A method and system for adaptive adjustment of underground intelligent dampers

Abstract

The present application belongs to the technical field of underground ventilation system regulation and control, and specifically provides an underground intelligent damper adaptive regulation method and system, the method mainly comprising: obtaining underground real-time monitoring data, the monitoring data including the damper opening original sequence, the wind speed original sequence, and the air pressure original sequence; pre-processing the damper opening original sequence and the wind speed original sequence to obtain a steady-state opening sequence and an effective wind speed sequence; correlating the steady-state opening sequence and the effective wind speed sequence to generate a characteristic parameter set that characterizes the dynamic equilibrium relationship between the opening and the wind speed; and regularizing the characteristic parameter set to generate a characteristic parameter set that adapts to the air pressure control logic. The present application can capture the coupling relationship between underground wind speed, air pressure, and damper opening in real time, and adaptively match the control strategies under different working conditions, thereby improving the regulation accuracy, accelerating the response speed, ensuring the stable operation of the underground ventilation system, and reducing safety risks.

Description

Technical Field

[0001] The present application belongs to the technical field of underground ventilation system regulation and control, and in particular relates to an underground intelligent damper adaptive regulation method and system. Background Art

[0002] In underground coal mine ventilation systems, dampers are critical for controlling airflow and ensuring safe production. Traditional damper adjustment relies primarily on manual experience or preset parameters. For example, when wind speed exceeds a certain threshold, the damper opening is automatically increased. When air pressure changes, the damper is adjusted according to a fixed ratio.

[0003] However, the underground environment is highly complex and dynamic, with parameters like wind speed and air pressure influencing each other and changing in real time. Monitoring and adjusting a single parameter cannot fully reflect the true state of underground airflow, and can easily lead to delayed or over-adjusted adjustments. For example, adjusting the damper based solely on wind speed may overlook the impact of air pressure changes on airflow, resulting in inaccurate adjustments and inefficient damper adjustment. This makes it difficult to ensure the stable operation of the underground ventilation system, increasing safety risks. Summary of the Invention

[0004] The present application provides an adaptive adjustment method and system for an underground intelligent damper, which effectively solves the problem that the monitoring and adjustment method of a single parameter in the existing technology cannot fully reflect the actual state of the underground airflow, which easily leads to adjustment lag or over-adjustment. It captures the coupling relationship between underground wind speed, air pressure and damper opening in real time, and adaptively matches the control strategy under different working conditions, thereby improving the adjustment accuracy, accelerating the response speed, ensuring the stable operation of the underground ventilation system, and reducing safety risks.

[0005] In order to achieve the above objectives, this application adopts the following technical solutions:

[0006] In a first aspect, a method for adaptively adjusting an underground intelligent damper is provided, comprising:

[0007] Obtain downhole real-time monitoring data, which includes an original sequence of damper opening, an original sequence of wind speed, and an original sequence of air pressure.

[0008] The original sequence of damper opening and wind speed is preprocessed to obtain the steady-state opening sequence and effective wind speed sequence.

[0009] The steady-state opening sequence and the effective wind speed sequence are correlated to generate a set of characteristic parameters that characterize the dynamic balance relationship between the opening and wind speed.

[0010] The characteristic parameter set is parameter-regulated to generate a characteristic parameter set that is adapted to the air pressure control logic.

[0011] The characteristic parameter set and the original air pressure sequence are integrated to generate the final damper opening control instruction.

[0012] Furthermore, the original sequence of damper opening and the original sequence of wind speed are preprocessed to obtain a steady-state opening sequence and an effective wind speed sequence, including:

[0013] The original sequence of the throttle opening is processed by sliding average filtering to remove instantaneous jitter and generate a steady-state opening sequence.

[0014] The original wind speed sequence is processed with dynamic threshold sliding window to generate a valid wind speed sequence.

[0015] Furthermore, the original sequence of the damper opening is subjected to sliding average filtering to remove instantaneous jitter and generate a steady-state opening sequence, including:

[0016] The original sequence of damper opening is filtered using a fixed window sliding average filter to generate a preliminary filtered sequence.

[0017] The filtering window is dynamically adjusted based on the fluctuation amplitude of the preliminary filtering sequence.

[0018] The adjusted filter window is used to perform a secondary sliding average filter on the original sequence to generate an intermediate filtered sequence.

[0019] The change rate of adjacent data points in the intermediate filter sequence is checked, abnormal points whose change rate exceeds the preset first peak are deleted, and the gaps are filled by interpolating the previous and next data to obtain the steady-state opening sequence.

[0020] Furthermore, the original wind speed sequence is processed with a dynamic threshold sliding window to generate a valid wind speed sequence, including:

[0021] Obtain historical wind speed data, calculate the moving average and standard deviation based on the historical wind speed data, and dynamically generate the threshold range within the window.

[0022] The original wind speed sequence is segmented by sliding according to the preset window, and the data points in the window that exceed the preset threshold range are deleted.

[0023] The remaining data points are verified in multiple stages, and the verified data are spliced ​​together to generate a valid wind speed sequence.

[0024] Furthermore, the multi-stage verification includes a first-stage verification and a second-stage verification.

[0025] The first stage of verification includes: calculating the wind speed change rate based on the original wind speed sequence; deleting the mutation data whose wind speed change rate between adjacent data points exceeds the preset wind speed threshold.

[0026] The second stage of verification includes: based on the temporal matching of the wind speed and the steady-state opening sequence within the window, deleting the data with correlation lower than the preset correlation threshold.

[0027] Furthermore, the steady-state opening sequence and the effective wind speed sequence are correlated to generate a set of characteristic parameters that characterize the dynamic balance relationship between the opening and the wind speed, including:

[0028] An improved fully connected neural network model is used to process steady-state opening sequence and effective wind speed sequence. The improved fully connected neural network model includes: a dynamic time window module, a fully connected neural network layer, a first hidden layer, a second hidden layer and an output layer.

[0029] The dynamic time window module is used to split the input sequence into continuous time segments of fixed length.

[0030] The fully connected neural network layer is used to establish the time series correspondence between the opening change value and the wind speed fluctuation value in each time segment, and generate the opening-wind speed correlation relationship.

[0031] The first hidden layer is used to perform multivariate fitting of the correlation relationship through a nonlinear activation function, extract the dynamic interaction characteristics of the opening and wind speed, and obtain the intermediate features.

[0032] The second hidden layer is used to compress the dimensions of the intermediate features and filter key information, removing redundant features to generate compressed features.

[0033] The output layer is used to map the compressed features into a parameter set that characterizes the dynamic balance relationship between the opening and the wind speed, thereby obtaining a characteristic parameter set.

[0034] Furthermore, the characteristic parameter set is regularized to generate a characteristic parameter set adapted to the air pressure control logic, including:

[0035] Parameters in the feature parameter set that exceed the preset upper and lower limits are deleted to generate a truncated parameter set.

[0036] The truncated parameter set is linearly mapped to the air pressure control interval to generate a mapping parameter set.

[0037] The mapping parameter set is matched with the preset air pressure control rules to generate a feature parameter set that adapts to the air pressure control logic.

[0038] Furthermore, the characteristic parameter set and the original air pressure sequence are integrated to generate the final damper opening control instruction, including:

[0039] The original air pressure sequence is subjected to multi-sensor collaborative calibration processing to generate a synchronized air pressure sequence.

[0040] An improved convolutional neural network model is used to process the feature parameter set and the synchronized air pressure sequence. The improved convolutional neural network model includes: a convolution layer, a cross-dimensional feature fusion improvement module, a first pooling layer, a second pooling layer and a fully connected output layer.

[0041] The convolutional layer is used to extract the local fluctuation characteristics of the synchronous air pressure sequence.

[0042] The cross-dimensional feature fusion improvement module is used to horizontally splice the local fluctuation feature with the feature parameter set to generate a fusion feature.

[0043] The first pooling layer is used to compress redundant data in the fusion feature to generate compressed features.

[0044] The second pooling layer is used to extract key trend features from the compressed features.

[0045] The fully connected output layer is used to fuse key trend features to generate the final damper opening control instruction.

[0046] Furthermore, the original air pressure sequence is subjected to multi-sensor collaborative calibration processing to generate a synchronized air pressure sequence, including:

[0047] According to the acquisition time difference between adjacent air pressure sensors, delay compensation is performed on the original air pressure sequence.

[0048] The compensated data are phase-aligned according to the timestamps, and the pressure values ​​at the missing time points are filled in by the interpolation algorithm to obtain the pressure sequence.

[0049] Based on the action time points of the steady-state opening sequence, the air pressure sequence is filtered for action correlation, the air pressure fluctuation noise during the damper static period is deleted, and a synchronized air pressure sequence synchronized with the damper action is obtained.

[0050] In a second aspect, the present application provides an underground intelligent damper adaptive adjustment system, characterized in that it includes:

[0051] Data acquisition module: acquires real-time monitoring data from underground wells, including the original sequence of damper opening, the original sequence of wind speed, and the original sequence of air pressure.

[0052] Data preprocessing module: preprocess the original sequence of damper opening and wind speed to obtain steady-state opening sequence and effective wind speed sequence.

[0053] Dynamic correlation analysis module: associates the steady-state opening sequence and the effective wind speed sequence to generate a set of characteristic parameters that characterize the dynamic balance relationship between the opening and wind speed.

[0054] Parameter adaptation module: regularizes the characteristic parameter set to generate a characteristic parameter set that adapts to the air pressure control logic.

[0055] Multi-source feature fusion module: integrates the feature parameter set and the original air pressure sequence to generate the final damper opening control instruction.

[0056] In the third aspect, the present application provides an underground intelligent damper adaptive adjustment device, which includes a memory and a processor; the memory is used to store computer programs; the processor is used to implement the steps of the underground intelligent damper adaptive adjustment method as described in the first aspect when executing the computer program.

[0057] In a fourth aspect, the present application provides a storage medium storing computer program instructions. When the computer program instructions are read and executed by a processor, the steps of the underground intelligent damper adaptive adjustment method as described in the first aspect are executed.

[0058] Beneficial effects of this application:

[0059] This application collects damper opening, wind speed and air pressure data in real time, extracts steady-state and effective sequences after pre-processing to eliminate noise, generates balance parameters through dynamic correlation analysis, and then regularly adapts the air pressure control logic. Finally, it integrates the air pressure characteristics to output precise adjustment instructions. It effectively solves the problem that the monitoring and adjustment method of a single parameter in the existing technology cannot fully reflect the true state of the underground airflow, which easily leads to adjustment lag or over-adjustment. It can capture the coupling relationship between underground wind speed, air pressure and damper opening in real time, and adaptively match the control strategies under different working conditions, thereby improving the adjustment accuracy, accelerating the response speed, ensuring the stable operation of the underground ventilation system, and reducing safety risks.

[0060] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures indicated in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0062] Figure 1 A schematic diagram showing a flow chart of a method for adaptively adjusting an underground intelligent damper according to the present application is shown;

[0063] Figure 2 A module schematic diagram of an underground intelligent damper adaptive adjustment system of the present application is shown. DETAILED DESCRIPTION

[0064] In order to solve the problems raised by the background technology, this application realizes the automatic matching optimization of equipment load and process time consumption, precise coordination of production resource allocation and process requirements through the division of template matching feature set and equipment operation sequence set, and the step-by-step generation of process association matrix, process optimization parameter set, and equipment control instruction set, significantly reduces the production cycle prediction deviation, and simultaneously completes the real-time identification and early warning of abnormal processes, thereby ensuring the stability of the entire process of clothing template production.

[0065] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0066] In some embodiments, as Figure 1 As shown, the present application provides a method for adaptively adjusting an underground intelligent damper, comprising:

[0067] S1. Obtain real-time monitoring data downhole, including the original sequence of damper opening, the original sequence of wind speed, and the original sequence of air pressure.

[0068] S2. Preprocess the original sequence of damper opening and the original sequence of wind speed to obtain a steady-state opening sequence and an effective wind speed sequence.

[0069] S3. Perform correlation processing on the steady-state opening sequence and the effective wind speed sequence to generate a set of characteristic parameters representing the dynamic equilibrium relationship between the opening and the wind speed.

[0070] S4. Regularize the characteristic parameter set to generate a characteristic parameter set that is adapted to the air pressure control logic.

[0071] S5. Perform feature integration on the characteristic parameter set and the original air pressure sequence to generate a final damper opening control instruction.

[0072] In some embodiments, in S2, the original sequence of damper opening and the original sequence of wind speed are preprocessed to obtain a steady-state opening sequence and an effective wind speed sequence, including:

[0073] S21. Perform sliding average filtering on the original sequence of the throttle opening to remove instantaneous jitter and generate a steady-state opening sequence.

[0074] S22. Perform dynamic threshold sliding window processing on the original wind speed sequence to generate a valid wind speed sequence.

[0075] In some embodiments, a sliding average filter is performed on the original sequence of damper openings to remove instantaneous jitter and generate a steady-state opening sequence, including:

[0076] S211. Apply fixed window sliding average filtering to the original sequence of damper opening to generate a preliminary filtered sequence.

[0077] For each data point in the original sequence of the damper opening, a window of fixed length is intercepted with the current point as the center, and the arithmetic mean of all the data in the window is calculated as the filtered value of the current point.

[0078] For example, if the original sequence of damper opening is [10, 15, 20, 18, 12] (unit: %) and the window length is 3, then the filtered value of the second data point is , the filtered value of the third data point is .

[0079] Preliminary filtering can smooth short-term jitter, but may leave some local noise due to the fixed window, such as sudden fluctuations within the window that are not completely suppressed.

[0080] S212. Dynamically adjust the filtering window based on the fluctuation amplitude of the preliminary filtering sequence.

[0081] Specifically, the fluctuation amplitude of adjacent data points in the preliminary filtering sequence is calculated, that is, the absolute value of the difference between the latter data point and the previous data point.

[0082] Adjust the subsequent filtering window size according to the fluctuation amplitude: if the fluctuation amplitude exceeds the preset amplitude threshold, increase the window length; if the fluctuation amplitude exceeds the preset amplitude threshold, maintain the original window length; the amplitude threshold can be based on the fluctuation characteristics of the historical operation data of the downhole damper, and the critical value that can distinguish normal fluctuations from abnormal fluctuations is determined through statistical analysis.

[0083] For example, if the fixed window length is 5 data points, the fluctuation amplitude of a certain section in the preliminary filtering sequence is [2%, 6%, 3%], and the amplitude threshold is 5%, the second window length is increased, and the other window lengths remain unchanged. If the second window length is increased to 7, the corresponding window lengths are adjusted to [5, 7, 5].

[0084] The dynamic window can enhance the smoothing effect in high-fluctuation areas and retain details in low-fluctuation areas, avoiding the over-smoothing or under-smoothing problems caused by fixed windows.

[0085] S213. Use the adjusted filter window to perform a secondary sliding average filter on the original sequence to generate an intermediate filtered sequence.

[0086] For example, if an original sequence is [22, 25, 18, 20, 23] and the dynamic window is 7 points, the window exceeds the data range and is actually calculated based on the maximum available data.

[0087] Secondary filtering further suppresses the residual noise from the initial filtering and generates a smoother intermediate filter sequence.

[0088] S214. Perform change rate verification on adjacent data points of the intermediate filter sequence, delete abnormal points whose change rate exceeds the preset first peak value, and fill the gaps by interpolating the previous and next data to obtain a steady-state opening sequence.

[0089] Specifically, the change rate of adjacent data points in the intermediate filter sequence is calculated, where the change rate is the absolute value of the difference between the next data point and the previous data point divided by the value of the previous data point.

[0090] Data points with a rate of change exceeding the preset first peak are deleted and gaps are filled using linear interpolation. The preset first peak is determined by analyzing the variation patterns of adjacent points in the historical damper opening data, combining the mechanical performance limits of the equipment and underground ventilation safety requirements, and is used to distinguish normal regulation fluctuations from abnormal mutations.

[0091] For example, if the intermediate filter sequence is [20, 25, 35, 22], the first peak is 15%, and the change rate from the second point to the third point is If the value exceeds 15%, the third point 35% is deleted and the previous and next data are interpolated to .

[0092] By checking and filling, the abnormal jumps that still exist after filtering are eliminated, ensuring the continuity and reliability of the steady-state opening sequence.

[0093] In some embodiments, in S22, dynamic threshold sliding window processing is performed on the original wind speed sequence to generate a valid wind speed sequence, including:

[0094] S221. Obtain historical wind speed data, calculate the moving average and standard deviation based on the historical wind speed data, and dynamically generate a threshold range within the window.

[0095] The historical wind speed data may be selected from wind speed data of a historical period similar to the current working conditions, such as the past 30 days.

[0096] The historical data is segmented into sliding windows of the same length (e.g., 10 minutes), and the arithmetic mean μ and standard deviation σ of the wind speed data in each window are calculated. The threshold interval can be positioned as [μ-2σ, μ+2σ], or the 2.5% quantile and 97.5% quantile of the historical data can be taken as the interval boundaries, which is equivalent to covering approximately 95% of the data.

[0097] S222. Slidingly split the original wind speed sequence according to a preset window, and delete data points in the window that exceed a preset threshold range.

[0098] The original sequence of current wind speed is segmented by sliding according to preset windows, and each window contains continuous data points, such as a window of 10 minutes.

[0099] For example, if the threshold interval is: [μ-2σ, μ+2σ], when the wind speed value of a data point in the window exceeds [μ-2σ, μ+2σ], it is marked as abnormal and deleted.

[0100] S223. Perform multi-stage verification on the remaining data points, and splice the verified data to generate a valid wind speed sequence.

[0101] In some embodiments, the multi-stage verification in S223 includes a first-stage verification and a second-stage verification.

[0102] The first phase of verification includes: calculating the wind speed change rate based on the original wind speed sequence; deleting the sudden change data where the wind speed change rate between adjacent data points exceeds the preset wind speed threshold;

[0103] Specifically, the wind speed change rate of adjacent remaining data points is calculated. The wind speed change rate is the absolute value of the difference between the latter data point and the previous data point divided by the value of the previous data point. The mutation data whose change rate exceeds the preset wind speed threshold is deleted. The preset wind speed threshold can be determined through experimental testing and field verification to determine the critical change rate that can filter out wind speed mutation noise while retaining the actual wind speed change, such as 20%.

[0104] The second stage of verification includes: based on the temporal matching of the wind speed and the steady-state opening sequence within the window, deleting the data with correlation lower than the preset correlation threshold.

[0105] Specifically, based on the action time points of the steady-state opening sequence, the temporal correlation between the wind speed data within the window and the damper opening changes is verified. The Pearson correlation coefficient between the wind speed within the window and the steady-state opening sequence is calculated. If the Pearson correlation coefficient is less than a preset correlation threshold, the wind speed within the window and the steady-state opening sequence for that period are deleted. The correlation threshold is determined by analyzing the theoretical coupling relationship between the steady-state opening sequence and wind speed during the damper adjustment process, and combining it with the actual data correlation distribution to set the critical correlation.

[0106] The data after the two-stage verification are spliced ​​in chronological order to generate a valid wind speed sequence.

[0107] The first stage eliminates sudden noise, such as a sudden increase in wind speed caused by a false alarm. The second stage uses the temporal correlation between damper opening and wind speed to eliminate interference data unrelated to damper operation. If the damper does not operate but the wind speed fluctuates abnormally, it is considered invalid and deleted.

[0108] In some embodiments, in S3, the steady-state opening sequence and the effective wind speed sequence are correlated to generate a set of characteristic parameters representing the dynamic balance relationship between the opening and the wind speed, including:

[0109] An improved fully connected neural network model is used to process the steady-state opening sequence and the effective wind speed sequence. The improved fully connected neural network model includes: a dynamic time window module, a fully connected neural network layer, a first hidden layer, a second hidden layer and an output layer.

[0110] The dynamic time window module is used to split the input sequence into continuous time segments of fixed length.

[0111] Specifically, the input sequence is cut into continuous and non-overlapping time segments according to a fixed duration. The fixed duration parameter of the dynamic time window module is a configurable parameter, and its value can be flexibly adjusted according to factors such as the actual monitoring frequency of the underground ventilation system and the damper action response cycle, such as 10 minutes.

[0112] For example, if the steady-state opening sequence is [30%, 32%, 35%, 33%], and the effective wind speed sequence is [4m / s, 4.5m / s, 5m / s, 4.8m / s], two time segments are obtained after segmentation: Time segment 1: opening [30%, 32%], wind speed [4m / s, 4.5m / s]; Time segment 2: opening [35%, 33%], wind speed [5m / s, 4.8m / s].

[0113] The fully connected neural network layer is used to establish the time series correspondence between the opening change value and the wind speed fluctuation value in each time segment, and generate the opening-wind speed correlation relationship.

[0114] The opening change value is the difference between each opening data point in the current period and the previous point. For example, the opening change in period 1 is +2%.

[0115] The wind speed fluctuation value is the difference between each wind speed data point in the current period and the previous point. For example, the wind speed fluctuation in period 1 is +0.5m / s.

[0116] The opening change value and wind speed fluctuation value are input into the fully connected layer, and the correlation parameters are generated through weight calculation, and the correlation parameters are used as the opening-wind speed correlation relationship.

[0117] For example, if the weight of the opening is 0.3 and the weight of the wind speed is 0.7, the associated parameters are: .

[0118] The first hidden layer is used to perform multivariate fitting of the correlation relationship through a nonlinear activation function, extract the dynamic interaction characteristics of the opening and wind speed, and obtain the intermediate features.

[0119] The ReLU (Rectified Linear Unit) activation function is used to process the associated parameters, eliminate negative associations, and obtain dynamic interaction features, which are used as intermediate features.

[0120] The second hidden layer is used to compress the dimensions of the intermediate features and filter key information, removing redundant features to generate compressed features.

[0121] The multi-dimensional features are compressed into a low-dimensional space through the fully connected layer.

[0122] For example, if the dynamic interaction feature is [0.356, 0.5, 0.1], the weight matrix is: , then the compressed features are:

[0123] .

[0124] .

[0125] The compressed feature is [0.3336, 0.3768].

[0126] If the dimension weight is close to 0 after dimensionality reduction, it means that its contribution to the final result is low and can be regarded as redundant; where, close to 0 can specifically be less than 0.2.

[0127] The output layer is used to map the compressed features into a parameter set that characterizes the dynamic balance relationship between the opening and the wind speed, thereby obtaining a characteristic parameter set.

[0128] Mapping compression features to actual control parameters.

[0129] Specifically, the product of feature 1 and feature 2 and the reference multiple can be used as the opening adjustment coefficient and wind speed sensitivity respectively, thereby obtaining a feature parameter set including the opening adjustment coefficient and wind speed sensitivity.

[0130] The benchmark multiplier is determined through model training to map the compression feature to the actual control range. For example, if historical data indicates that the opening adjustment needs to be amplified by 2 to balance wind speed fluctuations, the benchmark multiplier is set to 2.

[0131] In some embodiments, the characteristic parameter set is parameter-regulated in S4 to generate a characteristic parameter set adapted to the air pressure control logic, including:

[0132] S41. Delete parameters in the feature parameter set that exceed preset upper and lower limits to generate a truncated parameter set.

[0133] Specifically, set upper and lower limits for parameters based on the safe range of underground air pressure, for example, 0.9 to 1.1 times standard atmospheric pressure. Deleting parameters that exceed these limits can prevent abnormal parameters from causing damper loss of control. For example, an opening adjustment coefficient of 1.5 can cause overshoot, improving system safety.

[0134] S42. Linearly map the truncated parameter set to the air pressure control interval to generate a mapping parameter set.

[0135] A target range is set according to the downhole gas pressure control requirements. For example, 0.9 to 1.1 times the standard atmospheric pressure corresponds to the numerical interval [0, 1]. The truncation parameter is proportionally scaled to the target range to obtain a mapping parameter set.

[0136] S43. Match the mapping parameter set with the preset air pressure control rules to generate a feature parameter set adapted to the air pressure control logic.

[0137] Define air pressure control rules based on underground ventilation safety strategies. Example air pressure control rules are as follows: When wind speed sensitivity is greater than 0.5, prioritize adjusting the damper opening; when wind speed sensitivity is less than or equal to 0.5, prioritize maintaining the current opening.

[0138] The mapping parameters are matched with the air pressure control rules to generate the final control parameters including the opening priority and wind speed sensitivity, and the final control parameters are used as the feature parameter set.

[0139] For example, if the wind speed sensitivity is 0.8 (>0.5), the characteristic parameter set is [opening priority 1, wind speed sensitivity 0.8]; if the wind speed sensitivity is 0.3, the characteristic parameter set is [opening priority 0, wind speed sensitivity 0.3].

[0140] In some embodiments, in S5, the characteristic parameter set and the original air pressure sequence are integrated to generate a final damper opening control instruction, including:

[0141] S51. Perform multi-sensor collaborative calibration processing on the original air pressure sequence to generate a synchronized air pressure sequence.

[0142] S52. An improved convolutional neural network model is used to process the feature parameter set and the synchronized air pressure sequence. The improved convolutional neural network model includes: a convolution layer, a cross-dimensional feature fusion improvement module, a first pooling layer, a second pooling layer, and a fully connected output layer.

[0143] The convolutional layer is used to extract the local fluctuation features of the synchronized air pressure series.

[0144] The synchronous air pressure sequence is divided into sliding parts according to a fixed time window, and the air pressure fluctuation value within each window is calculated; wherein, the fixed time window is the length of time during which the air pressure change caused by the damper action shows a significant trend by analyzing the historical air pressure data, such as the fixed time window can be 10 minutes.

[0145] For example, if the air pressure change (fluctuation value) in window 1 is 1 kPa; the air pressure change in window 2 is 1 kPa, then the local fluctuation feature is [+1 kPa, +1 kPa].

[0146] The cross-dimensional feature fusion improvement module is used to horizontally splice local fluctuation features and feature parameter sets to generate fusion features.

[0147] The air pressure fluctuation value and the control parameters are horizontally spliced ​​according to the dimension to form a fusion feature.

[0148] For example, if the local fluctuation feature is: [+1 kPa, +1 kPa]; the feature parameter set is: [opening priority 1, wind speed sensitivity 0.8], then the fusion feature is: [+1 kPa, +1 kPa, 1, 0.8].

[0149] The first pooling layer is used to compress the redundant data in the fusion features to generate compressed features.

[0150] The first pooling layer is used to filter out repeated or low-contribution features, for example, by retaining only unique fluctuation values ​​through maximum pooling.

[0151] For example, for the fused feature: [+1 kPa, +1 kPa, 1, 0.8], the compressed feature is: [+1 kPa, 1, 0.8].

[0152] The second pooling layer is used to extract key trend features from the compressed features.

[0153] The second pooling layer is used to analyze and identify the pressure fluctuation trends in continuous windows, such as continuous rise, continuous fall, and smooth fluctuation, thereby identifying the long-term change pattern of air pressure.

[0154] For example, if the fluctuation value of three consecutive windows is +1 kPa, it is determined to be "continuously rising", and the key trend feature is [continuously rising, 1, 0.8].

[0155] The fully connected output layer is used to fuse key trend features to generate the final damper opening control instruction.

[0156] The trend and control parameters are combined according to preset rules to generate control instructions. For example, if the trend is "continuously rising" and the opening priority is 1, the opening adjustment is calculated. The opening adjustment is the product of wind speed sensitivity and the reference adjustment value. The reference adjustment value is used to map characteristic parameters (such as wind speed sensitivity) to the actual opening adjustment, ensuring that the control range is adapted to the underground operating conditions. The reference adjustment value is obtained by fitting historical data during the training of the improved fully connected neural network model.

[0157] For example, if the wind speed sensitivity is 0.8 and the reference adjustment value is 10%, the opening adjustment amount is: , the final damper opening control instruction is: opening increase by 8%.

[0158] In some embodiments, performing multi-sensor collaborative calibration on the original air pressure sequence in S51 to generate a synchronized air pressure sequence includes:

[0159] S511. Perform delay compensation on the original air pressure sequence based on the acquisition time difference between adjacent air pressure sensors.

[0160] For example, if the acquisition delay of sensor 2 relative to sensor 1 is 1 second, the timestamp of sensor 2 is advanced by 1 second and aligned to the time axis of sensor 1.

[0161] S512. Phase alignment is performed on the compensated data according to the timestamps, and the pressure values ​​at the missing time points are filled in by an interpolation algorithm to obtain a pressure sequence.

[0162] Merge the data from sensor 1 and sensor 2, arrange them by a unified timestamp, and use a linear interpolation algorithm to fill in the pressure values ​​at the missing time points. For example, at 10:00:10, the pressure value of sensor 1 is missing, and the pressure values ​​at adjacent time points (10:00:05 and 10:00:15) are interpolated.

[0163] S513. Based on the action time points of the steady-state opening sequence, the air pressure sequence is subjected to action correlation filtering to delete the air pressure fluctuation noise during the damper stationary period, thereby obtaining a synchronized air pressure sequence synchronized with the damper action.

[0164] Extract air pressure data for a specified time period (e.g., 5 minutes) before and after the damper actuation time point (e.g., 10:00:05). If the air pressure fluctuates by more than a preset fluctuation threshold (e.g., ±0.5 kPa) during the damper actuation period (e.g., 10:00:03 to 10:00:07), it is considered valid fluctuation. If the air pressure fluctuates by more than the fluctuation threshold during the damper inactivity period (e.g., 10:00:00 to 10:00:03), it is considered noise. The fluctuation threshold can be the minimum amplitude of valid air pressure fluctuation, calibrated through simulation and measured data, based on the characteristics of the underground ventilation system and the influence of damper actuation on air pressure.

[0165] Invalid fluctuation data during the damper stationary period is deleted, and only valid air pressure fluctuations related to the damper movement are retained to suppress environmental noise interference.

[0166] In some embodiments, as Figure 2 As shown, the present application provides a downhole intelligent damper adaptive adjustment system, which includes:

[0167] Data acquisition module: obtains real-time monitoring data from underground wells, including the original sequence of damper opening, original sequence of wind speed, and original sequence of air pressure.

[0168] Data preprocessing module: preprocess the original sequence of damper opening and wind speed to obtain steady-state opening sequence and effective wind speed sequence.

[0169] Dynamic correlation analysis module: associates the steady-state opening sequence and the effective wind speed sequence to generate a set of characteristic parameters that characterize the dynamic balance relationship between the opening and wind speed.

[0170] Parameter adaptation module: regularizes the characteristic parameter set to generate a characteristic parameter set that adapts to the air pressure control logic.

[0171] Multi-source feature fusion module: integrates the feature parameter set and the original air pressure sequence to generate the final damper opening control instruction.

[0172] In some embodiments, the present application provides a downhole intelligent damper adaptive adjustment device, which includes a memory and a processor; the memory is used to store computer programs; the processor is used to implement the steps of the downhole intelligent damper adaptive adjustment method when executing the computer program.

[0173] In some embodiments, the present application provides a storage medium storing computer program instructions. When the computer program instructions are read and executed by a processor, the steps of the downhole intelligent damper adaptive adjustment method are executed.

[0174] Among them, any reference to memory, storage, database or other media used in the various embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (PROM), electrically erasable programmable ROM (PROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.

[0175] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or elements that are inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device that includes the element.

[0176] Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for adaptively adjusting an underground intelligent damper, characterized in that: include: Acquire downhole real-time monitoring data, including an original sequence of damper opening, an original sequence of wind speed, and an original sequence of air pressure; Preprocess the original sequence of damper opening and wind speed to obtain steady-state opening sequence and effective wind speed sequence; The steady-state opening sequence and the effective wind speed sequence are correlated to generate a set of characteristic parameters that characterize the dynamic balance relationship between opening and wind speed. Regularize the characteristic parameter set to generate a characteristic parameter set that is suitable for the air pressure control logic; Perform feature integration on the characteristic parameter set and the original air pressure sequence to generate the final damper opening control instruction; Among them, the characteristic parameter set and the original air pressure sequence are integrated to generate the final damper opening control instruction, including: Perform multi-sensor collaborative calibration on the original air pressure sequence to generate a synchronized air pressure sequence; An improved convolutional neural network model is used to process the feature parameter set and the synchronized air pressure sequence. The improved convolutional neural network model includes: a convolution layer, a cross-dimensional feature fusion improvement module, a first pooling layer, a second pooling layer, and a fully connected output layer; The convolutional layer is used to extract the local fluctuation characteristics of the synchronous air pressure sequence; The cross-dimensional feature fusion improvement module is used to horizontally splice the local fluctuation feature with the feature parameter set to generate a fusion feature; The first pooling layer is used to compress redundant data in the fusion feature to generate compressed features; The second pooling layer is used to extract key trend features in the compressed features; The fully connected output layer is used to fuse key trend features to generate the final damper opening control instruction.

2. The underground intelligent damper adaptive adjustment method according to claim 1, characterized in that: The original sequence of damper opening and wind speed is preprocessed to obtain the steady-state opening sequence and effective wind speed sequence, including: Perform sliding average filtering on the original sequence of throttle opening to remove instantaneous jitter and generate a steady-state opening sequence; The original wind speed sequence is processed with dynamic threshold sliding window to generate a valid wind speed sequence.

3. The underground intelligent damper adaptive adjustment method according to claim 2, characterized in that: The original sequence of the damper opening is processed by sliding average filtering to remove instantaneous jitter and generate a steady-state opening sequence, including: The original sequence of damper opening is filtered using a fixed window sliding average filter to generate a preliminary filtered sequence. Dynamically adjust the filter window based on the fluctuation amplitude of the preliminary filter sequence; The adjusted filter window is used to perform a secondary sliding average filter on the original sequence to generate an intermediate filtered sequence; The change rate of adjacent data points in the intermediate filter sequence is checked, abnormal points whose change rate exceeds the preset first peak are deleted, and the gaps are filled by interpolating the previous and next data to obtain the steady-state opening sequence.

4. The underground intelligent damper adaptive adjustment method according to claim 3, characterized in that: The original wind speed sequence is processed with a dynamic threshold sliding window to generate a valid wind speed sequence, including: Obtain historical wind speed data, calculate the moving average and standard deviation based on the historical wind speed data, and dynamically generate the threshold range within the window; The original wind speed sequence is segmented by sliding according to the preset window, and the data points in the window that exceed the preset threshold range are deleted; The remaining data points are verified in multiple stages, and the verified data are spliced ​​together to generate a valid wind speed sequence.

5. The underground intelligent damper adaptive adjustment method according to claim 4, characterized in that: The multi-stage verification includes a first-stage verification and a second-stage verification; The first phase of verification includes: calculating the wind speed change rate based on the original wind speed sequence; deleting the sudden change data where the wind speed change rate between adjacent data points exceeds the preset wind speed threshold; The second stage of verification includes: based on the temporal matching of the wind speed and the steady-state opening sequence within the window, deleting the data with correlation lower than the preset correlation threshold.

6. The underground intelligent damper adaptive adjustment method according to claim 1, characterized in that: The steady-state opening sequence and the effective wind speed sequence are correlated to generate a set of characteristic parameters that characterize the dynamic balance relationship between opening and wind speed, including: An improved fully connected neural network model is used to process the steady-state opening sequence and the effective wind speed sequence. The improved fully connected neural network model includes: a dynamic time window module, a fully connected neural network layer, a first hidden layer, a second hidden layer and an output layer; The dynamic time window module is used to split the input sequence into continuous time segments with fixed duration; The fully connected neural network layer is used to establish the time series correspondence between the opening change value and the wind speed fluctuation value in each time segment, and generate the opening-wind speed correlation relationship; The first hidden layer is used to perform multivariate fitting of the correlation relationship through a nonlinear activation function, extract the dynamic interaction characteristics of the opening and wind speed, and obtain the intermediate features; The second hidden layer is used to compress the dimensions of the intermediate features and filter key information, removing redundant features to generate compressed features; The output layer is used to map the compressed features into a parameter set that characterizes the dynamic balance relationship between the opening and the wind speed, thereby obtaining a characteristic parameter set.

7. The underground intelligent damper adaptive adjustment method according to claim 1, characterized in that: Regularize the characteristic parameter set to generate a characteristic parameter set that adapts to the air pressure control logic, including: Delete the parameters in the feature parameter set that exceed the preset upper and lower limits to generate a truncated parameter set; Linearly mapping the truncated parameter set to the air pressure control interval to generate a mapping parameter set; The mapping parameter set is matched with the preset air pressure control rules to generate a feature parameter set that adapts to the air pressure control logic.

8. The underground intelligent damper adaptive adjustment method according to claim 1, characterized in that: Perform multi-sensor collaborative calibration on the original pressure sequence to generate a synchronized pressure sequence, including: According to the acquisition time difference of adjacent air pressure sensors, the original air pressure sequence is delayed and compensated; The compensated data is phase-aligned according to the timestamps, and the pressure values ​​at the missing time points are filled in by the interpolation algorithm to obtain the pressure sequence; Based on the action time points of the steady-state opening sequence, the air pressure sequence is filtered for action correlation, the air pressure fluctuation noise during the damper static period is deleted, and a synchronized air pressure sequence synchronized with the damper action is obtained.

9. An underground intelligent damper adaptive adjustment system, characterized in that: It includes: Data acquisition module: acquires real-time monitoring data downhole, including the original sequence of damper opening, wind speed, and air pressure; Data preprocessing module: preprocesses the original sequence of damper opening and wind speed to obtain steady-state opening sequence and effective wind speed sequence; Dynamic correlation analysis module: associates the steady-state opening sequence and the effective wind speed sequence to generate a set of characteristic parameters that characterize the dynamic balance relationship between opening and wind speed; Parameter adaptation module: regularizes the characteristic parameter set to generate a characteristic parameter set that adapts to the air pressure control logic; Multi-source feature fusion module: integrates the feature parameter set and the original air pressure sequence to generate the final damper opening control instruction; Among them, the characteristic parameter set and the original air pressure sequence are integrated to generate the final damper opening control instruction, including: Perform multi-sensor collaborative calibration on the original air pressure sequence to generate a synchronized air pressure sequence; An improved convolutional neural network model is used to process the feature parameter set and the synchronized air pressure sequence. The improved convolutional neural network model includes: a convolution layer, a cross-dimensional feature fusion improvement module, a first pooling layer, a second pooling layer, and a fully connected output layer; The convolutional layer is used to extract the local fluctuation characteristics of the synchronous air pressure sequence; The cross-dimensional feature fusion improvement module is used to horizontally splice the local fluctuation feature with the feature parameter set to generate a fusion feature; The first pooling layer is used to compress redundant data in the fusion feature to generate compressed features; The second pooling layer is used to extract key trend features in the compressed features; The fully connected output layer is used to fuse key trend features to generate the final damper opening control instruction.

Images