Intelligent damper window remote linkage control device and method

Abstract

The present application relates to the technical field of roadway ventilation control, and particularly relates to an intelligent air door and air window remote linkage control device and method, the method comprising deploying a multi-point differential pressure sensing unit array in a target roadway in a coal mine to obtain a plurality of grid-distributed measuring points; using a preset sampling rate and time alignment algorithm to obtain a high-resolution wind pressure slope field of the target roadway in real time; presetting a time window to extract trend characteristics of a current air flow state; inputting the obtained trend characteristics into a pre-constructed measuring point confidence judgment model to obtain a confidence level of the measuring point, the confidence level comprising a low confidence level, a medium confidence level and a high confidence level. The intelligent air door and air window remote linkage control method provided by the present application relies on a slope field and confidence area determination mechanism, and automatically enters a protection state when the air flow is not reliable to block the occurrence of dangerous operations, thereby reducing the risk of secondary disasters such as gas accumulation and dust surge, and improving the safety guarantee level of underground workers.

Description

Technical Field

[0001] This invention relates to the field of ventilation control technology, specifically to a remote linkage control method for intelligent air doors and windows. Background Technology

[0002] In underground coal mine ventilation control systems, intelligent air doors and windows typically rely on parameters such as changes in air pressure, airflow direction, and personnel passage status to achieve remote linkage control, ensuring stable airflow distribution in the roadways. However, in deep mining areas and connecting roadways of the mining face, where airflow fluctuates dramatically, the ventilation system may experience extremely low-probability nonlinear abrupt changes under specific dynamic scenarios, making it difficult for existing linkage logic to respond accurately.

[0003] For example, when the mine's main fan experiences a brief shutdown and rapid restart after an external power grid failure, the air pressure in the roadway will change drastically within a very short time. This change may cause a momentary reversal of the airflow direction in the roadway, forming a pulsed negative pressure air mass. Affected by the sudden change in air pressure, the air pressure sensor may experience problems such as output saturation and direction judgment jumps, making it impossible for the control system to correctly identify the actual airflow trend. In this situation, the linkage control unit may incorrectly execute preset actions, causing the downstream air door, which should have been closed, to open prematurely, while the upstream air door fails to close in time due to a delay in action. This results in a short-term through-flow negative pressure channel in the roadway. The negative pressure channel may cause residual gas and dust from the goaf to be drawn into the personnel working area in a short period of time, causing a sudden increase in local gas and dust concentrations within tens of seconds, creating a serious safety hazard.

[0004] Such wind reversal scenarios caused by sudden changes in wind pressure are characterized by their suddenness, short perception time, and difficulty in being captured by conventional logic. In existing remote linkage control methods, there is a lack of mechanisms to identify, filter, and autonomously judge abnormal states such as instantaneous disturbances, extreme wind pressure slopes, and directional change signals, which makes the system prone to malfunctions in the above extreme scenarios. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a remote linkage control device and method for intelligent air doors and windows.

[0006] This invention adopts the following technical solution: a remote linkage control method for intelligent air doors and windows, comprising:

[0007] A multi-point differential pressure sensor array is deployed in the target roadway of the coal mine to obtain multiple measuring points that are evenly distributed in a grid.

[0008] A high-resolution wind pressure slope field of the target roadway is obtained in real time using a preset sampling rate and time alignment algorithm;

[0009] A preset time window is used to obtain the sign and amplitude change data of the wind pressure slope within the preset time window based on the wind pressure slope field, and to extract the trend characteristics of the current wind flow state.

[0010] The acquired trend features are input into a pre-built confidence judgment model for measurement points to obtain the confidence level of the measurement point, which includes low confidence level, medium confidence level and high confidence level;

[0011] Traverse all measuring points, obtain the confidence level of all measuring points, register the measuring points with the roadway geometric model, map the measuring points to the roadway geometric model, and generate the roadway confidence model;

[0012] Based on the roadway confidence model, connectivity processing and merging processing are performed to divide the roadway into low-confidence, medium-confidence, and high-confidence regions.

[0013] Different remote linkage control optimization commands are generated based on different confidence levels.

[0014] As a further description of the above technical solution: the method for obtaining the high-resolution wind pressure slope field of the target roadway includes:

[0015] An absolute time reference is established at the main control station, and a preset sampling rate and timestamp format are sent to all differential pressure sensing units to ensure that each unit starts sampling with the same time base.

[0016] Synchronous testing was used to measure the transmission delay compensation amount of the round-trip delay from sampling to reception by the main control station for each differential pressure sensing unit. The sampling timestamp of each differential pressure sensing unit was corrected to obtain the corrected timestamp. Based on the corrected timestamp, all differential pressure data were mapped to the same absolute time axis.

[0017] The differential pressure sensing unit position is registered with the roadway geometric model, and the filtered differential pressure value is mapped to the roadway geometric model at each sampling time to generate a discrete wind pressure field with position constraints.

[0018] The time derivative is calculated using the central difference method on each time slice to generate the original slope field. Median filtering and total variation regularization are applied to the original slope field to preserve the slope abrupt boundary and suppress random noise, resulting in a smooth and boundary-preserving high-resolution wind pressure slope field.

[0019] As a further description of the above technical solution: the trend characteristics of the airflow state include the number of slope sign changes, the average amplitude, and the amplitude fluctuation range.

[0020] As a further description of the above technical solution: the method for obtaining the number of slope sign changes is as follows: based on a preset sampling rate, within a preset time window, the slope sign of each sampling at the measuring point is obtained. The slope sign includes positive and negative. When the sign changes once, it is recorded as one change. The number of slope sign changes is obtained by counting all the number of changes.

[0021] As a further description of the above technical solution: the method for obtaining the average amplitude includes: based on a preset sampling rate, within a preset time window, obtaining the absolute value of the slope of each sampling at the measuring point, which is recorded as the amplitude; summing all the sampled amplitudes and dividing by the number of samplings to obtain the average amplitude.

[0022] The method for obtaining the amplitude fluctuation range includes: calculating the deviation between each sampled amplitude and the average amplitude within the time window, calculating the standard deviation of these deviations, and obtaining the amplitude fluctuation range.

[0023] As a further description of the above technical solution: the training method of the measurement point confidence judgment model includes:

[0024] Q sets of training data are collected in advance, where Q is an integer greater than 0. The training data includes trend features and corresponding confidence levels of measurement points. The confidence levels include low confidence level, medium confidence level and high confidence level, and the corresponding labels are set to 0, 1 and 2 respectively.

[0025] Stratified sampling is used to divide the training data into training set, validation set and test set according to a preset ratio to ensure that the class distribution of each dataset is consistent.

[0026] Gradient boosting tree classifier is used as the confidence judgment model for test points. Initial hyperparameters are set, the model is initialized, and decision trees are built iteratively. Each new tree is fitted based on the negative gradient of the loss function of the training set. The optimal splitting feature and splitting point are selected by Gini impurity, and the samples are divided into different child nodes until the stopping condition is met.

[0027] The weights of the leaf nodes of each tree are optimized using gradient descent. The optimal combination is searched within the preset hyperparameter range using Bayesian optimization. The F1 score on the validation set is calculated every 20 trees. Training is stopped when the F1 score improves by less than 0.01 for three consecutive times. The trained model is then evaluated using the test set. If the model performance meets the evaluation criteria, it is deployed and applied.

[0028] As a further description of the above technical solution: the method for dividing the roadway into low-confidence, medium-confidence, and high-confidence regions includes:

[0029] Connectivity analysis was performed on spatially adjacent measurement points with the same confidence level in the tunnel geometric model, and initial connected regions with low confidence, medium confidence, and high confidence were extracted respectively.

[0030] A preset area threshold is used to mark connected regions with an area less than or equal to the area threshold as regions to be merged.

[0031] Map the regions to be merged to the roadway geometry model to obtain their actual location in the roadway. When the regions to be merged are located at specific key locations in the roadway, they are marked as isolated regions and no merging operation is performed.

[0032] The specific key locations include at least: air doors, air windows, intersections of branch roads, branch points, sharp bends, and abrupt changes in road cross-section.

[0033] When the region to be merged is not an isolated region, calculate the shortest axial distance between it and other connected regions of the same confidence level. When the shortest axial distance is less than or equal to a preset distance threshold, merge the region to be merged with the corresponding connected region of the same level, and synchronously mark the intermediate regions between them as the same whole region, so as to form a continuous confidence region.

[0034] When the shortest axial distance is greater than a preset distance threshold, the confidence level of the adjacent regions of the region to be merged is compared, and the region is merged with the connected region that is closest in confidence level and spatially adjacent.

[0035] For each connected region after merging, the number of measurement points with low confidence, medium confidence, and high confidence, as well as their proportion within the region, are counted.

[0036] A preset confidence level percentage threshold is set, and the final confidence level of the connected region is determined based on the comparison between the percentage of each level of measurement points and the preset confidence level percentage threshold.

[0037] As a further description of the above technical solution: the method for determining the final confidence level of the connected region based on the comparison result of the proportion of each level of measuring points with the preset confidence level proportion threshold includes:

[0038] When the proportion of low-confidence measurement points exceeds the preset confidence level proportion threshold, the connected region is marked as a low-confidence region.

[0039] If the proportion of confidence-level measurement points is greater than the preset confidence level proportion threshold, the connected region is marked as a medium confidence region.

[0040] When the proportion of high-confidence measurement points exceeds the preset confidence level proportion threshold, the connected region is marked as a high-confidence region;

[0041] When none of the three types of measurement points reach the preset confidence level percentage threshold, the connected region is divided into the corresponding confidence level based on the maximum percentage of the three types of measurement points.

[0042] As a further description of the above technical solution: the method for generating different remote linkage control optimization commands based on different confidence levels includes:

[0043] In the high confidence region, the airflow assessment is reliable and can be directly used for linkage control decisions;

[0044] For the medium confidence zone, there is a slight uncertainty in wind flow judgment, so a delayed remote linkage control is adopted; the delay time is set manually.

[0045] For low-confidence areas and isolated regions, airflow assessment is unreliable, so remote linkage control should be disconnected.

[0046] A remote linkage control device for intelligent dampers and windows, used to implement the aforementioned remote linkage control method for intelligent dampers and windows, the device comprising:

[0047] The multi-point measurement deployment module deploys a multi-point differential pressure sensing unit array in the target roadway within the coal mine to obtain multiple measurement points that are evenly distributed in a grid.

[0048] The slope field acquisition module uses a preset sampling rate and time alignment algorithm to acquire the high-resolution wind pressure slope field of the target roadway in real time;

[0049] The trend feature extraction module has a preset time window. Based on the wind pressure slope field, it obtains the sign and amplitude change data of the wind pressure slope within the preset time window and extracts the trend features of the current wind flow state.

[0050] The measurement point confidence determination module inputs the acquired trend features into the pre-built measurement point confidence determination model to obtain the confidence level of the measurement point, which includes low confidence level, medium confidence level and high confidence level;

[0051] The tunnel model generation module iterates through all measuring points, obtains the confidence level of all measuring points, registers the measuring points with the tunnel geometric model, maps the measuring points to the tunnel geometric model, and generates the tunnel confidence model.

[0052] The confidence partitioning module performs connectivity and merging processing based on the lane confidence model, partitions the lanes, and obtains low-confidence, medium-confidence, and high-confidence zones.

[0053] The linkage command optimization module generates different remote linkage control optimization commands based on different confidence levels.

[0054] Beneficial effects:

[0055] The intelligent remote linkage control method for air doors and windows provided by this invention, by deploying a spatially distributed differential pressure sensing unit array in the roadway, enables the system to simultaneously capture wind pressure change information at multiple locations. A time alignment algorithm eliminates delay deviations between different measuring points, unifying all differential pressure data onto the same absolute time axis. The central difference method is introduced to calculate the wind pressure time slope, allowing the system to more sensitively identify the rate of wind pressure change. Then, a confidence zone model quantifies the reliability of airflow in different areas, dividing the roadway into high-confidence, medium-confidence, and low-confidence zones. This achieves a spatial expression of airflow reliability. The high-confidence zone reflects stable and reliable airflow, allowing execution according to normal logic; the medium-confidence zone indicates some interference, allowing for appropriate delays in execution; and the low-confidence zone indicates that the current airflow state is unreliable, cutting off remote linkage control. This mechanism can prevent accidental opening or closing of air doors in extreme and sudden scenarios, effectively preventing the formation of through-type negative pressure channels.

[0056] In summary, this method relies on the slope field and confidence zone determination mechanism to automatically enter a protection state when the airflow is unreliable, thereby preventing dangerous operations and reducing the risk of secondary disasters such as gas accumulation and sudden dust rise, and improving the safety level of underground workers. Attached Figure Description

[0057] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0058] Figure 1 This is a flowchart illustrating the remote linkage control method for intelligent air doors and windows provided in Embodiment 1 of the present invention.

[0059] Figure 2 This is a flowchart of a method for obtaining a high-resolution wind pressure slope field of a target roadway, as provided in Embodiment 1 of the present invention.

[0060] Figure 3 A flowchart of the method for dividing a roadway into low-confidence, medium-confidence, and high-confidence regions according to Embodiment 1 of the present invention;

[0061] Figure 4 This is a module connection diagram of the intelligent air door and window remote linkage control device provided in Embodiment 2 of the present invention. Detailed Implementation

[0062] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0063] Example 1:

[0064] Please see Figures 1-3This invention provides a technical solution: a remote linkage control method for intelligent air doors and windows, comprising:

[0065] A multi-point differential pressure sensor array is deployed in the target roadway of the coal mine to obtain multiple measuring points that are evenly distributed in a grid.

[0066] The method for deploying a multi-point differential pressure sensor array includes:

[0067] A grid array is laid out along the axis of the target roadway in the coal mine to form a grid unit. The axial spacing is set based on the lateral characteristic scale of the roadway (for example, one column is laid out every 10-20m). The density at turns, contractions, and branch roadways is increased to 1 / 3 of the axial spacing. At each layout point, at least two differential pressure sensors are installed, which are redundant and measure the lateral pressure difference in pairs.

[0068] Specifically, by deploying a spatially distributed differential pressure sensing unit array within the tunnel, the system can simultaneously capture wind pressure change information at multiple locations. Compared to traditional data models that rely on only a single point or a few nodes, this method greatly improves spatial coverage, significantly enhancing the ability to capture local airflow disturbances, pulsed negative pressure, and instantaneous reversals, providing a comprehensive and reliable data foundation for subsequent wind pressure slope calculation and airflow trend determination.

[0069] A high-resolution wind pressure slope field of the target roadway is obtained in real time using a preset sampling rate and time alignment algorithm.

[0070] Specifically, due to network latency differences in downhole sensors, unprocessed data cannot be directly used to determine the overall trend of airflow changes. By using a time alignment algorithm to eliminate the latency deviation between different measuring points, all differential pressure data are unified to the same absolute time axis, improving the comparability and synchronization of information from multiple measuring points. This step ensures the accuracy of the wind pressure slope field construction, laying the foundation for accurately identifying real airflow changes in abrupt scenarios.

[0071] It should be noted that high-resolution wind pressure slope fields are mainly used to characterize the spatial variation trend and local non-uniformity of wind pressure. By the magnitude and direction of the slope, the disturbance amplitude, rotation characteristics, and flow direction change trend of local airflow can be characterized. It can reflect key flow characteristics such as vortices, shear flow, and recirculation zones. By using features such as gradient anomalies, slope abrupt changes, and direction shifts in the slope field, it can be used to locate wind field anomaly points, fault points, or hidden hazard sources.

[0072] The method for obtaining the high-resolution wind pressure slope field of the target roadway includes:

[0073] Establish an absolute time reference at the master control station Send a preset sampling rate to all differential pressure sensing units With the timestamp format, ensure that all units begin sampling on the same time base;

[0074] It should be noted that the timestamp format is as follows: ;in, In the formula, For the first The differential pressure sensing unit in the first The timestamp of each sampling point This indicates the absolute time reference established by the master control station. For the preset sampling rate, This is the sampling point number.

[0075] Synchronous testing was employed to measure the transmission delay compensation amount for the round-trip delay from sampling to reception at the main control station for each differential pressure sensing unit. The sampling timestamp of each differential pressure sensing unit is corrected to obtain the corrected timestamp. Based on the corrected timestamps, all differential pressure data are mapped to the same absolute time axis;

[0076] It should be noted that, , This is the corrected timestamp. This is the amount of compensation for transmission delay;

[0077] The differential pressure sensing unit position is registered with the roadway geometric model, and the filtered differential pressure value is mapped to the roadway geometric model at each sampling time to generate a discrete wind pressure field with position constraints.

[0078] It should be noted that the expression for the discrete wind pressure field is: ;in, , and For the first The spatial coordinates of each measuring point in the tunnel model Let be the differential pressure value at the i-th measuring point at time t;

[0079] The time derivative is calculated using the central difference method on each time slice to generate the original slope field. Median filtering and total variation regularization are applied to the original slope field to preserve the slope abrupt boundary and suppress random noise, resulting in a smooth and boundary-preserving high-resolution wind pressure slope field.

[0080] Specifically, this method introduces the central difference method to calculate the wind pressure time slope, enabling the system to more sensitively identify the rate of wind pressure change. The slope field, as a first derivative feature, can quickly reflect dynamic features such as wind flow reversal, peak pulses, and shear disturbances. Even in extreme wind flow events, it can still ensure the timely extraction of trend features and improve the ability to identify sudden dangerous states.

[0081] It should be noted that the expression for calculating the time derivative and generating the original slope field using the central difference method is as follows:

[0082] ;

[0083] For sampling time The original wind pressure time slope field is shown below. The sampling time interval, The spatial coordinates of the measuring point, To be at the sampling time Discrete wind pressure field values, For sampling time Discrete wind pressure field values.

[0084] It should be noted that the expression for applying median filtering and total variation regularization to the original slope field to preserve the slope abrupt boundary and suppress random noise is as follows:

[0085] ;

[0086] In the formula, For high-resolution wind pressure slope field, Search High-resolution wind pressure slope field with the largest value , Represents the high-resolution wind pressure slope field With respect to the original wind pressure time slope field The difference is the L1 norm over the entire spatial domain. The total variation regularization coefficient is . This represents the total variation of the slope field gradient, used to preserve the boundary conditions. This indicates that the total variation of the slope field is obtained by volume integration over the gradient magnitude of the entire spatial domain.

[0087] A preset time window is used to obtain the sign and amplitude of the wind pressure slope within the preset time window based on the wind pressure slope field, and to extract the trend characteristics of the current airflow state.

[0088] The trend characteristics of the airflow state include the number of times the slope sign changes, the average amplitude, and the amplitude fluctuation range;

[0089] The method for obtaining the number of slope sign changes is as follows: based on a preset sampling rate, within a preset time window, the slope sign of each sampling at the measuring point is obtained. The slope sign includes positive and negative. When the sign changes once, it is recorded as one change. All changes are counted to obtain the number of slope sign changes.

[0090] The method for obtaining the average amplitude includes: based on a preset sampling rate, within a preset time window, obtaining the absolute value of the slope of each sampling at the measuring point, which is recorded as the amplitude; summing all the sampled amplitudes and dividing by the number of samplings to obtain the average amplitude.

[0091] The method for obtaining the amplitude fluctuation range includes: calculating the deviation between each sampled amplitude and the average amplitude within the time window, calculating the standard deviation of these deviations, and obtaining the amplitude fluctuation range.

[0092] It should be noted that if the slope sign remains consistent within the time window, it indicates that the airflow direction is stable, increasing the confidence level; if the sign changes frequently, it may indicate a transient disturbance or abnormal measurement point, reducing the confidence level.

[0093] It should be noted that the amplitude of the slope reflects the degree of drastic change in wind pressure. The smaller the amplitude, the slower the change in airflow and the lower the confidence level. The larger the amplitude, the lower the confidence level due to sudden disturbances or sensor malfunctions. An amplitude within a reasonable range corresponds to a higher confidence level.

[0094] The acquired trend features are input into the pre-built measurement point confidence judgment model to obtain the confidence level of the measurement point;

[0095] The training method for the measurement point confidence judgment model includes:

[0096] Q sets of training data are collected in advance, where Q is an integer greater than 0. The training data includes trend features and corresponding confidence levels of measurement points. The confidence levels include low confidence level, medium confidence level and high confidence level, and the corresponding labels are set to 0, 1 and 2 respectively.

[0097] Stratified sampling is used to divide the training data into training, validation, and test sets according to a preset ratio (e.g., 6:3:1) to ensure that the class distribution of each dataset is consistent.

[0098] Gradient boosting tree classifier (GBDT classifier) ​​is used as the confidence judgment model for test points. The initial hyperparameters are set as follows: number of decision trees: 100-150, maximum depth of a single tree: 4-6, minimum number of samples for node split: 8-12, maximum number of features considered during split: 2-3 (based on 3 types of input features, some features are randomly selected to improve model diversity), the split criterion is Gini impurity, and the learning rate is 0.05-0.1.

[0099] Initialize the model, iteratively build a decision tree, fit each new tree based on the negative gradient of the loss function of the training set, select the optimal splitting feature and splitting point through Gini impurity, divide the samples into different child nodes, until the stopping condition is met (such as reaching the maximum depth or the number of samples in the child node is less than the minimum number of samples).

[0100] The weights of the leaf nodes of each tree are optimized using gradient descent, and the learning rate is used to control the contribution of new trees to the final prediction results, thus avoiding overfitting caused by excessive weights in a single tree.

[0101] The Bayesian optimization method is used to search for the optimal combination within a preset range of hyperparameters. The optimization range includes: 80-180 decision trees, learning rate 0.03-0.12, regularization coefficient 0.05-0.2, and the optimization objective is the weighted F1 score of the validation set (balancing the differences in the proportion of samples in each category).

[0102] For every 20 trees in each iteration, calculate the F1 score on the validation set; when the F1 score improves by less than 0.01 for 3 consecutive iterations (60 trees), stop training to avoid overfitting the model; save the model parameters with the highest F1 score on the validation set (including the splitting rules of all decision trees) to ensure that there is still a stable ability to determine the types of abnormal heat sources that have not been seen before.

[0103] The trained model is evaluated using a test set, and precision, recall, F1 score, and confusion matrix are calculated. If the precision is ≥90%, recall is ≥95%, and F1 score is ≥92%, the model performance evaluation is satisfactory, and it can be deployed and applied.

[0104] Traverse all measuring points, obtain the confidence level of all measuring points, register the measuring points with the roadway geometric model, map the measuring points to the roadway geometric model, and generate the roadway confidence model;

[0105] Connectivity and merging processes are performed based on the roadway confidence model. The roadway is divided into low-confidence, medium-confidence, and high-confidence regions. Different linkage optimization instructions are generated based on different confidence regions.

[0106] Methods for dividing roadways into low-confidence, medium-confidence, and high-confidence regions include:

[0107] Connectivity analysis was performed on spatially adjacent measurement points with the same confidence level in the tunnel geometric model, and initial connected regions with low confidence, medium confidence, and high confidence were extracted respectively.

[0108] A preset area threshold is used to mark connected regions with an area less than or equal to the threshold as regions to be merged.

[0109] Map the regions to be merged to the roadway geometry model to obtain their actual location in the roadway. When the regions to be merged are located at specific key locations in the roadway, they are marked as isolated regions and no merging operation is performed.

[0110] The specific key locations include at least: air doors, air windows, intersections of branch roads, branch points, sharp bends, and abrupt changes in road cross-section.

[0111] When the region to be merged is not an isolated region, calculate the shortest axial distance between it and other connected regions of the same confidence level. When the shortest axial distance is less than or equal to a preset distance threshold, merge the region to be merged with the corresponding connected region of the same level, and synchronously mark the intermediate regions between them as the same whole region, so as to form a continuous confidence region.

[0112] It should be noted that the shortest axial distance is the minimum interval distance measured along a single axis direction between the region to be merged and other connected regions of the same confidence level in a preset coordinate axis direction (such as the X-axis, Y-axis, or Z-axis direction).

[0113] When the shortest axial distance is greater than a preset distance threshold, the confidence level of the adjacent regions of the region to be merged is compared, and the region is merged with the connected region that is closest in confidence level and spatially adjacent.

[0114] For each connected region after merging, the number of measurement points with low confidence, medium confidence, and high confidence, as well as their proportion within the region, are counted.

[0115] A preset confidence level percentage threshold is set, and the final confidence level of the connected region is determined based on the comparison between the percentage of each level of measurement points and the preset confidence level percentage threshold.

[0116] The method for determining the final confidence level of a connected region based on a comparison of the proportion of each level of measurement points with the preset confidence level proportion threshold includes:

[0117] When the proportion of low-confidence measurement points exceeds a preset threshold, the connected region is marked as a low-confidence region.

[0118] If the proportion of confidence-level measurement points is greater than a preset threshold, the connected region is marked as a medium confidence region.

[0119] When the proportion of high-confidence measurement points exceeds a preset threshold, the connected region is marked as a high-confidence region;

[0120] When none of the three types of measurement points reach the preset confidence level percentage threshold, the connected region is divided into the corresponding confidence level based on the maximum percentage of the three types of measurement points.

[0121] Methods for generating different remote linkage control optimization commands based on different confidence levels include:

[0122] For the high confidence region, the airflow assessment is reliable and can be directly used for linkage control decisions;

[0123] For the medium confidence zone, there is a slight uncertainty in the wind flow judgment, so a delayed remote linkage control is adopted; the delay time is set manually.

[0124] In the low confidence zone, airflow judgment is unreliable, so remote linkage control should be cut off.

[0125] For isolated areas, remote linkage control should be disconnected.

[0126] Specifically, this method quantifies the reliability of airflow in different regions using a confidence zone model, dividing the roadway into high-confidence, medium-confidence, and low-confidence zones. This achieves a spatial expression of airflow reliability. The high-confidence zone reflects stable and reliable airflow, allowing execution according to normal logic. The medium-confidence zone indicates some interference, requiring appropriate delays in execution. The low-confidence zone indicates that the current airflow state is unreliable, cutting off remote linkage control. This mechanism can prevent accidental opening or closing of air doors in extreme and sudden scenarios, effectively preventing the formation of negative pressure channels. In special scenarios such as short-term restarts of the main fan, airflow reversal, and sudden changes in air pressure, traditional systems are prone to erroneous opening and closing, leading to short-term roadway connection. This method, relying on the slope field and confidence zone determination mechanism, automatically enters a protection state when airflow is unreliable, preventing dangerous operations and reducing the risk of secondary disasters such as gas accumulation and sudden dust rises, thereby improving the safety level of underground workers.

[0127] Example 2

[0128] A remote linkage control device for intelligent dampers and windows, used to implement the aforementioned remote linkage control method for intelligent dampers and windows, the device comprising:

[0129] The multi-point measurement deployment module deploys a multi-point differential pressure sensing unit array in the target roadway within the coal mine to obtain multiple measurement points that are evenly distributed in a grid.

[0130] The slope field acquisition module uses a preset sampling rate and time alignment algorithm to acquire the high-resolution wind pressure slope field of the target roadway in real time;

[0131] The trend feature extraction module has a preset time window. Based on the wind pressure slope field, it obtains the sign and amplitude change data of the wind pressure slope within the preset time window and extracts the trend features of the current wind flow state.

[0132] The measurement point confidence determination module inputs the acquired trend features into the pre-built measurement point confidence determination model to obtain the confidence level of the measurement point, which includes low confidence level, medium confidence level and high confidence level;

[0133] The tunnel model generation module iterates through all measuring points, obtains the confidence level of all measuring points, registers the measuring points with the tunnel geometric model, maps the measuring points to the tunnel geometric model, and generates the tunnel confidence model.

[0134] The confidence partitioning module performs connectivity and merging processing based on the lane confidence model, partitions the lanes, and obtains low-confidence, medium-confidence, and high-confidence zones.

[0135] The linkage command optimization module generates different remote linkage control optimization commands based on different confidence levels.

[0136] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A remote linkage control method for intelligent air doors and windows, characterized in that, include: A multi-point differential pressure sensor array is deployed in the target roadway of the coal mine to obtain multiple measuring points that are evenly distributed in a grid. A high-resolution wind pressure slope field of the target roadway is obtained in real time using a preset sampling rate and time alignment algorithm; A preset time window is used to obtain the sign and amplitude change data of the wind pressure slope within the preset time window based on the wind pressure slope field, and to extract the trend characteristics of the current wind flow state. The acquired trend features are input into a pre-built confidence judgment model for measurement points to obtain the confidence level of the measurement point, which includes low confidence level, medium confidence level and high confidence level; Traverse all measuring points, obtain the confidence level of all measuring points, register the measuring points with the roadway geometric model, map the measuring points to the roadway geometric model, and generate the roadway confidence model; Based on the roadway confidence model, connectivity processing and merging processing are performed to divide the roadway into low-confidence, medium-confidence, and high-confidence regions. Different remote linkage control optimization commands are generated based on different confidence levels.

2. The remote linkage control method for intelligent air doors and windows according to claim 1, characterized in that, The method for obtaining the high-resolution wind pressure slope field of the target roadway includes: An absolute time reference is established at the main control station, and a preset sampling rate and timestamp format are sent to all differential pressure sensing units to ensure that each unit starts sampling with the same time base. Synchronous testing was used to measure the transmission delay compensation amount of the round-trip delay from sampling to reception by the main control station for each differential pressure sensing unit. The sampling timestamp of each differential pressure sensing unit was corrected to obtain the corrected timestamp. Based on the corrected timestamp, all differential pressure data were mapped to the same absolute time axis. The differential pressure sensing unit position is registered with the roadway geometric model, and the filtered differential pressure value is mapped to the roadway geometric model at each sampling time to generate a discrete wind pressure field with position constraints. The time derivative is calculated using the central difference method on each time slice to generate the original slope field. Median filtering and total variation regularization are applied to the original slope field to preserve the slope abrupt boundary and suppress random noise, resulting in a smooth and boundary-preserving high-resolution wind pressure slope field.

3. The remote linkage control method for intelligent air doors and windows according to claim 1, characterized in that, The trend characteristics of the airflow state include the number of slope sign changes, the average amplitude, and the amplitude fluctuation range.

4. The intelligent air door and window remote linkage control method according to claim 3, characterized in that, The method for obtaining the number of slope sign changes is as follows: based on a preset sampling rate, within a preset time window, the slope sign of each sampling at the measuring point is obtained. The slope sign includes positive and negative. When the sign changes once, it is recorded as one change. All changes are counted to obtain the number of slope sign changes.

5. The remote linkage control method for intelligent air doors and windows according to claim 3, characterized in that, The method for obtaining the average amplitude includes: based on a preset sampling rate, within a preset time window, obtaining the absolute value of the slope of each sampling at the measuring point, which is recorded as the amplitude; summing all the sampled amplitudes and dividing by the number of samplings to obtain the average amplitude. The method for obtaining the amplitude fluctuation range includes: calculating the deviation between each sampled amplitude and the average amplitude within the time window, calculating the standard deviation of these deviations, and obtaining the amplitude fluctuation range.

6. The remote linkage control method for intelligent air doors and windows according to claim 1, characterized in that, The training method for the measurement point confidence judgment model includes: Q sets of training data are collected in advance, where Q is an integer greater than 0. The training data includes trend features and corresponding confidence levels of measurement points. The confidence levels include low confidence level, medium confidence level and high confidence level, and the corresponding labels are set to 0, 1 and 2 respectively. Stratified sampling is used to divide the training data into training set, validation set and test set according to a preset ratio to ensure that the class distribution of each dataset is consistent. Gradient boosting tree classifier is used as the confidence judgment model for test points. Initial hyperparameters are set, the model is initialized, and decision trees are built iteratively. Each new tree is fitted based on the negative gradient of the loss function of the training set. The optimal splitting feature and splitting point are selected by Gini impurity, and the samples are divided into different child nodes until the stopping condition is met. The weights of the leaf nodes of each tree are optimized using gradient descent. The optimal combination is searched within the preset hyperparameter range using Bayesian optimization. The F1 score on the validation set is calculated every 20 trees. Training is stopped when the F1 score improves by less than 0.01 for three consecutive times. The trained model is then evaluated using the test set. If the model performance meets the evaluation criteria, it is deployed and applied.

7. The intelligent air door and window remote linkage control method according to claim 1, characterized in that, Methods for dividing roadways into low-confidence, medium-confidence, and high-confidence regions include: Connectivity analysis was performed on spatially adjacent measurement points with the same confidence level in the tunnel geometric model, and initial connected regions with low confidence, medium confidence, and high confidence were extracted respectively. A preset area threshold is used to mark connected regions with an area less than or equal to the area threshold as regions to be merged. Map the regions to be merged to the roadway geometry model to obtain their actual location in the roadway. When the regions to be merged are located at specific key locations in the roadway, they are marked as isolated regions and no merging operation is performed. The specific key locations include at least: air doors, air windows, intersections of branching roadways, branch points, sharp bends, and abrupt changes in roadway cross-sections; When the region to be merged is not an isolated region, calculate the shortest axial distance between it and other connected regions of the same confidence level. When the shortest axial distance is less than or equal to a preset distance threshold, merge the region to be merged with the corresponding connected region of the same level, and synchronously mark the intermediate regions between them as the same whole region, so as to form a continuous confidence region. When the shortest axial distance is greater than a preset distance threshold, the confidence level of the adjacent regions of the region to be merged is compared, and the region is merged with the connected region that is closest in confidence level and spatially adjacent. For each connected region after merging, the number of measurement points with low confidence, medium confidence, and high confidence, as well as their proportion within the region, are counted. A preset confidence level percentage threshold is set, and the final confidence level of the connected region is determined based on the comparison between the percentage of each level of measurement points and the preset confidence level percentage threshold.

8. The remote linkage control method for intelligent air doors and windows according to claim 7, characterized in that, The method for determining the final confidence level of the connected region based on the comparison between the proportion of each level of measuring points and the preset confidence level proportion threshold includes: When the proportion of low-confidence measurement points exceeds the preset confidence level proportion threshold, the connected region is marked as a low-confidence region. If the proportion of confidence-level measurement points is greater than the preset confidence level proportion threshold, the connected region is marked as a medium confidence region. When the proportion of high-confidence measurement points exceeds the preset confidence level proportion threshold, the connected region is marked as a high-confidence region; When none of the three types of measurement points reach the preset confidence level percentage threshold, the connected region is divided into the corresponding confidence level based on the maximum percentage of the three types of measurement points.

9. The intelligent air door and window remote linkage control method according to claim 1, characterized in that, The method for generating different remote linkage control optimization commands based on different confidence levels includes: In the high confidence region, the airflow assessment is reliable and can be directly used for linkage control decisions; For the medium confidence zone, there is a slight uncertainty in the wind flow assessment, so a delayed remote linkage control is adopted; For low-confidence areas and isolated regions, airflow assessment is unreliable, so remote linkage control should be disconnected.

10. A remote linkage control device for intelligent dampers and windows, used to implement the remote linkage control method for intelligent dampers and windows as described in any one of claims 1-9, characterized in that, The device includes: The multi-point measurement deployment module deploys a multi-point differential pressure sensing unit array in the target roadway within the coal mine to obtain multiple measurement points that are evenly distributed in a grid. The slope field acquisition module uses a preset sampling rate and time alignment algorithm to acquire the high-resolution wind pressure slope field of the target roadway in real time; The trend feature extraction module has a preset time window. Based on the wind pressure slope field, it obtains the sign and amplitude change data of the wind pressure slope within the preset time window and extracts the trend features of the current wind flow state. The measurement point confidence determination module inputs the acquired trend features into the pre-constructed measurement point confidence determination model to obtain the confidence level of the measurement point, which includes low confidence level, medium confidence level and high confidence level; The tunnel model generation module iterates through all measuring points, obtains the confidence level of all measuring points, registers the measuring points with the tunnel geometric model, maps the measuring points to the tunnel geometric model, and generates the tunnel confidence model. The confidence partitioning module performs connectivity and merging processing based on the lane confidence model, partitions the lanes, and obtains low-confidence, medium-confidence, and high-confidence zones. The linkage command optimization module generates different remote linkage control optimization commands based on different confidence levels.

Images