Coal seam gas anomaly detection method and system based on machine learning, electronic equipment and storage medium

Abstract

The invention discloses a coal seam gas anomaly detection method and system based on machine learning, electronic equipment and a storage medium, and the method comprises the steps: obtaining the real-time gas concentration data of a plurality of concentration sensors in a coal mine, and carrying out the time-space sequence analysis of the real-time gas concentration data, the change trend of the gas concentration in time and space dimensions is obtained; calculating the gas concentration gradient of the position of each concentration sensor within the specified time period based on the change trend to obtain a gradient change sequence; aiming at the gradient change sequence, analyzing a dynamic evolution mode of the gradient in a time dimension by utilizing a long-short-term memory network model to obtain gradient dynamic characteristics; acquiring historical gradient change characteristics and historical construction disturbance characteristics of the positions where the concentration sensors are located, and constructing a disturbance distinguishing model; the gradient dynamic characteristics are analyzed based on a disturbance distinguishing model, whether gradient changes are caused by normal construction disturbance or not is judged, and if not, it is judged that gas concentration changes are abnormal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coal seam gas detection, and in particular to a coal seam gas anomaly detection method and system based on machine learning, an electronic device and a storage medium. BACKGROUND

[0002] Coal seam gas anomaly detection is a key research direction in the field of coal mine safety production, directly related to mine operation safety and personnel life protection. As the most harmful gas in coal mines, gas concentration anomalies can cause explosions or poisoning accidents, so accurate identification of gas anomalies is the core of disaster prevention. However, existing detection methods lack adaptability in complex geological environments, especially when faced with frequent changes in coal seam geological structure and uneven gas distribution. Traditional methods often fail to capture dynamic change characteristics, relying on fixed thresholds or single sensor data, making it difficult to respond to rapid fluctuations in gas concentration in time and space, resulting in frequent false negatives or false positives, increasing safety hazards.

[0003] In actual mine environments, the gradient change of gas concentration is an important representation of abnormal events. Gradient changes reflect the unevenness of gas distribution in space and the dynamic evolution over time, but existing technologies face two major challenges in handling gradient features. First, the spatiotemporal dynamics of gas concentration gradients are difficult to accurately capture. Gas distribution in mines is influenced by multiple factors such as geological conditions and ventilation conditions, and the concentration gradient may change dramatically in a short time. For example, near the excavation face, gas concentration may quickly rise due to sudden outflow from local geological fissures, but existing methods cannot accurately distinguish whether this mutation is abnormal. Second, there is insufficient correlation analysis between gradient changes and abnormal events. Gradient mutations may be caused by normal ventilation adjustments or may be a precursor to abnormal gas outflow, and lack of effective models to distinguish between the two reduces the reliability of early warning systems in complex scenarios. For example, during excavation in a mine, gas concentration suddenly increased from normal values within a few minutes, but existing systems could not determine whether this was an abnormal outflow caused by geological fissures or normal construction disturbance, thus missing the best early warning opportunity.

[0004] Therefore, how to analyze the gradient change trend of gas concentration in space and time dimensions, accurately identify abnormal mutation points and establish their correlation with gas abnormal events, becomes a key problem to improve detection accuracy and early warning capability. SUMMARY

[0005] To solve the above technical problems, the present application provides a coal seam gas anomaly detection method based on machine learning, comprising the following steps:

[0006] Real-time gas concentration data from several concentration sensors in a coal mine are acquired, and spatiotemporal sequence analysis is performed on the real-time gas concentration data to obtain the changing trends of gas concentration in time and space.

[0007] Based on the aforementioned trend, a sliding window is used to calculate the gas concentration gradient at the location of each concentration sensor within a specified time period, thereby obtaining a gradient change sequence.

[0008] For the gradient change sequence, the dynamic evolution pattern of the gradient in the time dimension is analyzed using a long short-term memory network model to obtain the gradient dynamic characteristics.

[0009] Historical gradient change characteristics and historical construction disturbance characteristics of the locations of each concentration sensor are obtained, and a disturbance discrimination model is constructed based on the historical gradient change characteristics and the historical construction disturbance characteristics.

[0010] The gradient dynamic characteristics are analyzed based on the disturbance differentiation model to determine whether the gradient change is caused by normal construction disturbance. If not, the gas concentration change is determined to be abnormal.

[0011] Preferably, the method for obtaining the trend of change includes:

[0012] Real-time gas concentration data from several concentration sensors in a coal mine are acquired, and the real-time gas concentration data is cleaned and missing values ​​are filled in to obtain a complete gas concentration dataset.

[0013] Based on the complete gas concentration dataset, a gas concentration distribution model in the spatial dimension is constructed using the Kriging interpolation method, and the spatial distribution characteristics of gas are obtained using the gas concentration distribution model.

[0014] Time series data are extracted from the spatial distribution characteristics of the gas, and the ARIMA model is used to analyze the time series data to obtain the temporal variation trend of gas concentration.

[0015] The spatial distribution characteristics of the gas and the temporal variation trend of the gas concentration are weighted and fused to obtain the variation trend of the gas concentration in both time and space.

[0016] Preferably, the method for obtaining the gradient change sequence includes:

[0017] Based on the changing trend, the time series of gas concentration at the location of each concentration sensor is obtained;

[0018] Based on the gas concentration time series, the concentration change rate within a specified time period is calculated using a sliding window to obtain the concentration gradient;

[0019] If the concentration gradient exceeds a first preset threshold, then a local weighted regression is performed on the time series of the concentration sensor to obtain the smoothed gradient change sequence.

[0020] Preferably, the method for obtaining the gradient dynamic features includes:

[0021] Obtain the gradient change sequence and extract the gradient values ​​in the time dimension to obtain the initial sequence dataset;

[0022] A long short-term memory network model is constructed and trained, and the dynamic evolution of the initial sequence dataset in the time dimension is analyzed to obtain the gradient dynamic features.

[0023] The present invention also provides a coal seam gas anomaly detection system based on machine learning. The system applies the above-mentioned method and includes: a spatiotemporal sequence analysis module, a gradient change calculation module, a gradient feature analysis module, a model construction module, and an anomaly judgment module.

[0024] The spatiotemporal sequence analysis module is used to acquire real-time gas concentration data from several concentration sensors in the coal mine, and to perform spatiotemporal sequence analysis on the real-time gas concentration data to obtain the changing trend of gas concentration in time and space.

[0025] Based on the changing trend, the gradient change calculation module uses a sliding window to calculate the gas concentration gradient at the location of each concentration sensor within a specified time period, thereby obtaining a gradient change sequence.

[0026] The gradient feature analysis module is used to analyze the dynamic evolution pattern of the gradient in the time dimension using a long short-term memory network model for the gradient change sequence, and obtain the gradient dynamic features.

[0027] The model building module is used to obtain the historical gradient change characteristics and historical construction disturbance characteristics of the location of each concentration sensor, and to build a disturbance differentiation model based on the historical gradient change characteristics and the historical construction disturbance characteristics;

[0028] The anomaly detection module analyzes the gradient dynamic characteristics based on the disturbance differentiation model to determine whether the gradient change is caused by normal construction disturbance. If not, the gas concentration change is determined to be abnormal.

[0029] Preferably, the workflow of the spatiotemporal sequence analysis module includes:

[0030] Real-time gas concentration data from several concentration sensors in a coal mine are acquired, and the real-time gas concentration data is cleaned and missing values ​​are filled in to obtain a complete gas concentration dataset.

[0031] Based on the complete gas concentration dataset, a gas concentration distribution model in the spatial dimension is constructed using the Kriging interpolation method, and the spatial distribution characteristics of gas are obtained using the gas concentration distribution model.

[0032] Time series data are extracted from the spatial distribution characteristics of the gas, and the ARIMA model is used to analyze the time series data to obtain the temporal variation trend of gas concentration.

[0033] The spatial distribution characteristics of the gas and the temporal variation trend of the gas concentration are weighted and fused to obtain the variation trend of the gas concentration in both time and space.

[0034] Preferably, the workflow of the gradient change calculation module includes:

[0035] Based on the changing trend, the time series of gas concentration at the location of each concentration sensor is obtained;

[0036] Based on the gas concentration time series, the concentration change rate within a specified time period is calculated using a sliding window to obtain the concentration gradient;

[0037] If the concentration gradient exceeds a first preset threshold, then a local weighted regression is performed on the time series of the concentration sensor to obtain the smoothed gradient change sequence.

[0038] Preferably, the workflow of the gradient feature analysis module includes:

[0039] Obtain the gradient change sequence and extract the gradient values ​​in the time dimension to obtain the initial sequence dataset;

[0040] A long short-term memory network model is constructed and trained, and the dynamic evolution of the initial sequence dataset in the time dimension is analyzed to obtain the gradient dynamic features.

[0041] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the aforementioned machine learning-based coal seam gas anomaly detection method.

[0042] The present invention also provides a computer-readable storage medium storing a computer program that, when executed, implements the aforementioned machine learning-based coal seam gas anomaly detection method.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] This invention discloses a method for accurately predicting the spatiotemporal variations of methane concentration and the risk of abnormal outbursts in mines. It integrates spatiotemporal series analysis, sliding window technology, and long short-term memory (LSTM) networks to construct a complete technical chain from data acquisition to anomaly diagnosis. First, spatiotemporal series analysis captures the temporal and spatial trends of methane concentration. Then, sliding window technology is used to calculate the concentration gradient sequence at each sensor location. Subsequently, LTM networks are employed to mine the dynamic evolution characteristics of the gradient and determine whether the gradient changes are caused by normal construction activities. This invention significantly improves the accuracy and real-time performance of predicting abnormal methane outbursts through multi-model collaboration and multi-source data fusion, providing efficient technical support for mine safety management. Attached Figure Description

[0045] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0047] Figure 2 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention.

[0048] Explanation of reference numerals in the attached figures:

[0049] 1010, Processor; 1020, Memory; 1030, Input / Output Interface; 1040, Communication Interface; 1050, Bus. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0052] Example 1

[0053] In this embodiment, as Figure 1 As shown, a machine learning-based method for detecting coal seam gas anomalies includes the following steps:

[0054] S1. Obtain real-time gas concentration data from several concentration sensors in the coal mine, perform spatiotemporal sequence analysis on the real-time gas concentration data, and obtain the changing trend of gas concentration in time and space.

[0055] The method for obtaining the changing trend includes: acquiring real-time gas concentration data from several concentration sensors in the coal mine; cleaning and imputing missing values ​​in the real-time gas concentration data to obtain a complete gas concentration dataset; constructing a spatial gas concentration distribution model based on the complete gas concentration dataset using the Kriging interpolation method, and obtaining the spatial distribution characteristics of gas using the gas concentration distribution model; extracting time series data from the spatial distribution characteristics of gas, analyzing the time series data using the ARIMA model to obtain the temporal changing trend of gas concentration; and weighted fusion of the spatial distribution characteristics of gas and the temporal changing trend of gas concentration to obtain the changing trend of gas concentration in both time and space.

[0056] In this embodiment, when acquiring real-time gas concentration data from multiple sensors within the mine, a multi-point distributed sensor network is deployed. Specifically, 10 sensors are arranged in key areas such as mine roadways and mining faces, collecting gas concentration data once per second, expressed as a percentage. The data is then cleaned using a box plot method to remove outliers. For example, if a sensor collects a concentration value of 10.5%, which is significantly higher than the normal range of 0.5%-2%, it is considered an anomaly and removed. If the cleaned dataset contains missing values, for short-term missing values ​​from a single sensor, interpolation (linear or spline) from nearby time points or simple prediction based on the sensor's historical data (such as a moving average) can be used to obtain a complete gas concentration dataset. Based on a complete gas concentration dataset, a spatial gas concentration distribution model is constructed using the Kriging interpolation method. The Kriging method estimates the concentration at unknown points through spatial autocorrelation. For example, in a certain area of ​​the mine, if the concentrations at sensor points A and B are 1.2% and 1.5% respectively, Kriging can estimate the concentration at the intermediate unknown point to be approximately 1.35%, generating a continuous concentration distribution map and thus obtaining the spatial distribution characteristics of gas. From the gas spatial distribution characteristics generated by Kriging, for a specific location (such as a sensor location or a grid point of interest), the predicted concentration value sequence over a period of time (minutes, hours, days) is extracted; the sequence of statistical indicators such as average concentration, maximum concentration, and concentration standard deviation over time for a certain area (such as the entire working face or a certain roadway) is also extracted. Then, the ARIMA model is used to analyze the time series data to obtain the temporal trend of gas concentration. The spatial distribution characteristics and the temporal trend of gas concentration are then weighted and fused to obtain the temporal and spatial trends of gas concentration.

[0057] S2. Based on the changing trend, the gas concentration gradient at the location of each concentration sensor within a specified time period is calculated using a sliding window to obtain the gradient change sequence.

[0058] The method for obtaining the gradient change sequence includes: obtaining the gas concentration time series of each concentration sensor location based on the change trend; calculating the concentration change rate within a specified time period using a sliding window based on the gas concentration time series to obtain the concentration gradient; if the concentration gradient exceeds a first preset threshold, performing local weighted regression on the time series of the concentration sensor to obtain a smoothed gradient change sequence.

[0059] In this embodiment, the time series data of gas concentration from each concentration sensor (e.g., 10 points) is extracted from the spatiotemporal fusion change trend obtained in step S1 and denoted as C. i(t), where i is the sensor number and t is the timestamp; the gas concentration time series data is normalized (e.g., Z-score normalization) to eliminate the dimensional differences between different sensors, resulting in the gas concentration time series. Based on the characteristics of the coal mine production cycle (e.g., mining machine operation cycle, ventilation cycle), the sliding window size T and step size Δt are set, and the data within the window are arranged in chronological order to form a subsequence C. i (t k ), t k Within the window [tT, t], calculate the gas concentration gradient G for each sensor i. i (t), characterizing the rate and direction of concentration change:

[0060]

[0061] Set a first preset threshold θ1 (e.g., 0.5% / min, according to coal mine safety regulations). If |G i (t)|>θ1, perform a weighted fitting on the time period [tT, t] corresponding to this concentration, assigning higher weights to neighboring data points to generate a smoothed gradient. Then, concatenate the smoothed gradients in chronological order to form the smoothed gradient change sequence for each sensor.

[0062] S3. For gradient change sequences, the dynamic evolution pattern of the gradient in the time dimension is analyzed using a long short-term memory network model to obtain the dynamic characteristics of the gradient.

[0063] Methods for obtaining gradient dynamic features include: acquiring gradient change sequences and extracting gradient values ​​in the time dimension to obtain an initial sequence dataset; constructing and training a long short-term memory network model and analyzing the dynamic evolution of the initial sequence dataset in the time dimension to obtain gradient dynamic features.

[0064] In this embodiment, the Long Short-Term Memory (LSTM) network model employs a multi-layer LSTM structure. The input layer has a dimension of 10 (corresponding to 10 sensors), the hidden layer has 32 neurons, and the output layer is mapped to 10-dimensional gradient predictions through a fully connected layer. A Dropout layer is introduced to prevent overfitting, and Layer Normalization is added to accelerate training convergence. The Mean Squared Error (MSE) is used as the loss function, the Adam optimizer is selected, and the training data is divided into a 70% training set, a 15% validation set, and a 15% test set.

[0065] Specifically, the gradient change sequence is obtained, and gradient values ​​in the time dimension are extracted from the original data to generate an initial sequence dataset. The trained network model is then used to analyze the dynamic evolution of the initial sequence dataset in the time dimension, extracting gradient dynamic features. If the fluctuation amplitude of the gradient dynamic features exceeds a preset threshold, the features are normalized to generate a normalized feature set. Based on the normalized feature set, principal component analysis is used to reduce the dimensionality of the feature data, obtaining dimensionality-reduced feature vectors. These dimensionality-reduced feature vectors are then used to analyze the temporal evolution of the gradient dynamic patterns, generating dynamic pattern descriptions. Finally, cluster analysis is used to classify the dynamic pattern descriptions, obtaining the classification results of the gradient dynamic features.

[0066] S4. Obtain the historical gradient change characteristics and historical construction disturbance characteristics of the locations of each concentration sensor, and construct a disturbance differentiation model based on the historical gradient change characteristics and historical construction disturbance characteristics.

[0067] In this embodiment, complete gradient dynamic feature sequences (including original gradient values, normalized features, principal component dimensionality reduction vectors, and clustering classification labels) for each sensor location over the past N days (e.g., 30 days) are extracted from the historical gradient dynamic feature library. These features are simultaneously acquired from historical construction disturbance features in the mine operation log, including: start-up and shutdown times of mining machinery, location coordinates, power load curves, blasting operation time, location, explosive equivalent, ventilation system wind speed / volume adjustment records, and spatiotemporal markers for engineering activities such as roadway support / equipment movement. Using a 1-minute time granularity, gradient features and construction disturbance features are matched by timestamp and sensor location to construct a labeled joint dataset. An initial disturbance discrimination model is constructed using a gradient boosting decision tree (LightGBM) as the main architecture, and the initial disturbance discrimination model is trained using the labeled joint dataset to obtain the final disturbance discrimination model.

[0068] S5. Analyze the gradient dynamic characteristics based on the disturbance differentiation model to determine whether the gradient change is caused by normal construction disturbance. If not, determine that the gas concentration change is abnormal.

[0069] In this embodiment, gradient dynamic feature data is input into a pre-built and trained disturbance discrimination model. The input gradient dynamic features are analyzed and processed to determine whether the gradient change is caused by normal construction disturbance. For example, in an area undergoing mining operations, normal construction operations such as starting and stopping mining machinery and adjusting the ventilation system will cause changes in the methane concentration gradient. If the model analysis finds that the current gradient change pattern highly matches the gradient change pattern caused by normal construction disturbance, then the gradient change is determined to be caused by normal construction disturbance. Conversely, if the gradient change pattern does not match the characteristics of normal construction disturbance, exceeds a certain threshold range, or exhibits characteristics that do not conform to conventional construction patterns, then the methane concentration change can be determined to be abnormal.

[0070] Once an abnormal change in methane concentration is detected, an alarm will be immediately issued, reminding relevant personnel to take appropriate measures, including increasing the monitoring frequency of methane concentration in the area, suspending related operations, adjusting the ventilation system, and organizing the evacuation of personnel, to ensure safety underground in the coal mine. Simultaneously, the abnormal situation will be recorded for further analysis and research, continuously optimizing the disturbance differentiation model and the methane monitoring system to improve the accuracy of identifying abnormal changes in methane concentration and the timeliness of response.

[0071] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0072] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, it should be understood that the sequence number of each step in the above embodiments does not imply the order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. The actions or steps recorded in the claims can be performed in a different order than that in the above embodiments and can still achieve the desired result. In addition, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0073] Example 2

[0074] In this embodiment, a machine learning-based coal seam gas anomaly detection system includes: a spatiotemporal sequence analysis module, a gradient change calculation module, a gradient feature analysis module, a model construction module, and an anomaly judgment module.

[0075] The spatiotemporal sequence analysis module is used to acquire real-time gas concentration data from several concentration sensors in the coal mine, perform spatiotemporal sequence analysis on the real-time gas concentration data, and obtain the changing trend of gas concentration in time and space.

[0076] The workflow of the spatiotemporal series analysis module includes: acquiring real-time gas concentration data from several concentration sensors in the coal mine; cleaning and imputing missing values ​​in the real-time gas concentration data to obtain a complete gas concentration dataset; constructing a spatial gas concentration distribution model based on the complete gas concentration dataset using the Kriging interpolation method, and obtaining the spatial distribution characteristics of gas using the gas concentration distribution model; extracting time series data from the spatial distribution characteristics of gas, analyzing the time series data using the ARIMA model to obtain the temporal variation trend of gas concentration; and weighted fusion of the spatial distribution characteristics of gas and the temporal variation trend of gas concentration to obtain the variation trend of gas concentration in both time and space.

[0077] The gradient change calculation module calculates the gas concentration gradient at the location of each concentration sensor within a specified time period based on the change trend and using a sliding window, thus obtaining the gradient change sequence.

[0078] The workflow of the gradient change calculation module includes: obtaining the gas concentration time series of each concentration sensor location based on the change trend; calculating the concentration change rate within a specified time period using a sliding window based on the gas concentration time series to obtain the concentration gradient; if the concentration gradient exceeds a first preset threshold, performing local weighted regression on the time series of the concentration sensor to obtain a smoothed gradient change sequence.

[0079] The gradient feature analysis module is used to analyze the dynamic evolution pattern of gradients over time using a long short-term memory network model to obtain gradient dynamic features for gradient change sequences.

[0080] The workflow of the gradient feature analysis module includes: acquiring gradient change sequences and extracting gradient values ​​in the time dimension to obtain the initial sequence dataset; constructing and training a long short-term memory network model and analyzing the dynamic evolution of the initial sequence dataset in the time dimension to obtain gradient dynamic features.

[0081] The model building module is used to obtain the historical gradient change characteristics and historical construction disturbance characteristics of the locations of each concentration sensor, and to build a disturbance differentiation model based on the historical gradient change characteristics and historical construction disturbance characteristics.

[0082] The anomaly detection module analyzes the dynamic characteristics of the gradient based on the disturbance differentiation model to determine whether the gradient change is caused by normal construction disturbance. If not, the change in gas concentration is determined to be abnormal.

[0083] The system described in the above embodiments is used to implement the corresponding machine learning-based coal seam gas anomaly detection method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0084] It should be noted that the aforementioned machine learning-based coal seam gas anomaly detection system is presented in the form of functional units. The term "module" here can be implemented in software and / or hardware, without specific limitations.

[0085] For example, a "module" can be a software program, hardware circuit, or a combination of both that implements the above functions. Hardware circuits may include application-specific integrated circuits (ASICs), electronic circuits, processors (e.g., shared processors, proprietary processors, or group processors) and memory for executing one or more software or firmware programs, combined logic circuits, and / or other suitable components that support the described functions.

[0086] Example 3

[0087] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the machine learning-based coal seam gas anomaly detection method described in any of the above embodiments.

[0088] Figure 2 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0089] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0090] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0091] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0092] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB (Universal Serial Bus), network cable, etc.) or wireless means (such as mobile network, WIFI (Wireless Fidelity), Bluetooth, etc.).

[0093] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0094] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0095] The system described in the above embodiments is used to implement the corresponding machine learning-based coal seam gas anomaly detection method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0096] Example 4

[0097] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the machine learning-based coal seam gas anomaly detection method as described in any of the above embodiments.

[0098] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0099] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the machine learning-based coal seam gas anomaly detection method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0100] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.

[0101] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuitry) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that the embodiments of this disclosure may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0102] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0103] Therefore, the units of the various examples described in the embodiments of this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0104] This disclosure is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for detecting coal seam gas anomalies based on machine learning, characterized in that, Includes the following steps: Real-time gas concentration data from several concentration sensors in a coal mine are acquired, and spatiotemporal sequence analysis is performed on the real-time gas concentration data to obtain the changing trends of gas concentration in time and space. Based on the aforementioned trend, a sliding window is used to calculate the gas concentration gradient at the location of each concentration sensor within a specified time period, thereby obtaining a gradient change sequence. For the gradient change sequence, the dynamic evolution pattern of the gradient in the time dimension is analyzed using a long short-term memory network model to obtain the gradient dynamic characteristics. Historical gradient change characteristics and historical construction disturbance characteristics of the locations of each concentration sensor are obtained, and a disturbance discrimination model is constructed based on the historical gradient change characteristics and the historical construction disturbance characteristics. The gradient dynamic characteristics are analyzed based on the disturbance differentiation model to determine whether the gradient change is caused by normal construction disturbance. If not, the gas concentration change is determined to be abnormal.

2. The method for detecting coal seam gas anomalies based on machine learning according to claim 1, characterized in that, The methods for obtaining the aforementioned trend of change include: Real-time gas concentration data from several concentration sensors in a coal mine are acquired, and the real-time gas concentration data is cleaned and missing values ​​are filled in to obtain a complete gas concentration dataset. Based on the complete gas concentration dataset, a gas concentration distribution model in the spatial dimension is constructed using the Kriging interpolation method, and the spatial distribution characteristics of gas are obtained using the gas concentration distribution model. Time series data are extracted from the spatial distribution characteristics of the gas, and the ARIMA model is used to analyze the time series data to obtain the temporal variation trend of gas concentration. The spatial distribution characteristics of the gas and the temporal variation trend of the gas concentration are weighted and fused to obtain the variation trend of the gas concentration in both time and space.

3. The method for detecting coal seam gas anomalies based on machine learning according to claim 1, characterized in that, The methods for obtaining the gradient change sequence include: Based on the changing trend, the time series of gas concentration at the location of each concentration sensor is obtained; Based on the gas concentration time series, the concentration change rate within a specified time period is calculated using a sliding window to obtain the concentration gradient; If the concentration gradient exceeds a first preset threshold, then a local weighted regression is performed on the time series of the concentration sensor to obtain the smoothed gradient change sequence.

4. The method for detecting coal seam gas anomalies based on machine learning according to claim 1, characterized in that, The methods for obtaining the gradient dynamic features include: Obtain the gradient change sequence and extract the gradient values ​​in the time dimension to obtain the initial sequence dataset; A long short-term memory network model is constructed and trained, and the dynamic evolution of the initial sequence dataset in the time dimension is analyzed to obtain the gradient dynamic features.

5. A machine learning-based coal seam gas anomaly detection system, wherein the system applies the method described in any one of claims 1-4, characterized in that, include: The module includes a spatiotemporal sequence analysis module, a gradient change calculation module, a gradient feature analysis module, a model building module, and an anomaly detection module. The spatiotemporal sequence analysis module is used to acquire real-time gas concentration data from several concentration sensors in the coal mine, and to perform spatiotemporal sequence analysis on the real-time gas concentration data to obtain the changing trend of gas concentration in time and space. Based on the changing trend, the gradient change calculation module uses a sliding window to calculate the gas concentration gradient at the location of each concentration sensor within a specified time period, thereby obtaining a gradient change sequence. The gradient feature analysis module is used to analyze the dynamic evolution pattern of the gradient in the time dimension using a long short-term memory network model for the gradient change sequence, and obtain the gradient dynamic features. The model building module is used to obtain the historical gradient change characteristics and historical construction disturbance characteristics of the location of each concentration sensor, and to build a disturbance differentiation model based on the historical gradient change characteristics and the historical construction disturbance characteristics; The anomaly detection module analyzes the gradient dynamic characteristics based on the disturbance differentiation model to determine whether the gradient change is caused by normal construction disturbance. If not, the gas concentration change is determined to be abnormal.

6. The machine learning-based coal seam gas anomaly detection system according to claim 5, characterized in that, The workflow of the spatiotemporal sequence analysis module includes: Real-time gas concentration data from several concentration sensors in a coal mine are acquired, and the real-time gas concentration data is cleaned and missing values ​​are filled in to obtain a complete gas concentration dataset. Based on the complete gas concentration dataset, a gas concentration distribution model in the spatial dimension is constructed using the Kriging interpolation method, and the spatial distribution characteristics of gas are obtained using the gas concentration distribution model. Time series data are extracted from the spatial distribution characteristics of the gas, and the ARIMA model is used to analyze the time series data to obtain the temporal variation trend of gas concentration. The spatial distribution characteristics of the gas and the temporal variation trend of the gas concentration are weighted and fused to obtain the variation trend of the gas concentration in both time and space.

7. The machine learning-based coal seam gas anomaly detection system according to claim 5, characterized in that, The workflow of the gradient change calculation module includes: Based on the changing trend, the time series of gas concentration at the location of each concentration sensor is obtained; Based on the gas concentration time series, the concentration change rate within a specified time period is calculated using a sliding window to obtain the concentration gradient; If the concentration gradient exceeds a first preset threshold, then a local weighted regression is performed on the time series of the concentration sensor to obtain the smoothed gradient change sequence.

8. The machine learning-based coal seam gas anomaly detection system according to claim 5, characterized in that, The workflow of the gradient feature analysis module includes: Obtain the gradient change sequence and extract the gradient values ​​in the time dimension to obtain the initial sequence dataset; A long short-term memory network model is constructed and trained, and the dynamic evolution of the initial sequence dataset in the time dimension is analyzed to obtain the gradient dynamic features.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the machine learning-based coal seam gas anomaly detection method as described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the machine learning-based coal seam gas anomaly detection method as described in any one of claims 1 to 4.

Images