion model, which has learned the mapping relationship between fusion features and roof pressure (also called mine pressure). In some embodiments, the fusion features can be input into the mine pressure prediction model; the Mamba module in the mine pressure prediction model generates long-step prediction features for future periods based on the fusion features; the LSTM module in the mine pressure prediction model performs mine pressure prediction on the long-step prediction features to obtain the mine pressure prediction result for future periods output by the mine pressure prediction model.

[0049] For example, this mine pressure prediction model combines a Mamba module and an LSTM module. The Mamba module can be used for long-step prediction of features, and the LSTM module can be used to accurately predict mine pressure in future periods based on the long-step prediction results output by the Mamba module. The fused features can be input into the Mamba module, which uses its dimensionality-invariant properties and long-sequence modeling capabilities to generate future features. The Mamba module is represented as follows:

[0050] in, These are parameters for the Mamba module; For the predicted long-step features (i.e., the long-step predicted features mentioned above). Mamba is a novel Large Language Model (LLM) architecture that integrates a Structured State-Space Sequence (S4) model to manage verbose data sequences. It combines the best features of recursive, convolutional, and continuous-time models. S4 can effectively model long-term dependencies, which enables it to handle irregular sampled data with infinite context and maintain computational efficiency throughout training and testing.

[0051] Long-step prediction features obtained from Mamba output Then, the long step size can be used to predict features. The data is input into an LSTM module to further generate mining pressure prediction results for future periods. For example, this LSTM module is represented as follows:

[0052] in, These are the parameters for the LSTM module; This indicates the final predicted pressure value; Indicates from time arrive Long step size prediction features.

[0053] Optionally, in some embodiments, a rolling prediction model for future mine pressure can be used to generate multi-step mine pressure prediction results based on recent monitoring data (including pressure monitoring data of hydraulic supports and external data). The model output prediction values ​​can include the roof pressure change trend over several subsequent time steps, which can be used to determine the intensity and timing of upcoming periodic pressure, assisting on-site personnel in taking safety measures such as support and reinforcement in advance.

[0054] In the above embodiments, a multi-dimensional feature system is constructed by fusing multi-source information such as time, space, and external features to comprehensively capture the dynamic changes and spatial coupling extrema of support pressure. Based on this, a mine pressure prediction model is adopted, integrating the advantages of the Mamba and LSTM modules to perform optimized calculations for both short-term and long-term predictions. The entire process is efficient and stable, solving the problems of insufficient accuracy and high noise interference in long-term time-series predictions using traditional methods. It provides a reliable and accurate mine pressure prediction solution, effectively mitigating the accuracy decay problem caused by increasing prediction step size, while balancing the accuracy of short-term predictions with the stability of long-term predictions. Furthermore, compared to existing methods, this application not only significantly improves the accuracy of mine pressure prediction, especially the long-term prediction capability, but also effectively reduces the negative impact of noise interference and sequence cumulative errors on the prediction results, enhancing the model's generalization ability and stability. This is beneficial for early warning of periodic roof pressure in mines, reducing the risk of mine safety accidents, improving mine production safety and mining efficiency, and has significant economic and social benefits.

[0055] It should be noted that the above-mentioned mine pressure prediction model can be pre-trained. For example, the mine pressure prediction model can be trained based on historical pressure monitoring data of hydraulic supports and historical external data. Figure 2 This is a flowchart illustrating the training method for the mine pressure prediction model provided in the embodiments of this application.

[0056] It should be noted that the execution entity of the mine pressure prediction model training method in this application embodiment can be a mine pressure prediction model training device, which can be implemented by software and / or hardware, and can be configured in an electronic device. For example, the electronic device may include, but is not limited to, a terminal, a server, etc.

[0057] like Figure 2 As shown, the training method for this mine pressure prediction model may include, but is not limited to, the following steps.

[0058] In step 201, historical pressure monitoring data of the hydraulic supports operating in the fully mechanized coal mining face and adjacent hydraulic supports are obtained.

[0059] Optionally, the implementation of step 201 can be found in the implementation of step 101 above, and will not be repeated here.

[0060] In step 202, external data related to the mine pressure of the fully mechanized coal mining face is acquired. The external data includes at least one of the following: the advance speed, position, and direction of the coal mining machine.

[0061] Optionally, the implementation of step 202 can be found in the implementation of step 102 above, and will not be repeated here.

[0062] In step 203, feature extraction is performed based on historical pressure monitoring data of hydraulic supports to obtain time-domain features, feature extraction is performed based on historical pressure monitoring data of neighboring hydraulic supports to obtain spatial features, and feature extraction is performed based on external data to obtain external features.

[0063] Optionally, the implementation of step 203 can be found in the implementation of step 103 above, and will not be repeated here.

[0064] In step 204, the temporal features, spatial features, and external features are fused to obtain fused features.

[0065] Optionally, the implementation of step 204 can be found in the implementation of step 104 above, and will not be repeated here.

[0066] In step 205, the fused features are input into the mine pressure prediction model to be trained to obtain the mine pressure prediction results for future periods output by the mine pressure prediction model; wherein, the mine pressure prediction model is a Mamba-LSTM fusion model.

[0067] Optionally, the implementation of step 205 can be found in the implementation of step 105 above, and will not be repeated here.

[0068] In step 206, a loss value is generated based on the mine pressure prediction results and the actual mine pressure monitoring values, and the mine pressure prediction model is trained based on the loss value.

[0069] In the embodiments of this application, the mining pressure prediction results are minimized. Compared with actual monitoring values ​​of mine pressure The error between them is expressed using a loss function. This allows the parameters of the mine pressure prediction model to converge.

[0070] In the above embodiments, this application significantly improves the model's ability to characterize the complex nonlinear characteristics of mine pressure by deeply analyzing the temporal and spatial coupling relationship contained in the hydraulic support pressure monitoring data, uniformly processing the spatiotemporal features from different sources and enhancing their consistency; by combining Mamba and LSTM to obtain the mine pressure prediction model architecture, it can achieve high accuracy and stability in both short-term and long-term predictions, enabling early warning of roof pressure in fully mechanized mining faces, ensuring safe mine production and improving mining efficiency.

[0071] Figure 3 This is a block diagram of a mine pressure spatiotemporal prediction device based on a Mamba-LSTM fusion model, provided in an embodiment of this application. Figure 3As shown, the mine pressure spatiotemporal prediction device may include: a first acquisition module 301, a second acquisition module 302, a feature extraction module 303, a feature fusion module 304, and a prediction module 305.

[0072] The first acquisition module 301 is used to acquire the pressure monitoring data of the hydraulic supports and adjacent hydraulic supports working in the fully mechanized coal mining face.

[0073] The second acquisition module 302 is used to acquire external data related to the mine pressure of the fully mechanized coal mining face. The external data includes at least one of the following: the advance speed, position, and direction of the coal mining machine.

[0074] The feature extraction module 303 is used to extract features based on the pressure monitoring data of the hydraulic support to obtain time-domain features, extract features based on the pressure monitoring data of the adjacent hydraulic support to obtain spatial features, and extract features based on external data to obtain external features.

[0075] The feature fusion module 304 is used to fuse temporal features, spatial features and external features to obtain fused features.

[0076] The prediction module 305 is used to input the fused features into the pre-trained mine pressure prediction model to obtain the mine pressure prediction results for future periods output by the mine pressure prediction model; wherein, the mine pressure prediction model is a Mamba-LSTM fusion model, and the mine pressure prediction model has learned the mapping relationship between the fused features and the roof pressure.

[0077] In some embodiments, the feature extraction module 303 is used to: perform window sliding operation on the pressure monitoring data of the hydraulic support in the manner of sliding window, and calculate the mean, variance, maximum and minimum values ​​of the pressure monitoring data within the window, and use the mean, variance, maximum and minimum values ​​of the pressure monitoring data within the window as time domain features.

[0078] In some embodiments, the feature extraction module 303 is used to: calculate the average pressure, pressure difference, maximum pressure and minimum pressure of each adjacent hydraulic support based on the pressure monitoring data of each adjacent hydraulic support; construct spatial features based on the average pressure, pressure difference, maximum pressure and minimum pressure of each adjacent hydraulic support, and use the average pressure, pressure difference, maximum pressure and minimum pressure of each adjacent hydraulic support to quantify the spatial correlation between hydraulic supports.

[0079] In some embodiments, the feature extraction module 303 is further configured to: analyze the spatial distribution of hydraulic supports and adjacent hydraulic supports, and normalize the position information of each hydraulic support and adjacent hydraulic supports, so that the spatial feature analysis takes into account the relative position changes of the coal mining machine between hydraulic supports during the coal mining process, and to evaluate the impact of different hydraulic supports on the changes in mine pressure in a consistent manner; wherein, the formula used for normalization is as follows:

[0080] in, This represents the original absolute position value of the hydraulic support in space; This represents the left endpoint of the interval containing the original position value of the hydraulic support; This represents the right endpoint of the interval containing the original position value of the hydraulic support; The final value is [0,1].

[0081] In some embodiments, the mine pressure spatiotemporal prediction device may further include a spatiotemporal alignment module. This module is used to: normalize the physical positions of hydraulic supports operating on the longwall mining face of a coal mine, converting the spatial coordinates of the hydraulic supports into a uniform scale; and perform spatiotemporal alignment processing on the movement of the coal mining machine between the hydraulic supports, where the movement of the coal mining machine defines a step size, each step size consisting of multiple time steps. Based on the workflow of the coal mining machine, a corresponding spatial position is assigned to each time step. Through the mapping relationship between time steps and spatial step sizes, the temporal characteristics and spatial position information in the pressure monitoring data and external data are spatiotemporally aligned.

[0082] In some embodiments, the prediction module is used to: input the fused features into the mine pressure prediction model; generate long-step prediction features for future periods based on the fused features using the Mamba module in the mine pressure prediction model; and perform mine pressure prediction on the long-step prediction features using the LSTM module in the mine pressure prediction model to obtain the mine pressure prediction results for future periods output by the mine pressure prediction model.

[0083] It should be noted that the foregoing explanation of the embodiment of the mineral pressure spatiotemporal prediction method also applies to the mineral pressure spatiotemporal prediction device of this embodiment, and will not be repeated here.

[0084] Figure 4 A block diagram of a mine pressure prediction model training device provided in an embodiment of this application. Figure 4 As shown, the mine pressure prediction model training device may include: a first acquisition module 401, a second acquisition module 402, a feature extraction module 403, a feature fusion module 404, a prediction module 405, and a training module 406.

[0085] The first acquisition module 401 is used to acquire historical pressure monitoring data of the hydraulic supports working in the fully mechanized coal mining face and the adjacent hydraulic supports.

[0086] The second acquisition module 402 is used to acquire external data related to the mine pressure of the fully mechanized coal mining face. The external data includes at least one of the following: the advance speed, position, and direction of the coal mining machine.

[0087] The feature extraction module 403 is used to extract features based on historical pressure monitoring data of hydraulic supports to obtain time-domain features, extract features based on historical pressure monitoring data of neighboring hydraulic supports to obtain spatial features, and extract features based on external data to obtain external features.

[0088] The feature fusion module 404 is used to fuse temporal features, spatial features and external features to obtain fused features.

[0089] The prediction module 405 is used to input the fused features into the mine pressure prediction model to be trained, and obtain the mine pressure prediction results for future periods output by the mine pressure prediction model; wherein, the mine pressure prediction model is a Mamba-LSTM fusion model.

[0090] Training module 406 is used to generate loss values ​​based on the mine pressure prediction results and actual mine pressure monitoring values, and to train the mine pressure prediction model based on the loss values.

[0091] It should be noted that the foregoing explanation of the embodiment of the mine pressure prediction model training method also applies to the mine pressure prediction model training device of this embodiment, and will not be repeated here.

[0092] Figure 5 A block diagram of an electronic device provided in an embodiment of this application. (e.g.) Figure 5 As shown, the electronic device may include at least one processor 501 and a memory 502. The memory 502 is communicatively connected to the at least one processor 501. The memory 502 stores instructions executable by the at least one processor 501, which, when executed by the at least one processor 501, enable the at least one processor 501 to execute the spatiotemporal prediction method for mining pressure or the training method for mining pressure prediction models provided in this application.

[0093] Based on the above embodiments, this application also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed on an electronic device, causes the electronic device to execute the spatiotemporal prediction method for mining pressure or the training method for mining pressure prediction models provided in this application.

[0094] To implement the above embodiments, this application also proposes a program product, including at least one of a program and instructions, wherein when the at least one of the program and instructions is executed by a processor, it implements the spatiotemporal prediction method for mining pressure or the training method for mining pressure prediction models in the above method embodiments.

[0095] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0096] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0097] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0098] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0099] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0100] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0101] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0102] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A mine pressure spatio-temporal prediction method based on a Mamba-LSTM fusion model, characterized in that, include: Acquire pressure monitoring data of the hydraulic supports operating in the fully mechanized coal mining face and adjacent hydraulic supports; Acquire external data associated with the mine pressure at the fully mechanized coal mining face, wherein the external data includes at least one of the following: the advance speed, position, and direction of the coal mining machine; Feature extraction based on the pressure monitoring data of the hydraulic support to obtain time-domain features includes: performing window sliding operation on the pressure monitoring data of the hydraulic support using a sliding window method, and calculating the mean, variance, maximum value and minimum value of the pressure monitoring data within the window, and using the mean, variance, maximum value and minimum value of the pressure monitoring data within the window as the time-domain features; Feature extraction is performed based on the pressure monitoring data of the adjacent hydraulic supports to obtain spatial features, including: calculating the average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support based on the pressure monitoring data of each adjacent hydraulic support; constructing the spatial features based on the average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support, wherein the average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support are used to quantify the spatial correlation between hydraulic supports; External features are obtained by extracting features from the external data. The temporal features, spatial features, and external features are fused to obtain fused features; The process of inputting the fused features into a pre-trained mining pressure prediction model to obtain the mining pressure prediction results for future periods output by the mining pressure prediction model includes: inputting the fused features into the mining pressure prediction model; generating long-step prediction features for future periods based on the fused features using the Mamba module in the mining pressure prediction model; and performing mining pressure prediction on the long-step prediction features using the LSTM module in the mining pressure prediction model to obtain the mining pressure prediction results for future periods output by the mining pressure prediction model. The mine pressure prediction model is based on the Mamba-LSTM fusion model, which has learned the mapping relationship between the fusion features and the roof pressure.

2. The method of claim 1, wherein, The method further includes: The spatial distribution of hydraulic supports and adjacent hydraulic supports is analyzed, and the positional information of each hydraulic support and its adjacent supports is normalized to ensure that the spatial feature analysis takes into account the relative positional changes of the coal mining machine among the hydraulic supports during the coal mining process, thus consistently evaluating the impact of different hydraulic supports on mine pressure changes. The formula used for the normalization process is as follows: in, This represents the original absolute position value of the hydraulic support in space; This represents the left endpoint of the interval containing the original position value of the hydraulic support; This represents the right endpoint of the interval containing the original position value of the hydraulic support; The final value is [0,1].

3. The method of claim 1, wherein, The method further includes: The physical position of the hydraulic support working on the fully mechanized mining face of the coal mine is normalized to convert the spatial coordinates of the hydraulic support into a uniform scale. The movement of the coal mining machine between the hydraulic supports is defined by a step length, each step length consisting of multiple time steps. Through the working process of the coal mining machine, a corresponding spatial position is assigned to each time step. Through the mapping relationship between time steps and spatial step lengths, the time characteristics and spatial position information in the pressure monitoring data and the external data are spatiotemporally aligned.

4. A mine pressure prediction model training method, characterized in that, include: Acquire historical pressure monitoring data for hydraulic supports operating in the fully mechanized coal mining face and for adjacent hydraulic supports. Acquire external data associated with the mine pressure at the fully mechanized coal mining face, wherein the external data includes at least one of the following: the advance speed, position, and direction of the coal mining machine; Feature extraction based on the historical pressure monitoring data of the hydraulic support to obtain time-domain features includes: performing window sliding operation on the pressure monitoring data of the hydraulic support using a sliding window method, and calculating the mean, variance, maximum value and minimum value of the pressure monitoring data within the window, and using the mean, variance, maximum value and minimum value of the pressure monitoring data within the window as the time-domain features; Feature extraction is performed based on the historical pressure monitoring data of the adjacent hydraulic supports to obtain spatial features, including: calculating the average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support based on the pressure monitoring data of each adjacent hydraulic support; constructing the spatial features based on the average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support, wherein the average pressure, pressure difference, maximum pressure, and minimum pressure of each adjacent hydraulic support are used to quantify the spatial correlation between hydraulic supports; External features are obtained by extracting features from the external data. The temporal features, spatial features, and external features are fused to obtain fused features; The process of inputting the fused features into a mine pressure prediction model to obtain the mine pressure prediction result for the future time period output by the mine pressure prediction model includes: inputting the fused features into the mine pressure prediction model; generating long-step prediction features for the future time period based on the fused features using the Mamba module in the mine pressure prediction model; and performing mine pressure prediction on the long-step prediction features using the LSTM module in the mine pressure prediction model to obtain the mine pressure prediction result for the future time period output by the mine pressure prediction model; wherein, the mine pressure prediction model is a Mamba-LSTM fusion model; Based on the predicted mine pressure results and the actual monitored mine pressure values, a loss value is generated, and the mine pressure prediction model is trained based on the loss value.

5. A mine pressure space-time prediction device based on a Mamba-LSTM fusion model, characterized in that, include: The first acquisition module is used to acquire the pressure monitoring data of the hydraulic supports working in the fully mechanized coal mining face and the adjacent hydraulic supports. The second acquisition module is used to acquire external data related to the mine pressure of the fully mechanized coal mining face, wherein the external data includes at least one of the advancing speed, position, and direction of the coal mining machine; The feature extraction module is used to extract features based on the pressure monitoring data of the hydraulic supports to obtain time-domain features, extract features based on the pressure monitoring data of neighboring hydraulic supports to obtain spatial features, and extract features based on the external data to obtain external features. Specifically, extracting features based on the pressure monitoring data of the hydraulic supports to obtain time-domain features includes: performing a sliding window operation on the pressure monitoring data of the hydraulic supports, calculating the mean, variance, maximum value, and minimum value of the pressure monitoring data within the window, and using the mean, variance, maximum value, and minimum value of the pressure monitoring data within the window as the time-domain features. Extracting features based on the pressure monitoring data of neighboring hydraulic supports to obtain spatial features includes: calculating the mean pressure, pressure difference, maximum pressure value, and minimum pressure value of each neighboring hydraulic support based on the pressure monitoring data of each neighboring hydraulic support; constructing the spatial features based on the mean pressure, pressure difference, maximum pressure value, and minimum pressure value of each neighboring hydraulic support, where the mean pressure, pressure difference, maximum pressure value, and minimum pressure value of each neighboring hydraulic support are used to quantify the spatial correlation between the hydraulic supports. The feature fusion module is used to fuse the temporal features, the spatial features, and the external features to obtain fused features; The prediction module is used to input the fused features into a pre-trained mine pressure prediction model to obtain the mine pressure prediction result for the future period output by the mine pressure prediction model. This includes: inputting the fused features into the mine pressure prediction model; generating long-step prediction features for the future period based on the fused features using the Mamba module in the mine pressure prediction model; and performing mine pressure prediction on the long-step prediction features using the LSTM module in the mine pressure prediction model to obtain the mine pressure prediction result for the future period output by the mine pressure prediction model. The mine pressure prediction model is a Mamba-LSTM fusion model, and the mine pressure prediction model has learned the mapping relationship between the fused features and the roof pressure. 6.A mining pressure prediction model training apparatus, characterized in that, include: The first acquisition module is used to acquire historical pressure monitoring data of the hydraulic supports working in the fully mechanized coal mining face and the adjacent hydraulic supports. The second acquisition module is used to acquire external data related to the mine pressure of the fully mechanized coal mining face, wherein the external data includes at least one of the advancing speed, position, and direction of the coal mining machine; The feature extraction module is used to extract features based on the historical pressure monitoring data of the hydraulic support to obtain time-domain features, extract features based on the historical pressure monitoring data of neighboring hydraulic supports to obtain spatial features, and extract features based on the external data to obtain external features. Specifically, extracting features based on the historical pressure monitoring data of the hydraulic support to obtain time-domain features includes: using a sliding window to slide the pressure monitoring data of the hydraulic support, and calculating the mean, variance, maximum, and minimum values ​​of the pressure monitoring data within the window, using these values ​​as the time-domain features. Extracting features based on the historical pressure monitoring data of neighboring hydraulic supports to obtain spatial features includes: calculating the mean pressure, pressure difference, maximum pressure, and minimum pressure of each neighboring hydraulic support based on their pressure monitoring data; constructing the spatial features based on these values, where the mean pressure, pressure difference, maximum pressure, and minimum pressure of each neighboring hydraulic support are used to quantify the spatial correlation between the hydraulic supports. The feature fusion module is used to fuse the temporal features, the spatial features, and the external features to obtain fused features; A prediction module is used to input the fused features into a mine pressure prediction model to be trained, and obtain the mine pressure prediction result for the future time period output by the mine pressure prediction model. This includes: inputting the fused features into the mine pressure prediction model; generating long-step prediction features for the future time period based on the fused features using a Mamba module in the mine pressure prediction model; and performing mine pressure prediction on the long-step prediction features using an LSTM module in the mine pressure prediction model to obtain the mine pressure prediction result for the future time period output by the mine pressure prediction model. The mine pressure prediction model is a Mamba-LSTM fusion model. The training module is used to generate a loss value based on the predicted mine pressure results and the actual monitored mine pressure values, and to train the mine pressure prediction model based on the loss value.

7. An electronic device, comprising: include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-3 and 4.

Images