A ventilation data detection and analysis method for coal mine production

Abstract

The present invention relates to the technical field of ventilation data detection and analysis, and particularly to a method for detecting and analyzing ventilation data for coal mine production. The method includes: obtaining a first sequence, a second sequence, and a third sequence; using an initial time window to intercept historical data from the current moment of the first sequence towards the previous moment, and calculating the historical data difference; further obtaining an optimized time window, and calculating the historical weighted mean of the first sequence; calculating the conditional variance at each moment within the optimized time window, and obtaining the conditional variance at future moments to form a future sequence; calculating the correlation between the conditional variance at each moment within the optimized time window and the future sequence, and then calculating the anomaly degree of the first sequence; similarly obtaining the anomaly degrees of the second and third sequences; judging whether the ventilation is abnormal according to the anomaly degrees of the first, second, and third sequences, and if not, further judging using the carbon dioxide concentration. The present invention can accurately detect ventilation anomalies in coal mines.

Description

Technical Field

[0001] The present invention relates to the technical field of ventilation data detection and analysis, and particularly to a method for detecting and analyzing ventilation data for coal mine production. Background Art

[0002] Ventilation systems are widely used in the coal mine production industry, and their operation safety and efficiency issues have become the focus of industry attention. At present, the monitoring of ventilation systems often faces problems such as difficulty in identifying potential risks and frequent false alarms in complex environments, making it difficult to meet the requirements of high-precision monitoring in modern chemical environments.

[0003] At present, the monitoring of ventilation systems faces many challenges in complex environments, mainly manifested as difficulty in identifying potential risks and frequent false alarms and missed alarms. In coal mine production, the collection of ventilation data mainly relies on sensors, often involving key parameters such as wind speed, wind pressure, and carbon dioxide concentration. However, existing monitoring methods usually analyze the collected multi-dimensional feature data and use the Generalized Autoregressive Conditional Heteroskedasticity algorithm (GARCH algorithm) to predict the volatility of ventilation parameters. This algorithm identifies the volatility of data within the historical window by introducing lag terms and conditional variances, gradually analyzes the multi-dimensional features within the historical window, and sets a threshold range to distinguish normal and abnormal fluctuations. This algorithm can accurately model the dynamic fluctuation characteristics of ventilation parameters, comprehensively analyze the correlation between multi-features, provide accurate analysis of fluctuation trends and risk assessments, but due to its limitation to the prediction results within the historical window, it is easily interfered by external factors, leading to misjudgment or failure to detect potential risks in a timely manner. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a method for detecting and analyzing ventilation data for coal mine production, and the specific technical solutions adopted are as follows:

[0005] An embodiment of the present invention provides a method for detecting and analyzing ventilation data for coal mine production, the method comprising:

[0006] Collecting wind speed data, wind pressure data, and carbon dioxide concentration data and performing preprocessing to obtain a first sequence, a second sequence, and a third sequence respectively;

[0007] Using an initial time window to intercept historical data from the current moment of the first sequence to previous moments, and calculating the historical data difference of the first sequence according to the historical data at each moment in the initial time window and the smoothed value obtained by the exponential smoothing algorithm;

[0008] Adjusting the time length of the initial time window according to the historical data difference to obtain an optimized time window corresponding to the first sequence; calculating the historical weighted mean of the first sequence based on the historical data within the optimized time window;

[0009] Calculate the conditional variance at each moment within the optimized time window using the historical weighted mean through the GARCH algorithm; calculate the conditional variance at future moments by sliding the optimized time window backward from the current moment; take a preset number of conditional variances starting from the first future moment after the current moment to form a future sequence;

[0010] Calculate the correlation between the conditional variances at each moment within the optimized time window and the future sequence; calculate the abnormality degree of the first sequence based on the correlation, the conditional variances at the current moment and the next moment; similarly obtain the abnormality degrees of the second and third sequences;

[0011] Judge whether the ventilation is abnormal based on the abnormality degrees of the first, second, and third sequences. If it is not abnormal, then judge whether the ventilation is abnormal using the carbon dioxide concentration between the outside of the exhaust vent and the middle of the ventilation duct inside the exhaust vent.

[0012] Preferably, collect wind speed data, wind pressure data, and carbon dioxide concentration data and perform preprocessing to obtain the first sequence, the second sequence, and the third sequence respectively, including:

[0013] The preprocessing includes filling in missing values, removing outliers, noise reduction, and normalization processing, and forming three different sequences from the normalized wind speed data, wind pressure data, and carbon dioxide concentration data respectively.

[0014] Preferably, the calculation formula for the historical data difference of the first sequence is:

[0015] ,

[0016] where, represents the historical data difference of the first sequence; θ and represent weight coefficients and have equal values; N represents the number of moments within the initial time window; Norm represents the normalization operation; represents the historical data at the i-th moment within the initial time window; represents the smoothed value obtained by the exponential smoothing algorithm at the i-th moment within the initial time window; δ represents the sensitivity parameter; λ and μ represent the first scaling coefficient and the second scaling coefficient; e represents the natural constant; σ represents the third scaling coefficient; represents the wind speed data at the current moment.

[0017] Preferably, adjust the time length of the initial time window according to the historical data difference to obtain the optimized time window corresponding to the first sequence, including:

[0018] Set a differential weight coefficient. If the difference of historical data is greater than or equal to the threshold, the differential weight coefficient takes the value of the first preset value. If the difference of historical data is less than the threshold, the differential weight coefficient takes the value of the second preset value; Multiply the differential weight coefficient by the difference of historical data, add the result of the multiplication to the third preset value and take the reciprocal to obtain the window time length adjustment coefficient; Multiply the window time length adjustment coefficient by the time length of the initial time window and round to obtain the time length of the optimized time window.

[0019] Preferably, calculating the historical weighted mean of the first sequence based on the historical data within the optimized time window includes:

[0020] Subtract the time of a historical data within the optimized time window from the current time to obtain a difference result; Set a time distance scaling parameter, and normalize the value obtained by multiplying the time distance scaling parameter by the difference result using the exponential function with the natural constant as the base to obtain the average weight of the historical data; Perform weighted averaging based on each historical data within the optimized time window and the corresponding average weight to obtain the historical weighted mean of the first sequence.

[0021] Preferably, calculating the conditional variance at each moment within the optimized time window using the historical weighted mean through the GARCH algorithm includes:

[0022] Obtain the calculation formula of the conditional variance in the GARCH algorithm; Calculate the residuals at each moment within the optimized time window, and the residual at each moment is the difference between the data value at the previous moment and the historical weighted mean; Obtain the conditional variance at this moment according to the calculation formula of the conditional variance and the residual and conditional variance at the previous moment of this moment.

[0023] Preferably, using the optimized time window to slide backward from the current time to calculate the conditional variance at future times includes:

[0024] Use the residual and conditional variance at the current time and the calculation formula of the conditional variance to calculate the conditional variance at the first future time of the current time; When calculating the conditional variance at the second future time, the optimized time window slides backward from the current time by a preset step length, and use the mean of the residuals of other times within the optimized time window except the first future time as the residual in the calculation formula of the conditional variance to calculate the conditional variance at the second future time; When calculating the conditional variance at the third future time, the optimized time window slides backward from the current time by two preset step lengths, and use the mean of the residuals of other times within the optimized time window except the first and second future times as the residual in the calculation formula of the conditional variance to calculate the conditional variance at the third future time, and so on, to calculate the conditional variance at each future time.

[0025] Preferably, calculating the abnormality degree of the first sequence according to the correlation, the conditional variance at the current moment and the next moment includes:

[0026] Setting a first weight corresponding to the conditional variance at the current moment and a second weight corresponding to the conditional variance at the first future moment; performing weighted summation on the conditional variances at the current moment and the first future moment using the first weight and the second weight to obtain a summation result; performing a negative correlation mapping on the summation result using an exponential function with the natural constant as the base to obtain a mapping result; the difference between a third preset value and the mapping result is the abnormality degree of the first sequence; wherein, if the correlation is greater than zero, the values of the first weight and the second weight are respectively a fourth preset value and a fifth preset value, if the correlation is less than or equal to zero, the values of the first weight and the second weight are respectively the fifth preset value and the fourth preset value, and the fourth preset value is greater than the fifth preset value.

[0027] Preferably, judging whether the ventilation is abnormal according to the abnormality degrees of the first, second, and third sequences includes:

[0028] When the abnormality degrees of the first, second, and third sequences are all less than the reference threshold, the ventilation is normal, and when one of the abnormality degrees of the first, second, and third sequences is greater than or equal to the reference threshold, the ventilation is abnormal.

[0029] Preferably, judging whether the ventilation is abnormal by using the carbon dioxide concentration between the outside of the exhaust port and the middle of the ventilation duct inside the exhaust port includes:

[0030] Performing mapping on the difference in the carbon dioxide concentration between the outside of the exhaust port and the middle of the ventilation duct inside the exhaust port using the sigmod function to obtain a concentration difference; if the concentration difference is greater than the concentration difference threshold, the ventilation is normal, and if the concentration difference is less than or equal to the concentration difference threshold, the ventilation is abnormal.

[0031] The embodiments of the present invention have at least the following beneficial effects: The present application preprocesses the wind speed data, wind pressure data, and carbon dioxide concentration data to obtain the first sequence, the second sequence, and the third sequence, which can improve the data quality and the accuracy of subsequent analysis. Further, the historical data difference of the first sequence is calculated, and then the time length of the initial time window is adjusted according to the difference to obtain an optimized time window, ensuring that the change of the wind speed data in the first sequence can be accurately captured. The conditional variance at each moment within the optimized time window is calculated through the GARCH algorithm, and the conditional variance at future moments forms a future sequence. Then, the correlation between the conditional variance at each moment within the optimized time window and the future sequence is calculated, and the anomaly degree of the first sequence is calculated based on the correlation, the conditional variance at the current moment, and the conditional variance at the next moment. That is, the volatility of the membership data within the optimized time window and the volatility of the predicted data at future moments are analyzed, the correlation between the two is analyzed, and the data fluctuation at the current moment is further analyzed based on the correlation to obtain the anomaly degree, which can solve the defect of a single historical window in the traditional method, achieve a comprehensive and accurate detection effect, and obtain a more accurate anomaly degree. Finally, it is judged whether the ventilation is abnormal according to the anomaly degrees of the first, second, and third sequences. If not, the carbon dioxide concentration between the outside of the exhaust port and the middle of the ventilation duct inside the exhaust port is used to judge whether the ventilation is abnormal. When it is judged that the ventilation is normal, a further judgment is made to further improve the accuracy of the detection of ventilation anomalies. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 It is a flowchart of a method for detecting and analyzing ventilation data for coal mine production provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in combination with the drawings and preferred embodiments, details the specific embodiments, structures, features, and effects of a method for detecting and analyzing ventilation data for coal mine production proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this invention belongs.

[0036] The following specifically describes the specific solution of a ventilation data detection and analysis method provided by the present invention in combination with the accompanying drawings.

[0037] Example:

[0038] The main application scenario of the present invention is: The present invention is mainly applied to the analysis of key ventilation data in the ventilation system under the coal mine production scenario, and then to judge whether ventilation abnormalities occur, ensuring the production safety of coal mines.

[0039] Please refer to Figure 1 , which shows the method flow chart of a ventilation data detection and analysis method provided by an embodiment of the present invention. The method includes the following steps:

[0040] Step S1, collect wind speed data, air pressure data, and carbon dioxide concentration data and perform preprocessing to obtain the first sequence, the second sequence, and the third sequence respectively.

[0041] In the coal mine production environment, sensors are arranged at key positions such as main roadways, branch roadways, and fan outlets to collect the core parameters of the ventilation system in real time, including wind speed data, air pressure data, carbon dioxide concentration data, etc. These sensors need to meet the usage requirements of complex mine environments and have the ability to resist moisture, high temperature, and dust. The collected data is uploaded to the ground monitoring center through wireless transmission (such as LoRa or WiFi), and at the same time, edge computing devices are used for preliminary processing. The collection frequency is generally set to collect once every minute to ensure that the dynamic changes of the system operation can be captured.

[0042] For the collected wind speed data, air pressure data, and carbon dioxide concentration data, preprocessing is required. The preprocessing includes filling in missing values, removing outliers, noise reduction, and normalization processing.

[0043] Specifically, for missing values, interpolation can be used to fill them in; for outliers (such as wind speed <0 or carbon dioxide concentration exceeding the physical limit value), they are removed in combination with the upper and lower limit rules. All sensor data is aligned according to the time stamp to ensure the temporal consistency of different features. Kalman filtering is used to reduce noise in the data, smooth short-term fluctuations, and eliminate noise caused by sensor errors or environmental interference. For example, in the wind speed data, tiny irrelevant fluctuations caused by high equipment sensitivity are filtered out to make the result more stable. Finally, the wind speed data, air pressure data, and carbon dioxide concentration data are normalized to ensure that the scales of the three types of data are consistent, facilitating the comprehensive analysis of different features.

[0044] The normalized wind speed data, wind pressure data, and carbon dioxide concentration data are respectively composed of three different sequences to obtain the first sequence, the second sequence, and the third sequence.

[0045] Step S2: Use the initial time window to intercept historical data from the current moment of the first sequence to the previous moments, and calculate the historical data difference of the first sequence according to the historical data at each moment in the initial time window and the smoothed value obtained by the exponential smoothing algorithm.

[0046] In the coal mine production environment, to judge the operation status of the ventilation system, combined with multi-feature hierarchical analysis, progressive judgment is carried out on wind speed, wind pressure, and carbon dioxide concentration data.

[0047] The simple overview of the generalized autoregressive conditional heteroskedasticity algorithm is as follows:

[0048] Determine the structure of the time series data and assume that its mean conforms to a specific regression model (such as an autoregressive model). Model the variance of the residuals, assuming that the conditional variance depends on the sum of the squares of the historical residuals and the weighted sum of the past conditional variances;

[0049] Based on historical data, through maximum likelihood estimation or other optimization methods, fit the parameters of the model, including the regression coefficients and the weights of the conditional variances. According to the estimated parameters, gradually calculate the conditional variance to quantify the volatility of the time series at each moment;

[0050] Use the fitted model to predict the conditional variance in the future time period, analyze the future volatility trend. At the same time, combine historical volatility and predicted volatility to evaluate the risk characteristics of the data and provide support for decision-making.

[0051] Furthermore, it is necessary to conduct a preliminary analysis on the wind speed data, wind pressure data, and carbon dioxide concentration data, and then optimize and adjust the fixed historical time window so that it can more accurately quantify the characteristics of the data.

[0052] There are three types of data in this application, and the operations required for them are basically the same. Therefore, the wind speed data, that is, the first sequence, is taken as an example for analysis.

[0053] First, set a fixed historical time window, denoted as the initial time window , At the same time, represent the time length of the window. Use the initial time window to intercept historical data from the current moment of the first sequence to the previous moments, and obtain the historical data at each moment within the initial time window. The historical data at each moment within the initial time window is the wind speed data from the past moment to the current moment including and Wind speed data at a moment. Further, calculate the historical data difference of the first sequence based on the historical data at each moment in the initial time window and the smoothed value obtained through the exponential smoothing algorithm.

[0054] The specific calculation formula for its historical data difference is:

[0055] ,

[0056] where, represents the historical data difference of the first sequence, and can also be called the historical data difference corresponding to the current moment of the first sequence; θ and represent the weight coefficients, and their values are equal; N represents the number of moments within the initial time window; Norm represents the normalization operation; represents the historical data at the i-th moment within the initial time window; represents the smoothed value obtained through the exponential smoothing algorithm at the i-th moment within the initial time window; δ represents the sensitivity parameter; λ and μ represent the first scaling coefficient and the second scaling coefficient; e represents the natural constant; σ represents the third scaling coefficient; represents the wind speed data at the current moment.

[0057] In the above formula, the value of the weight coefficient is 0.5, and the implementer can adjust it according to the actual situation; represents the smoothed value at the i-th moment obtained through the exponential smoothing algorithm, which is a well-known technology, and the specific calculation formula is , represents the smoothed value at the i-th moment obtained through the exponential smoothing algorithm, represents the historical data at the i-th moment within the initial time window, that is, the real data at the i-th moment. The value of α can be corrected according to the smoothed value result. In the embodiment of the present invention, the reference value is given as 0.8, represents the smoothed value at the (i - 1)-th moment obtained through the exponential smoothing algorithm; the reference value of the sensitivity parameter δ can be taken as 2, which is to control the sensitivity of the difference between the actual historical data and the smoothed value. When predicting the smoothed value, the law of data mutation will not be known in advance. Therefore, if the wind speed data mutates at a moment, the difference between its actual value and the smoothed value will be relatively large. Therefore, the greater the difference between the smoothed value and the actual historical data, the more unstable the historical data; is to balance the difference between the actual historical data and the smoothed value, amplify the difference, and improve the sensitivity to abnormal data mutation. The values of λ and μ are 0.5 and 0.1 respectively. The role of μ is the scaling coefficient of the exponential function with e as the base, and the role of λ is the exponential function with e as the base used to adjust the growth and decay amplitude, which can determine the maximum influence degree of the exponential term. The implementer can adjust it according to the actual situation; It represents the value obtained by normalizing the difference between the wind speed data at the current moment and other data within the initial time window. σ is its scaling coefficient. The greater the difference between the wind speed data at the current moment and other data within the window, the more unstable the data indicates.

[0058] From this, the historical data difference corresponding to the first sequence can be obtained, and then the initial time window can be optimized and adjusted.

[0059] Step S3: Adjust the time length of the initial time window according to the historical data difference to obtain the optimized time window corresponding to the first sequence; calculate the historical weighted mean of the first sequence based on the historical data within the optimized time window.

[0060] In step S2, the historical data difference of the first sequence is obtained. When the historical data difference is large, it indicates that the data volatility is strong, which means that the system state may change rapidly, indicating that there are unstable fluctuations in the wind speed data. At this time, if the initial time window is too large, the smoothing effect of the historical data will reduce the sensitivity of the model to sudden changes, making the abnormal signal diluted. For example, if the ventilation system fails within a short period of time and the window is too large, the model may not be able to quickly identify this problem due to the long-term average effect. Narrowing the window can reduce the interference of historical data on the current decision and enhance the response speed to sudden changes. In the case of strong volatility, the short-term trend information of the data is more important. If the window is too large, the short-term changes will be masked by the historical data over a longer period of time, thus reducing the accurate assessment of the current state. In the coal mine ventilation system, if the wind speed data rises rapidly within a short period of time and the window is too large, the model may still consider the overall trend stable, resulting in a lag in anomaly detection. Narrowing the window can focus on the recent data and improve the perception ability of the latest changes.

[0061] When the historical data difference of the first sequence is small and the data fluctuation is weak, the system is usually in a relatively stable state. At this time, the short-term small fluctuations may be just random noise. If the window is too small, the model may be overly sensitive to these small fluctuations, resulting in false alarms. For example, when the ventilation system is operating normally, there may be slight fluctuations in the wind speed, but these fluctuations do not represent system anomalies. If the window is too small, the model may misjudge that there is a problem with the system, while a larger window can smooth these short-term disturbances and ensure that the model does not make wrong judgments about irrelevant fluctuations. The change of data at a single moment may not reflect the overall state of the system. Enlarging the window can provide more comprehensive time series information, enabling the model to more accurately identify trends.

[0062] Further, the time length of the initial time window is adjusted according to the historical data difference to obtain the optimized time window corresponding to the first sequence. Specifically, a difference weight coefficient is set. If the historical data difference is greater than or equal to the threshold, the difference weight coefficient takes the value of the first preset value. If the historical data difference is less than the threshold, the difference weight coefficient takes the value of the second preset value. Multiply the difference weight coefficient by the historical data difference, add the result of the multiplication to the third preset value and take the reciprocal to obtain the window time length adjustment coefficient. Multiply the window time length adjustment coefficient by the time length of the initial time window and round to obtain the time length of the optimized time window.

[0063] Specifically, the calculation formula is:

[0064] ,

[0065] where W represents the result of multiplying the window time length adjustment coefficient by the time length of the initial time window. Further, W needs to be rounded. The rounding principle follows the rounding rule to obtain the time length of the optimized time window , represents the time length of the initial time window; β represents the difference weight coefficient, represents the historical data difference. If the historical data difference is greater than or equal to the threshold, it is considered that the difference is large and the wind speed data fluctuates unstably. At this time, the time length of the initial time window needs to be reduced, and the value of the difference weight coefficient is the first preset value, and the first preset value is 0.9. If the historical data difference is less than the threshold, it is considered that the difference is small and the wind speed data is stable. At this time, the time length of the initial time window needs to be increased, and the value of the difference weight coefficient is the second preset value, and the second preset value is -1. The value of the threshold is 0.5. The implementer can adjust the values of the first preset value, the second preset value and the threshold according to the actual situation, but it should be noted that when adjusting the first preset value and the second preset value, it is necessary to follow the principle of reducing the window time length for large differences and increasing the window time length for small differences; the window time length adjustment coefficient is , and the third preset value is 1.

[0066] Next, after obtaining the optimized time window corresponding to the first sequence, it is necessary to analyze the historical data within the optimized time window and calculate the historical weighted mean of the first sequence to facilitate the application of the subsequent GARCH algorithm.

[0067] Among them, the data at the last moment within the optimized time window is the wind speed data at the current moment, and other data are the historical wind speed data before the current moment.

[0068] Subtract the current moment from the moment of a historical data within the optimized time window to obtain a difference result; set a time distance scaling parameter, and use the exponential function with the natural constant as the base to normalize the value obtained by multiplying the time distance scaling parameter by the difference result to obtain the average weight of the historical data; perform a weighted average based on each historical data within the optimized time window and the corresponding average weight to obtain the historical weighted mean of the first sequence.

[0069] Its specific calculation formula is:

[0070] ,

[0071] where, represents the historical weighted mean of the first sequence; represents the time length of the optimized time window of the first sequence, which can also represent the set of each moment within the optimized time window here; e represents the natural constant; p represents the time distance scaling parameter, and the reference value in the embodiments of the present invention is 0.5, is the difference result between the current moment and the moment of a historical data within the optimized time window. The farther away from the current moment, its average weight is smaller, represents the historical data at the t-th moment within the optimized time window. Thus, the historical weighted mean of the first sequence can be obtained, and the GARCH algorithm is applied based on the historical weighted mean.

[0072] Step S4, calculate the conditional variance of each moment within the optimized time window using the historical weighted mean through the GARCH algorithm; use the optimized time window to slide backward from the current moment to calculate the conditional variance of future moments; take a preset number of conditional variances starting from the first future moment after the current moment and form a future sequence backward.

[0073] In step S3, the historical weighted mean of the first sequence is obtained. To evaluate the volatility of historical data, it is necessary to calculate the conditional variance of each moment within the optimized time window based on the historical weighted mean of the first sequence using the GARCH algorithm.

[0074] Obtain the calculation formula of the conditional variance in the GARCH algorithm; calculate the residuals of each moment within the optimized time window. The residual of each moment is the difference between the data value of the previous moment and the historical weighted mean; obtain the conditional variance of the moment according to the calculation formula of the conditional variance and the residual and conditional variance of the previous moment of the moment.

[0075] Specifically, the calculation formula of the conditional variance is:

[0076] ,

[0077] where, represents the conditional variance at the t-th moment; , , represent the parameters of the model, and the initial reference values can be taken as 0.4, 0.3, 0.3; represents the residual at the previous moment, which is the difference between the data value at the previous moment and the historical weighted mean ; represents the conditional variance at the previous moment, that is, the volatility calculated based on the known information (historical data) at the (t - 1)-th moment, which is used to measure the fluctuation intensity of the data at the previous moment. Thus, the conditional variances at each moment within the optimized time window can be obtained, which is a well-known technology and will not be elaborated here.

[0078] Thus, the conditional variances at each moment within the optimized time window can be obtained. Further, it is necessary to judge the correlation between the volatility at future moments and the volatility at historical moments, and then analyze the operating state of the ventilation system at the current moment. Therefore, it is necessary to calculate the conditional variance at future moments.

[0079] Specifically, it is necessary to slide the optimized time window backward, set a preset step length. Preferably, in the embodiments of the present invention, the preset step length is set to one moment. Using the residual and conditional variance at the current moment and the calculation formula of the conditional variance to calculate the conditional variance at the first future moment of the current moment; when calculating the conditional variance at the second future moment, the optimized time window slides backward one preset step length from the current moment, and uses the mean value of the residuals of other moments except the first future moment within the optimized time window as the residual in the calculation formula of the conditional variance to calculate the conditional variance at the second future moment; when calculating the conditional variance at the third future moment, the optimized time window slides backward two preset step lengths from the current moment, and uses the mean value of the residuals of other moments except the first and second future moments within the optimized time window as the residual in the calculation formula of the conditional variance to calculate the conditional variance at the third future moment, and so on, to calculate the conditional variances at each future moment.

[0080] The residual is the core for calculating the conditional variance, which reflects the influence of past errors on the current fluctuation. However, since some historical residuals will exceed the current calculation window when the window slides, these residuals cannot be directly used. Using the mean value of the residuals in the historical window to replace the missing residuals has a smoothing effect, which can effectively reduce the influence of short-term fluctuations on the prediction. By using the mean value for approximation, it is possible to reduce the interference of sudden abnormal data on the model prediction. Especially when some specific residual information is missing, the mean value can provide a reasonable estimate to ensure the accuracy of the prediction of the conditional variance at future moments.

[0081] After obtaining the conditional variance at a future moment, a future sequence is formed by taking a preset number of conditional variances backward starting from the first future moment after the current moment, where the preset number is equal to the number of moments included in the optimization time window.

[0082] Step S5: Calculate the correlation between the conditional variances at each moment within the optimization time window and the future sequence; calculate the abnormality degree of the first sequence based on the correlation, the conditional variances at the current moment and the next moment; similarly, obtain the abnormality degrees of the second and third sequences.

[0083] In step S4, the future sequence is obtained, that is, the conditional variances corresponding to each future moment. Furthermore, the correlation between the conditional variances at each moment within the optimization time window and the conditional variances of the future sequence is analyzed, and the correlation between the volatility at historical moments and the volatility at future moments is analyzed. Specifically, calculate the Pearson correlation coefficient between the conditional variances at each moment within the optimization time window and the conditional variances of the future sequence. The larger this coefficient is, the more correlated the volatility at historical moments and the volatility at future moments are. The smaller this coefficient is, the less correlated or weakly correlated the volatility at historical moments and the volatility at future moments are.

[0084] Denote the correlation between the conditional variances at each moment within the optimization time window and the future sequence as , if is greater than 0, there is a positive correlation between the conditional variances at each moment within the optimization time window and the future sequence; if is less than or equal to 0, there is a negative correlation or weak correlation between the conditional variances at each moment within the optimization time window and the future sequence.

[0085] A positive correlation indicates that the greater the volatility of historical wind speed data, the greater the possible volatility of future wind speed data, and the smaller the volatility of historical wind speed data, the smaller the possible volatility of future wind speed data. By amplifying the weight of historical volatility, the volatility at the current moment can be more accurately represented. A negative correlation and a weak correlation indicate that the greater the volatility of historical wind speed data, the smaller or unchanged the possible volatility of future wind speed data, and the smaller the volatility of historical wind speed data, the greater or unchanged the possible volatility of future wind speed data. In this case, it is necessary to amplify the weight of the volatility of wind speed at future moments to weaken the influence of historical wind speed data and increase the attention to the volatility at future moments.

[0086] Further, calculate the abnormality degree of the first sequence according to the correlation, the conditional variance at the current moment, and the conditional variance at the next moment. Set the first weight corresponding to the conditional variance at the current moment and the second weight corresponding to the conditional variance at the first future moment; use the first weight and the second weight to perform weighted summation on the conditional variances at the current moment and the first future moment to obtain a summation result; use the exponential function with the natural constant as the base to perform a negative correlation mapping on the summation result to obtain a mapping result; the difference between the third preset value and the mapping result is the abnormality degree of the first sequence; wherein, if the correlation is greater than zero, the values of the first weight and the second weight are the fourth preset value and the fifth preset value respectively, if the correlation is less than or equal to zero, the values of the first weight and the second weight are the fifth preset value and the fourth preset value respectively, and the fourth preset value is greater than the fifth preset value.

[0087] Specifically, the calculation formula is:

[0088] ,

[0089] wherein, represents the abnormality degree of the first sequence, represents the natural constant, represents the result of performing weighted summation on the conditional variances at the current moment and the first future moment using the first weight and the second weight, represents the negative correlation mapping of the summation result, and respectively represent the conditional variances at the current moment and the first future moment. The smaller these two values are, the more stable the wind speed data is; and respectively represent the first weight and the second weight. If is greater than 0, the values are the fourth preset value 0.6 and the fifth preset value 0.4 respectively. This is to magnify the importance of historical wind speed data. If is less than or equal to 0, the values are the fifth preset value 0.4 and the fourth preset value 0.6 respectively to increase the attention to the future moment.

[0090] The larger the value of the abnormality degree, the higher the degree of fluctuation. At this time, the degree of abnormality is higher, and it is necessary to remind the staff to conduct further inspections. Similarly, by performing the same processing on the second sequence and the third sequence, the corresponding difference degrees can be obtained.

[0091] The abnormality degrees of the first, second, and third sequences are the difference degrees corresponding to the wind speed data, the wind pressure data, and the carbon dioxide concentration data respectively, and can also be called the abnormality degrees of each type of data at the current moment. As time goes on, an abnormality degree can be calculated for each moment, and thus the operating state of the ventilation system can be analyzed in real time.

[0092] Step S6: Determine whether the ventilation is abnormal based on the abnormality levels of the first, second, and third sequences. If it is not abnormal, then determine whether the ventilation is abnormal by using the carbon dioxide concentration between the outside of the exhaust vent and the middle of the ventilation duct inside the exhaust vent.

[0093] In step S5, the difference levels corresponding to the wind speed data, wind pressure data, and carbon dioxide concentration data are obtained. Thus, a comprehensive analysis can be performed based on these three different types of data. Set a reference threshold, with the initial value of the reference threshold being 0.8, denoted as. The reference threshold can be adjusted according to be adjusted, represents the mean of the abnormality levels of one of the three types of data in a stable state, represents the standard deviation of the abnormality level of one type of data in a stable state, and k represents the adjustment coefficient, with a value of 2 that can be adjusted according to the actual situation. Each of the three types of data corresponds to a reference threshold, and all of them can be adjusted subsequently.

[0094] When the abnormality levels of the first, second, and third sequences are all less than the reference threshold, the ventilation is normal. When one of the abnormality levels of the first, second, and third sequences is greater than or equal to the reference threshold, the ventilation is abnormal. If the ventilation is abnormal, prompt the staff to conduct an inspection to check whether there are abnormalities in the hardware equipment, circuits, ventilation ducts, etc. of the ventilation system.

[0095] When the ventilation is normal, for a more accurate judgment, further judgment is required. Specifically, install a carbon dioxide concentration detection sensor between the outside of the exhaust vent and the middle of the ventilation duct inside the exhaust vent to obtain the carbon dioxide concentration between the outside of the exhaust vent and the middle of the ventilation duct inside the exhaust vent; use the sigmod function to map the difference in the carbon dioxide concentration between the outside of the exhaust vent and the middle of the ventilation duct inside the exhaust vent to obtain the concentration difference. If the concentration difference is greater than the concentration difference threshold, the ventilation is normal. If the concentration difference is less than or equal to the concentration difference threshold, the ventilation is abnormal. The concentration difference threshold is 0.5.

[0096] Carbon dioxide will diffuse outward through the fan installed at the exhaust vent. At this time, if the ventilation is normal, the carbon dioxide concentration in the middle of the ventilation duct inside the exhaust vent or at the position of the ventilation duct far from the exhaust vent will be relatively low, and the difference in the carbon dioxide concentration between the outside of the vent and the middle of the ventilation duct inside the exhaust vent is greater than zero. After being mapped by the sigmod function, it is greater than 0.5. If the difference in the carbon dioxide concentration between the outside of the vent and the middle of the ventilation duct inside the exhaust vent is less than or equal to zero, generally, it is because the ventilation is not smooth, resulting in the carbon dioxide concentration in the middle of the ventilation duct being higher than that outside the fan exhaust vent. Thus, a more accurate judgment can be obtained.

[0097] Finally, regularly collect the latest wind speed, wind pressure, and carbon dioxide concentration data from the coal mine production environment, and use preprocessing techniques to remove noise and outliers to ensure data quality. By dynamically monitoring the prediction error of the model, evaluate its adaptability to the current environment. If the error gradually increases, indicating that the environmental conditions or data distribution have changed, then retrain the model with new data and adjust parameters such as the exponential smoothing factor or weighted window strategy to better capture the changing trend of the current features. In addition, during the update process, fully combine the results of dual-window analysis and volatility assessment to further optimize the feature extraction ability of the model and ensure that it can continuously and accurately judge the abnormal state of ventilation data. Through this dynamic parameter adjustment mechanism, it can meet the detection requirements of the coal mine ventilation system in the long term.

[0098] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. And the above has described specific embodiments of this specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0099] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

[0100] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A ventilation data detection and analysis method for coal mine production, characterized in that: The method includes: Collect wind speed data, wind pressure data and carbon dioxide concentration data and perform preprocessing to obtain the first sequence, the second sequence and the third sequence respectively; The initial time window is used to intercept historical data from the current moment of the first sequence to the previous moment, and the difference of historical data of the first sequence is calculated according to the historical data of each moment in the initial time window and the smoothed value obtained by the exponential smoothing algorithm; The time length of the initial time window is adjusted according to the difference of historical data to obtain the optimized time window corresponding to the first sequence; the historical weighted mean of the first sequence is calculated based on the historical data in the optimized time window; The conditional variance of each moment in the optimized time window is calculated by using the GARCH algorithm using the historical weighted mean; the conditional variance of future moments is calculated by sliding the optimized time window backward from the current moment; the future sequence is formed by taking a preset number of conditional variances from the first future moment after the current moment as the starting point; Calculate the conditional variance at each moment in the optimized time window and the correlation of the future sequence; calculate the degree of abnormality of the first sequence based on the correlation, the conditional variance at the current moment and the next moment; similarly, obtain the degree of abnormality of the second and third sequences; Whether the ventilation is abnormal is determined based on the degree of abnormality of the first, second and third sequences. If not, the carbon dioxide concentration in the middle of the ventilation duct outside the exhaust port and inside the exhaust port is used to determine whether the ventilation is abnormal.

2. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The wind speed data, wind pressure data and carbon dioxide concentration data are collected and preprocessed to obtain a first sequence, a second sequence and a third sequence respectively, including: Preprocessing includes filling in missing values, removing outliers, denoising and normalization. The normalized wind speed data, wind pressure data and carbon dioxide concentration data are respectively organized into three different sequences.

3. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The calculation formula of the historical data difference of the first sequence is: ， in, represents the difference of historical data of the first sequence; θ and ϑ represent weight coefficients, which are equal; N represents the number of moments in the initial time window; Norm represents the normalization operation; Represents the historical data at the i-th moment in the initial time window; represents the smoothed value obtained by the exponential smoothing algorithm at the i-th moment in the initial time window; δ represents the sensitivity parameter; λ and μ represent the first scaling factor and the second scaling factor; e represents the natural constant; σ represents the third scaling factor; Indicates the wind speed data at the current moment.

4. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The step of adjusting the time length of the initial time window according to the difference of historical data to obtain the optimized time window corresponding to the first sequence includes: Set a difference weight coefficient. If the difference of historical data is greater than or equal to the threshold, the difference weight coefficient takes the first preset value. If the difference of historical data is less than the threshold, the difference weight coefficient takes the second preset value. Multiply the difference weight coefficient by the difference of historical data, add the multiplication result to the third preset value and take the inverse to obtain the window time length adjustment coefficient. Multiply the window time length adjustment coefficient by the time length of the initial time window and round it to obtain the time length of the optimized time window.

5. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The calculating the historical weighted mean of the first sequence based on the historical data in the optimization time window includes: Subtract the current moment from the moment of a historical data in the optimized time window to obtain the difference result; set the time distance scaling parameter, and use an exponential function with a natural constant as the base to normalize the value obtained by multiplying the time distance scaling parameter and the difference result to obtain the average weight of the historical data; based on each historical data in the optimized time window and the corresponding average weight, perform weighted averaging to obtain the historical weighted mean of the first sequence.

6. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The method of using the historical weighted mean to calculate the conditional variance at each moment in the optimization time window through the GARCH algorithm includes: Obtain the calculation formula for the conditional variance in the GARCH algorithm; calculate the residuals at each moment in the optimization time window, where the residual at each moment is the difference between the data value at the previous moment and the historical weighted mean; obtain the conditional variance at a moment based on the calculation formula for the conditional variance and the residual and conditional variance at the previous moment.

7. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The method of calculating the conditional variance at future moments by sliding the optimized time window backward from the current moment includes: The residual and conditional variance at the current moment and the calculation formula of the conditional variance are used to calculate the conditional variance of the first future moment from the current moment; when calculating the conditional variance of the second future moment, the optimization time window slides backward by a preset step from the current moment, and the mean of the residuals at other moments in the optimization time window except the first future moment is used as the residual in the calculation formula of the conditional variance to calculate the conditional variance of the second future moment; when calculating the conditional variance of the third future moment, the optimization time window slides backward by two preset steps from the current moment, and the mean of the residuals at other moments in the optimization time window except the first and second future moments is used as the residual in the calculation formula of the conditional variance to calculate the conditional variance of the third future moment, and so on, the conditional variance of each future moment is calculated.

8. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The calculating the abnormality degree of the first sequence according to the correlation and the conditional variance at the current moment and the next moment comprises: A first weight corresponding to the conditional variance at the current moment and a second weight corresponding to the conditional variance at the first future moment are set; the conditional variances at the current moment and the first future moment are weightedly summed using the first weight and the second weight to obtain a summation result; an exponential function with a natural constant as the base is used to negatively map the summation result to obtain a mapping result; the difference between the third preset value and the mapping result is the abnormality degree of the first sequence; wherein, if the correlation is greater than zero, the values ​​of the first weight and the second weight are the fourth preset value and the fifth preset value respectively, and if the correlation is less than or equal to zero, the values ​​of the first weight and the second weight are the fifth preset value and the fourth preset value respectively, and the fourth preset value is greater than the fifth preset value.

9. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The step of judging whether ventilation is abnormal according to the abnormality levels of the first, second and third sequences includes: When the abnormality levels of the first, second and third sequences are all less than the reference threshold, the ventilation is normal; when one of the abnormality levels of the first, second and third sequences is greater than or equal to the reference threshold, the ventilation is abnormal.

10. A ventilation data detection and analysis method for coal mine production according to claim 1, characterized in that: The method of judging whether ventilation is abnormal by using the carbon dioxide concentration outside the exhaust port and in the middle of the ventilation duct inside the exhaust port includes: The sigmoid function is used to map the difference in carbon dioxide concentration between the outside of the exhaust port and the middle of the ventilation duct inside the exhaust port to obtain the concentration difference; if the concentration difference is greater than the concentration difference threshold, the ventilation is normal; if the concentration difference is less than or equal to the concentration difference threshold, the ventilation is abnormal.

Images