A hierarchical early warning control method of a mine automatic fire extinguishing system

Abstract

This invention provides a graded early warning control method for an automatic fire extinguishing system in mines, belonging to the field of graded early warning control technology. The method includes: acquiring measured value sequences from multiple sensors at various monitoring points within a mine monitoring area and predicted value sequences output by a baseline prediction model; calculating the difference between the two to generate a temporal deviation sequence for each monitoring point; analyzing the physical driving factors of deviation changes based on the spatial and temporal propagation characteristics of the temporal deviation sequences at each monitoring point, including the spatial propagation path of the deviation, multi-parameter physical consistency, contribution of environmental factors, and reversibility characteristics of the deviation; dynamically calculating a fire risk emergency index based on the comprehensive analysis results of the physical driving factors; and executing graded early warning control according to the fire risk emergency index. This invention solves the problems of insufficient accuracy in early anomaly identification and untimely early warning response in fire risk early warning in mine monitoring areas.

Description

Technical Field

[0001] This invention relates to the field of graded early warning and control technology, and specifically to a graded early warning and control method for an automatic fire extinguishing system in a mine. Background Technology

[0002] Currently, mining monitoring areas typically deploy multiple sensors to collect environmental parameters and upload the data to a monitoring platform for analysis to determine the presence of fire risks. Existing methods generally rely on single parameter threshold comparisons or independent data prediction models, triggering alarms when parameters exceed limits or predicted values ​​deviate significantly.

[0003] The above methods have certain limitations. Comparing threshold values ​​for a single parameter is insufficient to distinguish between normal environmental fluctuations and early abnormal changes, which can easily lead to false alarms. Independent prediction models do not fully utilize the spatial and physical relationships between multiple monitoring points and sensors, making it difficult to accurately identify anomalies in their early stages. Furthermore, existing methods lack analysis of anomaly propagation paths, the impact of environmental factors, and the reversibility of anomalies, resulting in insufficient early warning evidence and often delayed responses.

[0004] Therefore, existing methods for early warning of fire risks in mining monitoring areas are insufficient in terms of accuracy in early anomaly identification and timeliness of early warning response. Summary of the Invention

[0005] This invention provides a graded early warning control method for automatic fire extinguishing systems in mines, which addresses the problems of insufficient accuracy in early anomaly identification and untimely early warning response in fire risk early warning of mine monitoring areas.

[0006] In view of the above problems, the present invention provides a graded early warning control method for an automatic fire extinguishing system in mines, comprising: The measured value sequences of multiple sensors at multiple monitoring points within the mining monitoring area are obtained, along with the predicted value sequences output by the baseline prediction model. The difference between the two is calculated to generate the time-series deviation sequence for each monitoring point. Based on the spatial and temporal propagation characteristics of the time-series deviation sequences at each monitoring point, the physical driving factors of the deviation changes are analyzed. The physical driving factors include the spatial propagation path of the deviation, the physical consistency of multiple parameters, the contribution of environmental factors, and the reversibility characteristics of the deviation. Based on the comprehensive analysis results of physical driving factors, the fire risk emergency index is dynamically calculated, and graded early warning control is implemented according to the fire risk emergency index.

[0007] One or more technical solutions provided in this invention have at least the following technical effects or advantages: The technical solution of this invention first obtains the measured value sequences of multiple sensors at multiple monitoring points within a mining monitoring area and the predicted value sequences output by the baseline prediction model, calculates the difference between the two, and generates a time-series deviation sequence for each monitoring point. By converting the raw monitoring data into a time series deviating from the normal baseline state, a quantitative basis is provided for subsequent analysis.

[0008] Furthermore, based on the propagation characteristics of the time-series deviation sequence in space and time, the spatial propagation path, multi-parameter physical consistency, contribution of environmental factors, and reversibility characteristics of the deviation are analyzed. The spatial propagation credibility, physical consistency credibility, first correction coefficient, and second correction coefficient are obtained. The causes of deviation are quantitatively described from four dimensions: spatial propagation law, multi-parameter physical response, influence of environmental factors, and reversibility of deviation.

[0009] Finally, the spatial propagation credibility and physical consistency credibility are weighted and summed to obtain the first comprehensive value. The first correction coefficient and the second correction coefficient are weighted and summed to obtain the environmental correction value. The fire risk emergency index is obtained by subtracting the environmental correction value from the first comprehensive value. The graded early warning control is then implemented based on the fire risk emergency index.

[0010] In summary, the technical solution of this invention effectively distinguishes between early fire characteristics and deviations caused by environmental factors through the analysis of the spatiotemporal propagation physical mechanism of deviations from multiple monitoring points and multiple sensors, thereby improving the accuracy of early anomaly identification and reducing the false alarm rate. Through the dynamic calculation of the fire risk emergency index and the hierarchical early warning control, a closed-loop response is achieved from the quantification of deviation causes to the determination of early warning levels and the execution of on-site control actions, thereby improving the timeliness of early warning response. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a graded early warning control method for an automatic fire extinguishing system in a mine, provided by the present invention.

[0012] Figure 2 This is a schematic diagram of the analysis process of physical driving factors in the graded early warning control method of an automatic fire extinguishing system for mines provided by the present invention.

[0013] Figure 3 This is a logical diagram of the graded early warning control in the graded early warning control method of the automatic fire extinguishing system for mines provided by the present invention. Detailed Implementation

[0014] This invention provides a graded early warning and control method for automatic fire extinguishing systems in mines, which solves the problems of insufficient accuracy in early anomaly identification and untimely early warning response in fire risk early warning in mine monitoring areas.

[0015] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0016] Examples, such as Figure 1 As shown, this invention provides a graded early warning control method for an automatic fire extinguishing system in mines, the method comprising: S100: Obtain the measured value sequence of multiple sensors at multiple monitoring points within the mining monitoring area and the predicted value sequence output by the baseline prediction model, calculate the difference between the two, and generate the time series deviation sequence of each monitoring point.

[0017] In this step, multiple monitoring points are deployed within the mining monitoring area, and each monitoring point is equipped with multiple types of sensors, including temperature sensors, carbon monoxide sensors, smoke concentration sensors, and oxygen concentration sensors.

[0018] The deployment of monitoring points is based on the spatial range of the monitoring area, the direction of the roadway, the ventilation flow direction, and the effective monitoring radius of the sensors. The effective monitoring radius of temperature sensors is generally 3 to 5 meters, carbon monoxide sensors 5 to 8 meters, smoke concentration sensors 5 to 10 meters, and oxygen concentration sensors 3 to 5 meters. The spacing between adjacent monitoring points is determined based on the effective monitoring radii of the various sensors and the early-stage fire characteristics and their diffusion rate; for example, it can be set to 40 meters.

[0019] For example, in a tunnel 200 meters long and 5 meters wide, six monitoring points are set longitudinally at distances of 0 meters, 40 meters, 80 meters, 120 meters, 160 meters, and 200 meters from the tunnel entrance. When the early characteristics of a fire spread rapidly, the spacing between monitoring points can be shortened to increase the monitoring density.

[0020] The sampling period is determined based on the sensor response time and the rate of change of early fire characteristics. The response time for temperature sensors is typically 1 to 5 seconds, for carbon monoxide sensors 5 to 15 seconds, for smoke concentration sensors 5 to 20 seconds, and for oxygen concentration sensors 1 to 5 seconds. To ensure the capture of continuous changes in the parameters, the sampling period is set to 1 to 5 seconds.

[0021] For example, the sampling period is set to 1 second, and continuous sampling is performed for 60 seconds. Each type of sensor at each monitoring point obtains 60 measured values ​​within this time period.

[0022] In this step, the baseline prediction model is a pre-established time-series prediction model used to predict the sensor values ​​at the current moment based on historical measured values. It can be implemented using an autoregressive integral moving average model, a long short-term memory network, or a gated recurrent unit network.

[0023] Taking a Long Short-Term Memory (LSTM) network as an example, the specific model structure is as follows: the input layer receives a time series of length 120, with a feature dimension of 1 for each time step; the hidden layer is a single-layer LSM network with 64 hidden units, using tanh as the activation function and sigmoid as the recurrent activation function; after the hidden layer, a fully connected layer is connected, with an output dimension of 1 and a linear activation function; the output value is the sensor prediction value at the current time step. During model training, mean squared error is used as the loss function, Adam is used as the optimizer, the learning rate is set to 0.001, the batch size is 32, and the number of training epochs is 100.

[0024] The training data was acquired as follows: Under normal operating conditions in the mining monitoring area, historical sensor data were continuously collected for at least 7 days, with a sampling period consistent with the real-time monitoring phase, both being 1 second. After collection, data from periods of sensor failure, maintenance, and communication interruption were removed, and missing values ​​were filled in using linear interpolation. The cleaned data was then arranged in chronological order, with the first 80% used as the training set and the last 20% as the validation set.

[0025] Before training, the sensor measurements are normalized to their maximum and minimum values, mapping the data to the interval between 0 and 1. The normalization parameters are determined by the training set and applied to the validation set. After training, the predicted values ​​are restored to their original dimensions through inverse normalization. The model input sequence is constructed as follows: for each sampling time t in both the training and validation sets, the normalized measured values ​​of 120 consecutive sampling points from t-119 to t are taken as the input sequence, and the normalized measured value at time t+1 is used as the prediction target. When t is less than 120 or t+1 exceeds the data range, no input sequence is constructed for that sampling time, and it is not included in the training. During training, each training epoch iterates through all constructible input sequences in the training set, with each batch containing 32 input sequences.

[0026] At each sampling time, the measured value of each sensor is subtracted from the corresponding baseline prediction model output value to obtain the deviation value of that sensor at that time. A positive deviation value indicates that the measured value is higher than the predicted value, and a negative deviation value indicates that the measured value is lower than the predicted value. The absolute value of the deviation value reflects the degree of difference between the current measured state and the normal baseline state.

[0027] Finally, the deviation values ​​at each sampling time are arranged in chronological order to form a time-series deviation sequence for each type of sensor. Each monitoring point has multiple types of sensors, therefore each monitoring point corresponds to multiple time-series deviation sequences.

[0028] For example, the deviation values ​​of a temperature sensor at a certain monitoring point for the first 10 sampling times within a continuous 60 seconds are 0.1℃, 0.3℃, 0.2℃, -0.1℃, 0.4℃, 0.6℃, 0.8℃, 1.1℃, 1.5℃, and 2.0℃. Arranging this set of data in chronological order constitutes the first 10 deviation values ​​of the temperature sensor's time-series deviation sequence at that monitoring point. Similarly, the deviation values ​​of a carbon monoxide sensor at the same monitoring point for the same sampling time are 0.2ppm, 0.3ppm, 0.1ppm, 0.0ppm, 0.4ppm, 0.7ppm, 1.0ppm, 1.4ppm, 1.8ppm, and 2.3ppm. This set of data constitutes the first 10 deviation values ​​of the carbon monoxide sensor's time-series deviation sequence at that monitoring point. The deviation value of a smoke concentration sensor for the first 10 sampling times within the same time period is 0.01 mg / m³. 3 0.02 mg / m 3 0.01 mg / m 3 0.03 mg / m 3 0.05mg / m 3 0.07 mg / m 3 0.09 mg / m 3 0.12 mg / m 3 0.15mg / m 3 0.18 mg / m 3 The deviation values ​​of the oxygen concentration sensor for the first 10 sampling times within this time period were -0.01%, -0.02%, -0.01%, -0.03%, -0.05%, -0.07%, -0.09%, -0.12%, -0.15%, and -0.18%. Corresponding time-series deviation sequences were generated for each type of sensor at each monitoring point in the above manner for subsequent analysis.

[0029] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0030] In summary, this step transforms the raw monitoring data into a time series that reflects the degree to which the parameters at each monitoring point deviate from the normal baseline state, providing a quantitative basis for subsequent analysis.

[0031] S200: Based on the spatial and temporal propagation characteristics of the time-series deviation sequences at each monitoring point, analyze the physical driving factors of the deviation changes. The physical driving factors include the spatial propagation path of the deviation, the physical consistency of multiple parameters, the contribution of environmental factors, and the reversibility characteristics of the deviation.

[0032] like Figure 2As shown, this step analyzes the spatial propagation path of the deviation, the physical consistency of multiple parameters, the contribution of environmental factors, and the reversibility characteristics of the deviation, and transforms the time-series deviation sequence of each monitoring point into spatial propagation confidence, physical consistency confidence, first correction coefficient, and second correction coefficient.

[0033] This step analyzes the deviated spatial propagation path, including: Obtain the time-series deviation sequence of multiple monitoring points deployed along the wind flow direction within the monitoring area, and determine the deviation trigger time of each monitoring point. The deviation trigger time is the moment when the time-series deviation sequence first exceeds the preset deviation threshold. Based on the deviation trigger time and spatial location of each monitoring point, the propagation velocity vector of the deviation is calculated, and the direction of deviation propagation is determined based on the propagation velocity vector; Obtain the airflow direction during the current monitoring period, calculate the angle between the deviation from the propagation direction and the airflow direction, and determine the propagation direction consistency coefficient based on the angle. The propagation direction consistency coefficient is negatively correlated with the angle. The reliability of spatial propagation is determined based on the consistency coefficient of propagation direction and the deviation from propagation speed. The reliability of spatial propagation is positively correlated with the consistency coefficient of propagation direction.

[0034] Specifically, the time-series deviation sequences of multiple monitoring points deployed along the wind direction within the monitoring area are first obtained, and the deviation trigger time for each monitoring point is determined. The deviation trigger time is the moment when the time-series deviation sequence first exceeds a preset deviation threshold.

[0035] The preset deviation threshold can be set to three times the standard deviation of the upper limit of the normal fluctuation range of the sensors at each monitoring point. The specific calculation method is as follows: under normal operating conditions in the monitoring area, historical data of each type of sensor at each monitoring point is collected for a preset duration, the mean μ and standard deviation σ of the deviation value sequence are calculated, and the preset deviation threshold is set to μ+3σ.

[0036] Specifically, the 7-day historical data used to calculate the deviation value sequence only includes normal operating conditions. Data from abnormal operating conditions such as equipment start-up and shutdown, maintenance, and ventilation adjustment are removed during the cleaning phase. Abnormal operating conditions are determined as follows: periods when the equipment's start-up or shutdown status is in the process of starting or stopping; periods when the equipment's operating power exceeds the average normal operating power by ±5%; and periods when the wind speed exceeds the average normal operating wind speed by ±0.1 m / s.

[0037] For example, if the standard deviation of the temperature sensor's deviation value sequence under normal operating conditions is 0.15℃, then the temperature sensor deviation threshold is set to 3 × 0.15℃ = 0.45℃, which can be further rounded to 0.5℃. If the standard deviation of the carbon monoxide sensor's deviation value under normal operating conditions is 0.08ppm, then the deviation threshold is set to 3 × 0.08ppm = 0.24ppm, which can be further rounded to 0.25ppm. If the standard deviation of the smoke concentration sensor's deviation value under normal operating conditions is 0.005mg / m³, then the deviation threshold is set to 3 × 0.08ppm = 0.24ppm, which can be further rounded to 0.25ppm. 3 The deviation threshold is then set to 3 × 0.005 mg / m³. 3 =0.015mg / m 3 This can be further rounded to 0.02 mg / m³. 3 The standard deviation of the oxygen concentration sensor under normal operating conditions is 0.015%, so the deviation threshold is set to 3 × 0.015% = 0.045%, which can be further rounded to 0.05%. The oxygen concentration deviation trigger condition is that the deviation value is lower than -0.05%, that is, it is determined to be triggered when the deviation value is first less than -0.05%.

[0038] Setting the threshold using three times the standard deviation allows for a reasonable margin within normal fluctuations while ensuring a high probability that deviations caused by early fire characteristics will exceed the threshold. The standard deviation values ​​mentioned above are example values; in actual applications, they should be calculated and determined based on historical deviation data collected under normal operating conditions in each monitoring area.

[0039] For example, monitoring points A, B, C, and D are set sequentially along the airflow direction in the tunnel, with an adjacent spacing of 40 meters. The times when the temperature sensor timing deviation sequence of each monitoring point first exceeds 0.5℃ are: 18 seconds for monitoring point A, 27 seconds for monitoring point B, 36 seconds for monitoring point C, and 45 seconds for monitoring point D. Therefore, the temperature deviation trigger times for each monitoring point are 18 seconds, 27 seconds, 36 seconds, and 45 seconds, respectively.

[0040] Secondly, based on the deviation trigger time and spatial location of each monitoring point, the deviation propagation velocity vector is calculated. The deviation propagation direction is determined based on the propagation velocity vector. The propagation velocity between adjacent monitoring points is obtained by dividing the distance between adjacent monitoring points by the difference in deviation trigger time. The propagation velocities between adjacent monitoring points are synthesized along the monitoring point layout direction to obtain the deviation propagation velocity vector. When the magnitude and direction of the propagation velocity between adjacent monitoring points are consistent, the magnitude of the deviation propagation velocity vector is the propagation velocity, and the direction is along the monitoring point layout direction.

[0041] If there are differences in propagation speed between adjacent monitoring points, the overall propagation speed vector is calculated by time-weighted average of the speed vectors of each segment. The time-weighted average calculation method is: Overall propagation speed vector = Sum of the products of the propagation speed vectors between adjacent monitoring points and the corresponding deviation from the trigger time / Corresponding deviation from the trigger time.

[0042] For example, if monitoring points A and B are 40 meters apart with a trigger time difference of 9 seconds, the propagation speed from A to B is approximately 40 / 9 ≈ 4.44 m / s. Similarly, if B and C are 40 meters apart with a trigger time difference of 9 seconds, the propagation speed from B to C is also approximately 40 / 9 ≈ 4.44 m / s. And if C and D are 40 meters apart with a trigger time difference of 9 seconds, the propagation speed from C to D is also approximately 40 / 9 ≈ 4.44 m / s. The magnitude and direction of the propagation speed between adjacent monitoring points are consistent, with a deviation of 4.44 m / s from the propagation speed vector, and the direction is along the monitoring point layout from A to D.

[0043] Next, the wind direction during the current monitoring period is obtained, and the angle between the deviation from the propagation direction and the wind direction is calculated. The propagation direction consistency coefficient is determined based on this angle, and the consistency coefficient is negatively correlated with the angle. The wind direction is obtained from wind speed and direction sensors deployed within the monitoring area.

[0044] Among them, the wind speed and direction sensor is installed in the middle of the monitoring area at a height of 1.5 meters above the roadway floor. The sampling period is the same as that of other sensors, which is 1 second. The wind speed measurement accuracy is ±0.1m / s and the wind direction measurement accuracy is ±1°.

[0045] The wind direction during the current monitoring period is taken as the vector average of the wind direction values ​​output by the wind speed and direction sensors during that period. The vector average is calculated as follows: the wind direction angle is taken as 0° with the longitudinal centerline of the roadway pointing towards the exit of the monitoring area, and clockwise is positive. The wind direction angle θi of each sampling point is converted into a unit vector, with the x-component being cos(θi) and the y-component being sin(θi). The x-component and y-component of the unit vector are averaged separately, and then the angle is calculated from the average components using the formula θ = atan2(y0, x0), where y0 is the average value of the y-component, x0 is the average value of the x-component, and atan2 is the arctangent function in the four quadrants. The average wind direction is then obtained.

[0046] Specifically, the formula for calculating the propagation direction consistency coefficient is C1 = cosθ, where θ is the angle between the deviation from the propagation direction and the airflow direction. The propagation direction consistency coefficient ranges from -1 to 1, with a larger value indicating a more consistent deviation from the propagation direction and the airflow direction.

[0047] For example, if the wind speed and direction sensor measures the airflow direction along the roadway from A to D during the current monitoring period, which is consistent with the deviation from the propagation direction, and the angle between the two is 0 degrees, then C1 = cos0° = 1. If the angle is 90 degrees, then C1 = cos90° = 0, indicating that the deviation from the propagation direction is perpendicular to the airflow direction. If the angle is 180 degrees, then C1 = cos180° = -1, indicating that the deviation from the propagation direction is completely opposite to the airflow direction.

[0048] Finally, the spatial propagation reliability is determined based on the propagation direction consistency coefficient and the deviation propagation velocity. The spatial propagation reliability is positively correlated with the propagation direction consistency coefficient. The formula for calculating spatial propagation reliability is R1 = sigmoid(v × C1), where v is the normalized deviation propagation velocity and C1 is the propagation direction consistency coefficient. The deviation propagation velocity is normalized as v = v1 / v0, where v1 is the magnitude of the deviation propagation velocity vector and v0 is the reference velocity. The sigmoid function output ranges from 0 to 1; the closer R1 is to 1, the higher the spatial propagation reliability.

[0049] The reference velocity is set based on the fact that the normal ventilation airflow velocity in mine roadways is generally in the range of 1 m / s to 4 m / s. In the early stage of a fire, the diffusion velocity of hot smoke along the airflow direction is usually slightly higher than the normal airflow velocity. 5 m / s is taken as the normalized reference value so that the normalized value corresponding to the normal propagation velocity is distributed between 0.2 and 1.0.

[0050] Specifically, the reference speed v0 is determined as follows: the wind speed data recorded by the wind speed and direction sensors for 7 consecutive days under normal operating conditions in the monitoring area is statistically analyzed, and the maximum value of the daily average wind speed is taken. If the maximum value is less than 5 m / s, then v0 is taken as 5 m / s; if the maximum value is greater than or equal to 5 m / s, then v0 is taken as the value after rounding up the maximum value.

[0051] For example, if the deviation from the propagation speed is v1=4.44m / s, v=4.44 / 5=0.888, C1=1, then R1=sigmoid(0.888×1)≈0.708.

[0052] Among them, when the deviation from the propagation direction is consistent with the airflow direction and the propagation speed is close to the reference speed, the spatial propagation reliability is relatively high; when the deviation from the propagation direction is perpendicular to or opposite to the airflow direction, it may be a sensor malfunction or local equipment interference, and the spatial propagation reliability is relatively low.

[0053] This step analyzes multi-parameter physical consistency, including: Obtain temperature deviation sequences, gas concentration deviation sequences, and smoke concentration deviation sequences; Determine the trigger time for temperature deviation, and based on the trigger time and the preset time delay window, obtain the trigger status of gas concentration deviation and smoke concentration deviation within the time delay window; Calculate the first deviation between the actual trigger delay and the theoretical trigger delay of the gas concentration deviation, and calculate the second deviation between the actual trigger delay and the theoretical trigger delay of the smoke concentration deviation; The delay consistency coefficient is determined based on the first deviation and the second deviation. The delay consistency coefficient is negatively correlated with both the first deviation and the second deviation. Physical consistency reliability is determined based on the delay consistency coefficient, and physical consistency reliability is positively correlated with the delay consistency coefficient.

[0054] Specifically, the temperature deviation sequence, gas concentration deviation sequence, and smoke concentration deviation sequence are first obtained. The temperature deviation sequence is the time-series deviation sequence of the temperature sensor, the gas concentration deviation sequence is the time-series deviation sequence of the carbon monoxide sensor, and the smoke concentration deviation sequence is the time-series deviation sequence of the smoke concentration sensor. These sequences are derived from the deviation value records of the same monitoring point within the same time period.

[0055] Secondly, determine the trigger time of temperature deviation. Based on the trigger time and the preset time delay window, obtain the trigger status of gas concentration deviation and smoke concentration deviation within the time delay window. The temperature deviation trigger time is the moment when the temperature deviation sequence first exceeds the temperature deviation threshold.

[0056] Oxygen concentration deviations are not included in the multi-parameter physical consistency analysis because, in the early stages of a fire, the decrease in oxygen concentration is usually small and is greatly affected by ventilation conditions. The correlation between the direction of the deviation and the fire development is not as stable as that between temperature, carbon monoxide concentration, and smoke concentration. Therefore, the multi-parameter physical consistency analysis only uses three parameters: temperature, carbon monoxide concentration, and smoke concentration, in order to improve the reliability of the analysis results.

[0057] The time delay window is set based on the physical response time difference between changes in temperature and changes in gas and smoke concentrations in the early stages of a fire. Early-stage temperature increases typically precede increases in carbon monoxide and smoke concentrations; after temperature triggering, carbon monoxide and smoke concentrations deviate sequentially. The time delay window is set as a range of time following the temperature triggering moment, determined based on sensor response characteristics and the early fire spread rate.

[0058] Specifically, the delay window can be set to 60 seconds. This setting is based on the following: temperature sensor response time is generally 1 to 5 seconds, carbon monoxide sensor response time is generally 5 to 15 seconds, and smoke concentration sensor response time is generally 5 to 20 seconds. In the early stages of a fire, the time interval from when the temperature first exceeds the threshold to when carbon monoxide and smoke concentrations show detectable deviations is generally between 30 and 60 seconds, based on publicly available mining fire experimental data and engineering experience. The upper limit of this range, 60 seconds, is taken as the delay window to ensure that the triggering time of gas and smoke concentrations is covered in most early-stage fire scenarios.

[0059] The above experimental data on mine fires are derived from a publicly published experimental report on the early parameter response characteristics of mine roadway fires. The experimental conditions were a roadway cross-sectional area of ​​5m². 2 Up to 12m 2 The ventilation speed is 0.5 m / s to 4 m / s, and the fire source is a simulated fire source in the early stage of spontaneous combustion of coal. The experiment is repeated no less than 5 times.

[0060] For example, if the temperature deviation sequence at a certain monitoring point first exceeds the temperature deviation threshold of 0.5℃ at the 18th second of the monitoring period, the temperature deviation trigger time is the 18th second. The time delay window is set to 60 seconds after the temperature trigger time, that is, from the 18th second to the 78th second.

[0061] Next, calculate the first deviation between the actual trigger delay and the theoretical trigger delay for gas concentration deviation, and calculate the second deviation between the actual trigger delay and the theoretical trigger delay for smoke concentration deviation. The actual trigger delay is the time difference between the moment when the gas concentration deviation sequence or smoke concentration deviation sequence first exceeds the corresponding deviation threshold within the delay window and the temperature deviation trigger time. Within the delay window, if the carbon monoxide concentration deviation sequence never exceeds the carbon monoxide deviation threshold, then the carbon monoxide concentration deviation is not triggered, and the actual trigger delay is taken as the upper limit of the delay window, 60 seconds. If the smoke concentration deviation sequence never exceeds the smoke concentration deviation threshold, then the smoke concentration deviation is not triggered, and the actual trigger delay is taken as the upper limit of the delay window, 60 seconds. If neither is triggered, the actual trigger delay is 60 seconds for both, and the first and second deviations are both 60 seconds minus the theoretical trigger delay. If only one of them is triggered, the actual trigger delay for the non-triggered one is 60 seconds, and the delay for the triggered one is calculated based on the actual trigger time. The theoretical trigger delay is calibrated based on the physical response time difference of temperature, carbon monoxide concentration, and smoke concentration at the same monitoring point in the early stage of a fire under normal operating conditions.

[0062] The theoretical trigger delay is calibrated as follows: In the monitoring area or similar mine roadway environment, time-series data of temperature, carbon monoxide concentration, and smoke concentration are obtained through controlled early-stage fire simulation experiments. The times when the temperature, carbon monoxide concentration, and smoke concentration first exceed the deviation threshold are recorded respectively. The time difference between the carbon monoxide concentration trigger time and the temperature trigger time, as well as the time difference between the smoke concentration trigger time and the temperature trigger time, are calculated. The experiment is repeated multiple times, and the average of the time differences obtained from multiple experiments is taken as the theoretical trigger delay. If controlled experimental conditions are not available, the statistical average of the response time differences of temperature, carbon monoxide concentration, and smoke concentration in the early stage of a fire from publicly available data on similar mine environments is used as the theoretical trigger delay.

[0063] The identification parameters for similar mining environments include: roadway cross-sectional area, ventilation velocity range, type of ignition source, and sensor placement height. When the roadway cross-sectional area of ​​a certain mining environment is 5m²... 2 Up to 12m 2 When the ventilation velocity is within the range of 0.5 m / s to 4 m / s, the ignition source is a simulated ignition source in the early stage of coal spontaneous combustion, and the sensor placement height is consistent with the real-time monitoring stage, it is identified as a similar mining environment.

[0064] The publicly available data mentioned above are time-series data on early-stage fire temperature, carbon monoxide concentration, and smoke concentration recorded in publicly published experimental reports on the early-stage parameter response characteristics of mine roadway fires. Data sources include publicly published academic papers, industry experimental reports, or standardized experimental databases.

[0065] When there are significant differences in the response time differences of early fire temperature, carbon monoxide concentration, and smoke concentration among different publicly available data sources, the following methods are used for screening: Experimental data whose differences between the roadway cross-sectional area, ventilation velocity range, or fire source type and the real-time monitoring area exceed a preset range are removed; data with fewer than 3 experimental repetitions are removed; statistical analysis is performed on the response time differences in the remaining data, and outliers exceeding the mean ± 2 standard deviations are removed; the arithmetic mean of the remaining data is used as the theoretical trigger delay.

[0066] The theoretical trigger delay for carbon monoxide is obtained by extracting the times when the temperature and carbon monoxide concentration first exceed the deviation threshold from publicly available data, calculating the time difference between the two, and then taking the arithmetic mean of the filtered time differences. Similarly, the theoretical trigger delay for smoke is obtained by extracting the times when the temperature and smoke concentration first exceed the deviation threshold from publicly available data, calculating the time difference between the two, and then taking the arithmetic mean of the filtered time differences.

[0067] For example, the carbon monoxide deviation sequence at this monitoring point first exceeded the carbon monoxide deviation threshold of 0.25 ppm at the 32nd second within the time delay window, with an actual trigger delay of 32-18=14 seconds. The smoke concentration deviation sequence first exceeded the smoke concentration deviation threshold of 0.02 mg / m³ at the 45th second within the time delay window. 3 The actual trigger delay is 45-18=27 seconds. The theoretical trigger delay is obtained from the early controlled fire simulation experiment of this monitoring point during the calibration phase or from publicly available data on similar mining environments. The theoretical trigger delay for carbon monoxide concentration deviating from relative temperature is 12 seconds, and the theoretical trigger delay for smoke concentration deviating from relative temperature is 25 seconds. The first deviation is 14-12=2 seconds, and the second deviation is 27-25=2 seconds.

[0068] Next, the time delay consistency coefficient is determined based on the first deviation and the second deviation. The time delay consistency coefficient is negatively correlated with both the first and second deviations. The smaller the first and second deviations, the closer the actual triggering time of the gas concentration deviation and smoke concentration deviation is to the theoretical triggering time, the more the physical response relationship between multiple parameters conforms to the early characteristics of a fire, and the higher the time delay consistency coefficient.

[0069] Specifically, the formula for calculating the latency consistency coefficient is C2=e (-(|Δt1|+|Δt2|) / T)Δt1 is the first deviation, Δt2 is the second deviation, and T is the normalized time constant. The normalized time constant T is set based on the fact that the normal fluctuation range of temperature and gas / smoke deviations from the trigger delay in the early stage of a fire is generally within 10 seconds. Exceeding this range indicates that the multi-parameter physical response relationship deviates significantly from the characteristics of the early stage of a fire.

[0070] The normalized time constant T can be set to 10 seconds. The method for determining the value is as follows: In the theoretical trigger delay calibration experiment, the deviation between the actual trigger delay and the theoretical trigger delay of each experiment is recorded synchronously. The average and standard deviation of the absolute values ​​of the deviations from multiple experiments are calculated. The sum of the average and standard deviation is rounded up to obtain the value of T. If there is no experimental data, 10 seconds is directly taken as the default value.

[0071] For example, if the first deviation Δt1 is 2 seconds, the second deviation Δt2 is 2 seconds, and T is 10 seconds, then C2 = e (-(2+2) / 10) ≈0.67.

[0072] Finally, the physical consistency reliability is determined based on the delay consistency coefficient, which is positively correlated with the delay consistency coefficient. The physical consistency reliability R² is set as R² = C². The closer the physical consistency reliability is to 1, the better the multi-parameter physical consistency. When both the first and second deviations are 0, C² = 1, and the physical consistency reliability is 1; when the sum of the first and second deviations is much greater than the normalized time constant, the physical consistency reliability approaches 0.

[0073] For example, the physical consistency confidence level R² = C² = 0.67. The physical consistency confidence level is used for subsequent calculations of the fire risk urgency index. The theoretical trigger delay and normalized time constant mentioned above are example values; in practical applications, they should be determined based on historical early fire data or calibration experiments in the monitored area.

[0074] This step analyzes the contribution of environmental factors, including: Acquire the mine ventilation parameter sequence and equipment operating status parameter sequence during the monitoring period. The ventilation parameter sequence shall include at least the wind speed sequence. Based on the ventilation parameter sequence and the equipment operating status parameter sequence, determine the contribution of each parameter to the normal fluctuation range of temperature, gas concentration and smoke concentration; The contribution of environmental factors is determined by the ratio of the contribution amount to the deviation amount of the time series deviation sequence; The first correction coefficient of the fire risk emergency index is determined based on the contribution of environmental factors, and the first correction coefficient is positively correlated with the contribution of environmental factors.

[0075] Specifically, the mine ventilation parameter sequence and equipment operating status parameter sequence for the monitoring period are first acquired. The ventilation parameter sequence includes at least the wind speed sequence. The mine ventilation parameter sequence is obtained from wind speed and direction sensors and wind pressure sensors deployed within the monitoring area. The equipment operating status parameter sequence is obtained from the mine equipment operation monitoring system, including equipment start / stop status, operating power, and operating duration. The sampling period for the wind speed sequence is consistent with that of multiple sensors, which is 1 second. The acquisition period for the equipment operating status parameters is 1 second, synchronized with the sensor sampling period.

[0076] The wind pressure sensor is installed in the central part of the monitoring area, 1.5 meters above the roadway floor, parallel to the wind speed and direction sensors. Its wind pressure range is -100Pa to 100Pa, with a measurement accuracy of ±1Pa and a sampling period of 1 second. Equipment operating status parameters are obtained from equipment directly related to ventilation and thermal environment in the monitoring area within the mine equipment operation monitoring system. These include local ventilation fans, conveyors, lighting equipment, and electric valves within the monitoring area. Equipment operating power data is collected by power transmitters in each equipment control cabinet, with a sampling period of 1 second and a power measurement accuracy of ±1kW.

[0077] For example, during a certain monitoring period, the wind speed and direction sensors recorded wind speed sequences of 1.2 m / s, 1.3 m / s, 1.1 m / s, 1.4 m / s, 1.5 m / s, 1.6 m / s, 1.8 m / s, 2.0 m / s, 2.2 m / s, and 2.5 m / s, all in m / s. During the same period, the equipment operation monitoring system recorded equipment operating power sequences of 50 kW, 50 kW, 55 kW, 60 kW, 65 kW, 70 kW, 75 kW, 80 kW, 85 kW, and 90 kW. The wind pressure sensors recorded wind pressure sequences of -20 Pa, -22 Pa, -25 Pa, -28 Pa, -30 Pa, -35 Pa, -40 Pa, -45 Pa, -50 Pa, and -55 Pa. Negative wind pressure values ​​indicate that the monitored area is under negative pressure ventilation.

[0078] Secondly, based on the ventilation parameter sequence and equipment operating status parameter sequence, the contribution of each parameter to the normal fluctuation range of temperature, gas concentration, and smoke concentration is determined. The contribution is calculated by using the ventilation parameters and equipment operating status parameters as independent variables, and the deviations of temperature, carbon monoxide concentration, and smoke concentration as dependent variables, to establish a linear regression model.

[0079] Specifically, the linear regression model takes the form D1=a1×W+b1×P+c1×E+ε1, D2=a2×W+b2×P+c2×E+ε2, and D3=a3×W+b3×P+c3×E+ε3, where D1, D2, and D3 represent the temperature deviation, carbon monoxide concentration deviation, and smoke concentration deviation, respectively; W is the wind speed; P is the wind pressure; E is the equipment operating power; a1, b1, c1, a2, b2, c2, a3, b3, and c3 are the regression coefficients; and ε1, ε2, and ε3 are the error terms. The regression coefficients are estimated using the least squares method.

[0080] The regression model training samples were constructed as follows: From seven consecutive days of historical data under normal operating conditions, wind speed, wind pressure, equipment operating power, temperature deviation, carbon monoxide concentration deviation, and smoke concentration deviation were aligned according to the sampling time. Each sampling time constituted a sample, resulting in a total of 604,800 samples (7 days × 86,400 seconds / day). After removing samples with missing parameters or those marked as abnormal operating conditions after data cleaning, the remaining samples were used for training. When solving using the least squares method, the independent variables were not standardized; the original dimensional data were used directly to solve for the regression coefficients. The independence of the error term was verified using the Durbin-Watson test. If the test statistic deviated significantly from 2, a lag term was added to the regression model for correction. After the regression coefficients were calculated, the contribution of each parameter to the deviation of a certain type of sensor was calculated by multiplying the parameter value by the corresponding regression coefficient.

[0081] For example, based on 7 consecutive days of historical data under normal operating conditions in the monitoring area, a linear regression model was established for temperature deviation values ​​with respect to wind speed, wind pressure, and equipment operating power. The regression results show that for every 1 m / s increase in wind speed, the temperature deviation value decreases by an average of 0.08℃, i.e., a1 = -0.08℃·s / m; for every 1 Pa decrease in wind pressure, the temperature deviation value decreases by an average of 0.01℃, i.e., b1 = 0.01℃ / Pa. When the wind pressure is negative and continues to decrease during the current monitoring period, the temperature deviation value decreases accordingly; for every 1 kW increase in equipment operating power, the temperature deviation value increases by an average of 0.002℃, i.e., c1 = 0.002℃ / kW. When the wind speed is 2.0 m / s during the current monitoring period, the contribution of wind speed to the temperature deviation value is a1 × W = -0.08 × 2.0 = -0.16℃. When the equipment operating power is 80 kW, the contribution of equipment operating power to the temperature deviation value is c1 × E = 0.002 × 80 = 0.16℃. Adding the contributions of wind speed, wind pressure, and equipment operating power to the temperature deviation, we obtain the total contribution of environmental factors to the temperature deviation as -0.16 + 0.16 + contributions from other items. The contributions of the remaining items are calculated in the same way from the wind pressure contribution b1 × P.

[0082] Next, the contribution of environmental factors is determined based on the ratio of their contribution to the deviation in the time-series deviation sequence. The contribution of environmental factors is defined as the ratio of the absolute value of the total contribution of environmental factors to the absolute value of the deviation in the corresponding time-series deviation sequence. When this ratio is greater than 1, it indicates that the contribution of environmental factors is greater than the actual deviation, and the environmental factors can fully explain the deviation, suggesting a high probability that the deviation was caused by environmental factors. When this ratio is less than 1, it indicates that the environmental factors can only partially explain the deviation, and that factors other than environmental factors also exist in the deviation.

[0083] At the same monitoring point, the contributions of temperature (α1), carbon monoxide (α2), and smog (α3) to the environmental factors are calculated at the current time. The environmental factor contribution α for this monitoring point is the maximum value among the three, i.e., α = max(α1, α2, α3). The basis for taking the maximum value is that when the deviation of any parameter is mainly affected by environmental factors, the environmental factor contribution corresponding to that parameter is relatively high. Taking the maximum value ensures that the overall impact of environmental factors on the deviation is not underestimated, thus allowing for sufficient deduction of the deviation caused by environmental factors in the subsequent first correction coefficient.

[0084] For example, at the current monitoring time, the temperature deviation at a certain monitoring point is 0.8℃, and the total contribution of environmental factors to the temperature deviation is -0.4℃. The contribution of environmental factors to temperature is α1 = |-0.4| / |0.8| = 0.5. At the same monitoring point, the carbon monoxide deviation is 0.6ppm, and the total contribution of environmental factors to the carbon monoxide deviation is -0.36ppm. The contribution of environmental factors to carbon monoxide is α2 = |-0.36| / |0.6| = 0.6. At the same monitoring point, the smoke concentration deviation is 0.04mg / m³. 3 The total contribution of environmental factors to the deviation of smoke concentration was -0.015 mg / m³. 3 The contribution of environmental factors to the smoke pollution is α3 = |-0.015| / |0.04| = 0.375. The environmental factor contribution at this monitoring point is α = max(0.5, 0.6, 0.375) = 0.6. An environmental factor contribution of 0.6 indicates that the maximum proportion of deviations in the three parameters that can be explained by environmental factors is 60%, suggesting a high probability that the deviations are influenced by environmental factors.

[0085] Finally, the first correction coefficient for the fire risk emergency index is determined based on the contribution of environmental factors. This first correction coefficient is positively correlated with the contribution of environmental factors. The formula for calculating the first correction coefficient is K1 = α, where α is the contribution of environmental factors and K1 is the first correction coefficient. When the contribution of environmental factors is 0, K1 = 0, indicating that environmental factors contribute no to the deviation, and the fire risk emergency index is not lowered. When the contribution of environmental factors is 1, K1 = 1, indicating that the deviation is entirely caused by environmental factors, and the correction amount for the fire risk emergency index reaches its maximum. When the contribution of environmental factors is greater than 1, K1 is 1. When the contribution of environmental factors ranges from 0 to 1, the first correction coefficient also ranges from 0 to 1. The greater the contribution of environmental factors, the larger the first correction coefficient, and the greater the downward adjustment of the fire risk emergency index.

[0086] For example, when the temperature deviation is 0.8℃ and the environmental factor contribution is 0.5, the first correction coefficient K1 = 0.5. This correction coefficient is used to deduct the deviation caused by environmental factors when calculating the fire risk emergency index. The linear regression coefficients, contribution values, and environmental factor contribution values ​​mentioned above are all example values. In actual applications, a regression model is established and calculated based on historical data collected under normal operating conditions in the monitoring area.

[0087] This step analyzes deviations from reversibility characteristics, including: When the contribution of environmental factors exceeds the preset contribution threshold, the regression characteristics of the time-series deviation sequence of the ventilation parameter sequence after it has recovered to the normal state are obtained. The regression rate of the time series deviation sequence after the ventilation parameter sequence is recovered is calculated, and the deviation reversibility coefficient is determined based on the regression rate. The deviation reversibility coefficient is positively correlated with the regression rate. The second correction factor for the fire risk emergency index is determined based on the deviation reversibility coefficient, and the second correction factor is positively correlated with the deviation reversibility coefficient.

[0088] Specifically, when the contribution of environmental factors exceeds a preset contribution threshold, the regression characteristics of the time-series deviation sequence after the ventilation parameter sequence recovers to normal are first obtained. The preset contribution threshold, set to 0.5, is used to determine whether the deviation is mainly caused by environmental factors. When the contribution of environmental factors is greater than or equal to 0.5, it indicates that environmental factors account for the majority of the deviation. When the contribution of environmental factors is equal to 0.5, it is treated as exceeding the threshold, and the regression characteristics of the deviation after the ventilation parameters recover are further analyzed. The regression characteristics include whether the deviation value falls back, the magnitude of the fall, and the speed of the fall. The method for determining whether the ventilation parameters have recovered to normal is: the wind speed value returns to within ±0.1 m / s of the average wind speed under normal operating conditions at the monitoring point, and the equipment operating power returns to within ±5% of the average power under normal operating conditions.

[0089] The preset contribution threshold of 0.5 is set based on the following: when the contribution of environmental factors is 0.5, the contribution of environmental factors to the deviation and the actual deviation each account for half, and the credibility of the deviation being dominated by environmental factors has reached a level that can be determined; when it is lower than 0.5, the proportion of non-environmental factors in the deviation is larger, which does not meet the precondition of analyzing the reversibility of the deviation. This value can be adjusted according to the balance requirement of false alarm rate and false negative rate in practical applications.

[0090] The calculation method for the average wind speed and average power under normal operating conditions is as follows: Take the wind speed data and equipment operating power data from 7 consecutive days of historical data under normal operating conditions, after data cleaning to remove the periods of equipment start-up, shutdown, maintenance and ventilation adjustment, and calculate the average wind speed and average power respectively. The calculation method is arithmetic mean.

[0091] For example, the temperature deviation at a certain monitoring point is 0.8℃, with an environmental factor contribution of 0.6, which is greater than the preset contribution threshold of 0.5. Under normal operating conditions, the average wind speed at this monitoring point is 1.2 m / s, and the average equipment operating power is 50 kW. At the 30th second, the wind speed recovers from 2.0 m / s to 1.2 m / s, and the equipment operating power recovers from 80 kW to 50 kW, both meeting the criteria for returning to normal. The time-series deviation sequence of the temperature sensor at this monitoring point is obtained for the 60 seconds following the 30th second. This sequence is 0.8℃, 0.7℃, 0.6℃, 0.5℃, 0.4℃, 0.3℃, 0.2℃, 0.1℃, 0.0℃, and -0.1℃, indicating that the deviation gradually returns to the normal fluctuation range.

[0092] Secondly, the regression rate of the time-series deviation sequence after the ventilation parameter sequence recovers is calculated. The reversibility coefficient of the deviation is determined based on the regression rate, which is positively correlated with the regression rate. The regression rate is calculated as follows: starting from the time the ventilation parameters recover, a linear fit is performed on the absolute value of the time-series deviation sequence. The fitting function is in the form |D(t)|=r×t+b, where D(t) is the deviation value at the t-th sampling time after the ventilation parameters recover, t is the sampling point number starting from the time the ventilation parameters recover, r is the regression rate, and b is the fitting intercept. The least squares method is used for linear fitting. The sample size used for linear fitting is all 60 sampling points within 60 seconds after the ventilation parameters recover. If the deviation value has fallen back to the normal fluctuation range within 60 seconds, all 60 sampling points are still used for fitting without premature truncation. If the 60-second data collection after the environmental factors have recovered is not yet complete, K2 is temporarily set to 0, indicating that the deviation from the reversibility feature does not contribute to the environmental correction when the data is not complete. After the 60-second data collection is completed, the deviation reversibility coefficient C3 and the second correction coefficient K2 are calculated and updated.

[0093] The formula for calculating the deviation reversibility coefficient is C3 = 1 - e (-|r| / r0)Where r is the regression rate and r0 is the normalized regression rate constant. The deviation reversibility characteristic analysis is calculated solely based on the temperature deviation sequence; C3 is determined only by the regression rate of the temperature deviation sequence, and K2 is also determined only by C3. This is because temperature deviations respond most directly to environmental factor recovery in the early stages of a fire, with a stable and dimensionless decline process; carbon monoxide concentration deviations and smoke concentration deviations are significantly affected by diffusion and dilution after ventilation parameters recover, and their physical meaning differs from that of temperature, making it inappropriate to share the same normalized regression rate constant with temperature.

[0094] The setting of r0 is based on the following: the deviation recovery speed caused by environmental factors in the early stage of a fire generally does not exceed 0.05 units / second. Exceeding this speed indicates that the deviation recovery is too rapid, possibly caused by sensor malfunction or abnormal data transmission. Therefore, r0 is set to 0.05 units / second. The data source for r0 is experimental data on the recovery characteristics of environmental parameters in mine roadways. The experimental conditions involve artificially changing the wind speed or equipment power under normal ventilation conditions, restoring the parameters to normal after the deviation stabilizes, recording the deviation recovery process, and calculating the average absolute value of the recovery speed. This average value is taken as r0. If no experimental data is available, the default value of 0.05 units / second is used. When |r| is much greater than r0, C3 approaches 1, indicating high reversibility of the deviation; when |r| is close to 0, C3 approaches 0, indicating irreversible deviation.

[0095] In r0 = 0.05 units / second, the unit / second refers to temperature units per second, i.e., °C / s. The physical meaning of r0 is the normalized baseline value of the rate of temperature deviation decline, with dimensions in °C / s. During calculation, both the regression rate r and r0 adopt the dimensions of the temperature deviation value to ensure dimensional consistency. If the temperature deviation sequence uses other temperature units, r0 is converted accordingly to maintain dimensional matching.

[0096] For example, the regression rate of the temperature deviation sequence after the ventilation parameters recover was obtained through linear fitting. Starting from the moment the ventilation parameters recover, the absolute deviation values ​​of all 60 sampling moments within 60 seconds were used for fitting. The absolute deviation values ​​for the first 10 sampling moments were 0.8℃, 0.7℃, 0.6℃, 0.5℃, 0.4℃, 0.3℃, 0.2℃, 0.1℃, 0.0℃, and 0.1℃, with the absolute deviation values ​​decreasing sequentially to 0.05℃ at subsequent sampling moments. The sampling period was 1 second. Linear fitting of this sequence yielded a regression rate r = -0.088℃ / s and a fitting intercept b = 0.81℃. Taking the absolute value of the regression rate |r| = 0.088℃ / s and r0 = 0.05℃ / s, the deviation reversibility coefficient C3 = 1 - e^(-r / s) was calculated. (-0.088 / 0.05)≈1-0.172=0.828. The deviation reversibility coefficient is 0.828, which means that the deviation falls back quickly after the ventilation parameters are restored. The deviation is mainly caused by reversible environmental factors, and the possibility of early fire characteristics is low.

[0097] Secondly, a second correction coefficient for the fire risk emergency index is determined based on the deviation reversibility coefficient, which is positively correlated with the deviation reversibility coefficient. The formula for calculating the second correction coefficient is K2 = C3, where C3 is the deviation reversibility coefficient and K2 is the second correction coefficient. Both the deviation reversibility coefficient and the second correction coefficient range from 0 to 1. A larger deviation reversibility coefficient indicates a more reversible deviation, a greater likelihood that the deviation is caused by environmental factors, and a larger second correction coefficient, resulting in a greater downward adjustment of the fire risk emergency index. When the deviation reversibility coefficient is 0, K2 = 0, indicating an irreversible deviation, and the fire risk emergency index is not adjusted downwards. When the deviation reversibility coefficient is 1, K2 = 1, indicating a completely reversible deviation, and the fire risk emergency index correction reaches its maximum.

[0098] For example, when the reversibility coefficient C3 is 0.828, the second correction coefficient K2 is also 0.828. This correction coefficient is used to deduct reversibility deviations when calculating the fire risk emergency index. The preset contribution threshold, regression rate value, normalized regression rate constant, and reversibility coefficient value mentioned above are all example values. In actual application, they should be determined based on historical data of the monitoring area and calibration experiments.

[0099] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0100] In summary, this step quantifies the causes of deviation from four dimensions: spatial propagation patterns, multi-parameter physical response relationships, the degree of influence of environmental factors, and the recoverability of deviations. This provides a physical mechanism-level basis for the subsequent dynamic calculation of the fire risk emergency index.

[0101] S300: Based on the comprehensive analysis results of physical driving factors, dynamically calculate the fire risk emergency index and implement graded early warning control according to the fire risk emergency index.

[0102] like Figure 3 As shown, this step obtains a first comprehensive value by weighted summation of spatial propagation credibility and physical consistency credibility, obtains an environmental correction value by weighted summation of the first correction coefficient and the second correction coefficient, obtains a fire risk emergency index by subtracting the environmental correction value from the first comprehensive value, and performs graded early warning control based on the fire risk emergency index.

[0103] In this step, based on the comprehensive analysis results of physical driving factors, the fire risk emergency index is dynamically calculated, including: Obtain spatial propagation credibility, physical consistency credibility, first correction coefficient, and second correction coefficient; The first comprehensive value is obtained by weighted summation of the credibility of spatial propagation and the credibility of physical consistency. The environmental correction value is obtained by weighted summing of the first and second correction factors. The fire risk emergency index is obtained by subtracting the environmental correction value from the first comprehensive value.

[0104] Specifically, the spatial propagation reliability, physical consistency reliability, first correction coefficient, and second correction coefficient are first obtained. The spatial propagation reliability is R1 calculated in the spatial propagation path analysis step; the physical consistency reliability is R2 calculated in the multi-parameter physical consistency analysis step; the first correction coefficient is K1 calculated in the environmental factor contribution analysis step; and the second correction coefficient is K2 calculated in the deviation reversibility characteristic analysis step. These values ​​are dynamically updated each time the fire risk emergency index is calculated based on the latest deviation data and analysis results within the current monitoring period.

[0105] The fire risk emergency index is calculated based on the monitoring area, with a calculation cycle of 1 second, consistent with the sensor sampling cycle. When multiple monitoring points exist within the monitoring area, the spatial propagation reliability R1 is calculated by the spatial propagation path analysis step based on the deviation trigger time and airflow direction of each monitoring point, and is directly used as the spatial propagation reliability at the monitoring area level. The physical consistency reliability is first calculated for each monitoring point separately, and then the maximum value of the physical consistency reliability of each monitoring point is taken as R2. The first correction coefficient K1 is taken as the minimum value among the first correction coefficients corresponding to the environmental factor contribution of each monitoring point. The second correction coefficient K2 is determined as follows: when the environmental factor contribution of any monitoring point is greater than or equal to the preset contribution threshold of 0.5, the maximum value among the second correction coefficients corresponding to the deviation reversibility coefficient of each monitoring point is taken as K2. The basis for taking the maximum value is: the stronger the deviation reversibility, the more likely the deviation is caused by environmental factors, the larger the environmental correction value should be, and the greater the reduction of the fire risk emergency index. Taking the maximum value can ensure that the reversible deviation caused by environmental factors is fully deducted. When the contribution of environmental factors at all monitoring points does not exceed 0.5, the deviation from reversibility feature analysis step is not triggered, and K2 takes the default value of 0, indicating that the deviation from reversibility feature does not contribute to environmental correction.

[0106] For example, in a certain calculation, the spatial propagation confidence level R1 = 0.708, the physical consistency confidence level R2 = 0.67, the first correction coefficient K1 = 0.5, and the second correction coefficient K2 = 0.828.

[0107] Secondly, the spatial propagation credibility and physical consistency credibility are weighted and summed to obtain the first comprehensive value. The formula for calculating the first comprehensive value is S1=w1×R1+w2×R2, where w1 is the spatial propagation credibility weight and w2 is the physical consistency credibility weight.

[0108] The weighting is based on the following criteria: spatial propagation credibility reflects whether the spatial propagation pattern of the deviation is consistent with the early-stage fire diffusion characteristics; physical consistency credibility reflects whether the temporal response relationship of the multi-parameter deviation is consistent with the early-stage fire physical response characteristics. Both contribute equally to fire risk assessment, therefore w1 and w2 are both set to 0.5. The weights satisfy w1 + w2 = 1. The first comprehensive value ranges from 0 to 1; a larger value indicates that the deviation is more consistent with early-stage fire characteristics in both spatial propagation and multi-parameter physical response dimensions.

[0109] Specifically, the weights w1 and w2 are determined as follows: Based on historical early fire data and normal operating condition data in the monitoring area, with the constraint that the false alarm rate should not exceed the preset upper limit of the false alarm rate and the objective of minimizing the missed alarm rate, a grid search is performed on w1 and w2 within the range of 0.1 to 0.9 with a step size of 0.1, selecting the weight combination that satisfies the constraints and has the lowest missed alarm rate; if there is no historical fire data, then w1=0.5 and w2=0.5 are taken as default values. The preset upper limit of the false alarm rate is determined according to the safety management requirements of the mining monitoring area, generally ranging from 5% to 10%.

[0110] For example, the first composite value S1 = 0.5 × 0.708 + 0.5 × 0.67 = 0.689.

[0111] Next, the first correction factor and the second correction factor are weighted and summed to obtain the environmental correction value. The formula for calculating the environmental correction value is S2 = w3 × K1 + w4 × K2, where w3 is the weight of the first correction factor and w4 is the weight of the second correction factor.

[0112] The first correction factor reflects the contribution of environmental factors to the deviation, while the second correction factor reflects the reversibility of the deviation after the environmental factors are restored. Both are used to deduct deviations caused by environmental factors. Both w3 and w4 are set to 0.5, and the weights satisfy w3 + w4 = 1. The environmental correction value ranges from 0 to 1; a larger value indicates a higher degree of deviation caused by environmental factors, and a larger portion needs to be deducted from the fire risk.

[0113] Specifically, the weights w3 and w4 are determined in the same way as w1 and w2. They are determined by grid search based on historical fire data and normal operating condition data, with the false alarm rate not exceeding the preset upper limit of the false alarm rate as a constraint and the minimum missed alarm rate as the objective. If there is no historical fire data, w3=0.5 and w4=0.5 are taken as default values.

[0114] For example, the environmental correction value S2 = 0.5 × 0.5 + 0.5 × 0.828 = 0.25 + 0.414 = 0.664.

[0115] Next, the environmental correction value is subtracted from the first comprehensive value to obtain the fire risk emergency index. The fire risk emergency index is calculated using the formula F = S1 - S2, where S1 is the first comprehensive value and S2 is the environmental correction value. S1 ranges from 0 to 1, S2 ranges from 0 to 1, and the theoretical range of F is -1 to 1. When F is less than 0, F is set to 0, so the final output range of F is 0 to 1. When F is greater than 0, it indicates that the contribution of deviation from the early stage of a fire is greater than the contribution of environmental factors, and the larger the F, the higher the fire risk. When F equals 0, it indicates that the contribution of deviation from the early stage of a fire does not exceed the contribution of environmental factors, and the fire risk is low.

[0116] For example, the fire risk urgency index F = 0.689 - 0.664 = 0.025. F is greater than 0, indicating that the contribution of early fire characteristics to the deviation in the current monitoring period is slightly greater than the contribution of environmental factors, and the fire risk urgency index is 0.025.

[0117] In another scenario, the spatial propagation confidence level R1 = 0.2, the physical consistency confidence level R2 = 0.3, the first correction coefficient K1 = 0.9, and the second correction coefficient K2 = 0.8. The first comprehensive value S1 = 0.5 × 0.2 + 0.5 × 0.3 = 0.1 + 0.15 = 0.25. The environmental correction value S2 = 0.5 × 0.9 + 0.5 × 0.8 = 0.45 + 0.4 = 0.85. The fire risk urgency index F = 0.25 - 0.85 = -0.6. Since F is less than 0, setting F to 0 indicates that the current deviation is mainly caused by environmental factors, and the fire risk is low.

[0118] The weight values ​​and example values ​​mentioned above are for illustrative purposes only. In actual applications, the weights can be adjusted according to the historical fire data of the monitoring area and the false alarm rate requirements.

[0119] Furthermore, the warning level is determined based on the warning threshold range to which the F value belongs. The warning threshold ranges are set as follows: when F is less than 0.2, the warning level is normal; when F is greater than or equal to 0.2 and less than 0.4, the warning level is blue; when F is greater than or equal to 0.4 and less than 0.6, the warning level is yellow; when F is greater than or equal to 0.6 and less than 0.8, the warning level is orange; and when F is greater than or equal to 0.8, the warning level is red.

[0120] The calibration method for the warning threshold intervals is as follows: Based on historical early fire data and normal operating condition data for the monitoring area, calculate the F-values ​​corresponding to the early fire periods and normal operating condition periods. Use the 95th percentile of the F-value during the normal operating condition period as the boundary between normal and blue warnings, and the 5th percentile of the F-value during the early fire periods as the boundary between orange and red warnings. Divide the intervals between these boundary values ​​into two equal segments to obtain the boundary values ​​between blue and yellow, and between yellow and orange. If no historical fire data is available, use 0.2, 0.4, 0.6, and 0.8 as the default boundary values.

[0121] The F-value reflects the degree of contribution to the deviation from the early characteristics of a fire; a higher F-value indicates a greater fire risk. The division of warning threshold ranges is used to convert continuous risk values ​​into discrete warning levels, facilitating different response measures by on-site personnel and management systems according to the level.

[0122] Finally, corresponding control actions are executed based on the warning level. The graded warning control steps are executed synchronously with the fire risk emergency index calculation steps. Each calculation cycle determines the warning level based on the latest F value and executes the corresponding control actions. When the warning level changes, the corresponding control actions for the new level are executed immediately; when the warning level jumps up, there is no confirmation buffer period, and the new level control actions are executed directly.

[0123] Under normal conditions, the system maintains regular monitoring without triggering alarms or control actions.

[0124] Under a blue alert level, the system sends a notification to the monitoring platform, including the alert level, monitoring area number, and current F value, while simultaneously increasing the data recording frequency. Specifically, the sampling period of all sensors within the monitoring area is adjusted from 1 second to 0.5 seconds, and the duration of this increased frequency matches the duration of the blue alert. When the alert level returns to normal, the sampling period reverts to 1 second.

[0125] Under a yellow alert level, the system sends an alarm message to the monitoring platform and an activation signal to the on-site audible and visual alarms. The alarms are installed at both ends of the monitoring area, one at each end, 1.5 meters above the tunnel floor. Upon receiving the activation signal, the alarms emit a flashing yellow light and an intermittent buzzing sound. The flashing frequency is 2Hz, and the buzzing sound pressure level is 75dB. The alarm duration is the same as the yellow alert duration. When the alert level returns to blue or normal, the alarms stop sounding.

[0126] Under an orange alert level, the system sends an alarm message to the monitoring platform and activates the on-site audible and visual alarm. The alarm emits flashing orange lights and a continuous buzzing sound, with a flashing frequency of 4Hz and a buzzing sound pressure level of 85dB. Simultaneously, a pre-start signal is sent to the automatic fire suppression system, which then enters the pre-start state. Specifically, the pre-start state involves: the extinguishing medium release valve remaining closed, the extinguishing medium delivery pipeline completing pressure build-up, the nozzle protective cover opening, the control system entering a standby state, and the monitoring platform displaying a pre-start countdown. The duration of the pre-start state is the same as the duration of the orange alert. When the alert level returns to yellow, blue, or normal, the automatic fire suppression system exits the pre-start state, the nozzle protective cover closes, and the delivery pipeline pressure is released to a safe value.

[0127] Under a red alert, the system sends an alarm message to the monitoring platform and activates the on-site audible and visual alarm. The alarm emits flashing red lights and a continuous buzzing sound, with a flashing frequency of 5Hz and a buzzing sound pressure level of 90dB. Simultaneously, it sends a start signal to the automatic fire suppression system, which releases the extinguishing agent. The extinguishing agent is dry powder, released via nitrogen-driven spraying through pipelines and nozzles. The release area is a 20-meter radius before and after the location of the monitoring point with the highest F-value within the monitoring area. The release quantity is the amount of dry powder required for the design extinguishing concentration corresponding to the volume of the area, and the release duration is 30 seconds. After release, the automatic fire suppression system maintains a depressurized pipeline state until on-site safety is confirmed, at which point it is manually reset.

[0128] For example, within a certain calculation period, the fire risk emergency index F = 0.25. An F value greater than or equal to 0.2 and less than 0.4 indicates a blue alert. The system sends a notification to the monitoring platform: "The current fire risk emergency index for the third lane in the monitoring area is 0.25, blue alert. Please pay attention." Simultaneously, the sampling period of all sensors within this monitoring area is adjusted from 1 second to 0.5 seconds, continuously recording temperature, carbon monoxide concentration, smoke concentration, and oxygen concentration data.

[0129] In another example, F=0.52, where F is greater than or equal to 0.4 and less than 0.6, the warning level is determined to be a yellow warning. The system sends an alarm message to the monitoring platform and activates the audible and visual alarms at both ends of the monitoring area. The audible and visual alarms emit flashing yellow lights and intermittent buzzing sounds, with a flashing frequency of 2Hz and a buzzing sound pressure level of 75dB.

[0130] In another example, F=0.75, where F is greater than or equal to 0.6 and less than 0.8, the warning level is determined to be orange. The system sends an alarm message to the monitoring platform, activates the on-site audible and visual alarm, which emits flashing orange lights and a continuous buzzing sound at a flashing frequency of 4Hz and a sound pressure level of 85dB. Simultaneously, a pre-start signal is sent to the automatic fire extinguishing system, the extinguishing medium release valve remains closed, the nozzle protective cover opens, the delivery pipeline completes pressure build-up, and the monitoring platform displays a pre-start countdown.

[0131] In another example, F=0.85, and since the F value is greater than or equal to 0.8, the warning level is determined to be a red alert. The system sends an alarm message to the monitoring platform, activates the on-site audible and visual alarm, which emits flashing red lights and a continuous buzzing sound at a flashing frequency of 5Hz and a sound pressure level of 90dB. Simultaneously, an activation signal is sent to the automatic fire extinguishing system. The automatic fire extinguishing system uses nitrogen to drive dry powder extinguishing agent, which is then sprayed through the pipeline and nozzles. The release area is a 20-meter radius before and after the location of the monitoring point with the highest F value, and the release duration is 30 seconds. After the release is complete, the automatic fire extinguishing system maintains a depressurized pipeline state, and is manually reset after on-site confirmation of safety.

[0132] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0133] In summary, this step transforms the comprehensive analysis results of physical driving factors into continuous risk values, and then converts them into discrete early warning levels through threshold grading. This achieves a closed-loop response throughout the entire process, from quantifying the causes of deviations to determining early warning levels and executing on-site control actions. While effectively deducting interference from environmental factors, it provides graded and progressive early warning and fire suppression control for deviations caused by early fire characteristics.

[0134] The method provided by this invention further includes: Based on the deviation trigger time and spatial spacing of each monitoring point, the propagation speed of the deviation along the airflow direction is calculated; A preset fire spread rate threshold is obtained. When the spread rate exceeds the preset fire spread rate threshold, the fire is determined to be in a rapid spread stage, and the weight coefficient of spatial spread credibility in the calculation of the first comprehensive value is increased. When the spread rate is lower than the preset fire spread rate threshold, the fire is determined to be in the smoldering stage, and the current monitoring frequency is maintained.

[0135] Specifically, the propagation speed of the deviation along the airflow direction is calculated based on the deviation trigger time and spatial distance of each monitoring point. The deviation trigger time is the moment when the temporal deviation sequence of each monitoring point first exceeds the preset deviation threshold, and the spatial distance is the distance between adjacent monitoring points along the airflow direction.

[0136] Specifically, the propagation velocity is calculated as follows: the spatial distance between adjacent monitoring points is divided by the difference in the deviation from the trigger time to obtain the propagation velocity between those adjacent monitoring points. The propagation velocities between each adjacent monitoring point are then averaged over time, with the weight of the time-weighted average being the difference in the deviation from the trigger time corresponding to each adjacent monitoring point, to calculate the overall propagation velocity along the wind direction.

[0137] For example, monitoring points A, B, C, and D are set sequentially along the airflow direction in the tunnel, with an adjacent spacing of 40 meters. The temperature deviation trigger times for each monitoring point are 18 seconds, 27 seconds, 36 seconds, and 45 seconds, respectively. The propagation speed from adjacent monitoring points A to B is 40 / (27-18) = 40 / 9 ≈ 4.44 m / s, the propagation speed from B to C is 40 / (36-27) = 40 / 9 ≈ 4.44 m / s, and the propagation speed from C to D is 40 / (45-36) = 40 / 9 ≈ 4.44 m / s. The propagation speed is consistent among adjacent monitoring points, and the overall propagation speed is 4.44 m / s.

[0138] Secondly, a preset fire spread rate threshold is obtained. This threshold is determined based on experimental data and engineering experience regarding the early spread rate of fires in mine roadways. The threshold is set according to the critical spread rate at which a fire in a mine roadway transitions from the smoldering stage to the rapid spread stage. Based on mine fire experimental data, this critical spread rate is generally between 3 m / s and 5 m / s; 5 m / s is taken as the default value.

[0139] Finally, when the propagation speed exceeds the preset fire spread speed threshold, the fire is determined to be in a rapid spread stage, and the weight coefficient of spatial propagation credibility in the calculation of the first comprehensive value is increased. The spatial propagation credibility is R1 calculated in the spatial propagation path analysis step. The formula for calculating the first comprehensive value is S1=w1×R1+w2×R2, where w1 is the spatial propagation credibility weight and w2 is the physical consistency credibility weight.

[0140] When the fire is determined to be in a rapid spread phase, w1 is increased to 0.7, and w2 is adjusted accordingly to 0.3, maintaining w1+w2=1. The basis for the weight adjustment is: in the rapid spread phase of the fire, the rapid propagation deviating from the direction of airflow is an important characteristic of fire development, and the reliability of spatial propagation contributes more to the fire risk assessment, therefore w1 is increased.

[0141] When the propagation speed is lower than the preset fire spread speed threshold, the fire is determined to be in the smoldering stage, and the current monitoring frequency is maintained. The current monitoring frequency is 1 second, which is the sensor sampling period. During the smoldering stage, the fire spreads slowly, and the deviation from the propagation characteristics is not significant. The spatial propagation confidence weights remain unchanged at w1=0.5 and w2=0.5, and the monitoring frequency remains at once per second. When the propagation speed equals the preset fire spread speed threshold, the fire is determined to be in the smoldering stage, and the current monitoring frequency is maintained without adjusting the weights.

[0142] For example, the overall propagation speed is 4.44 m / s, which is less than the preset fire spread speed threshold of 5 m / s. The fire is determined to be in the smoldering stage. The current monitoring frequency is maintained at once per second, and the spatial propagation confidence weight w1 is maintained at 0.5.

[0143] In another example, the temperature deviation trigger times at each monitoring point are 15 seconds, 21 seconds, 28 seconds, and 36 seconds, with an adjacent distance of 40 meters. The propagation speed from adjacent monitoring points A to B is 40 / (21-15)=40 / 6≈6.67m / s, the propagation speed from B to C is 40 / (28-21)=40 / 7≈5.71m / s, and the propagation speed from C to D is 40 / (36-28)=40 / 8=5m / s. The time-weighted average calculation method is: overall propagation speed = (6.67×6+5.71×7+5×8) / (6+7+8)=(40.02+39.97+40) / 21≈5.71m / s. The propagation speed exceeds the preset fire spread speed threshold of 5m / s, indicating that the fire is in a rapid spread stage. In the first comprehensive value calculation, w1 is increased from 0.5 to 0.7, and w2 is adjusted from 0.5 to 0.3.

[0144] The preset fire spread rate threshold and weight adjustment values ​​mentioned above are all example values. In actual application, they should be calibrated based on historical fire data of the monitoring area and engineering experience.

[0145] The method provided by this invention further includes: Obtain the time-series deviation sequence of the target monitoring point and the time-series deviation sequence of adjacent monitoring points, and calculate the deviation coordination coefficient between the target monitoring point and adjacent monitoring points. The deviation coordination coefficient is negatively correlated with the time difference between the deviation trigger times of the two monitoring points. When the deviation coordination coefficient is less than the preset coordination threshold and the deviation trigger time of the target monitoring point is later than the deviation trigger time of the adjacent monitoring point, it is determined that the sensor of the target monitoring point is faulty or subject to local interference, and the monitoring weight of the target monitoring point is set to zero. When the deviation coordination coefficient is greater than or equal to the preset coordination threshold and the deviation trigger time of the target monitoring point is earlier than the deviation trigger time of the adjacent monitoring point, the target monitoring point is determined to be a candidate location of the fire source, and the monitoring priority and sampling frequency of the target monitoring point are increased.

[0146] Specifically, the time-series deviation sequence of the target monitoring point and the time-series deviation sequences of adjacent monitoring points are first obtained. The target monitoring point is the monitoring point whose data reliability needs to be assessed, and the adjacent monitoring points are those adjacent to the target monitoring point along the airflow direction. The deviation coordination coefficient is used to measure the temporal proximity of the deviation triggering at the target monitoring point and the deviation triggering at the adjacent monitoring points.

[0147] Secondly, calculate the deviation coordination coefficient between the target monitoring point and its adjacent monitoring points. Obtain the deviation trigger time t1 of the target monitoring point and the deviation trigger time t2 of the adjacent monitoring points, and calculate the absolute value of the time difference between them, Δt1 = |t1 - t2|. The deviation coordination coefficient is calculated using the formula C4 = e (-Δt1 / T1) , where T1 is the cooperative time constant.

[0148] The coordination time constant T1 is set based on the following: the time difference in the propagation of early fire deviation along the airflow direction between adjacent monitoring points generally does not exceed the distance between adjacent monitoring points divided by the minimum spread velocity. When the adjacent distance is 40 meters and the minimum spread velocity is 3 m / s, the time difference is approximately 13.3 seconds. Considering the difference in sensor response time, T1 is set to 15 seconds. The deviation coordination coefficient ranges from 0 to 1. The smaller the time difference, the closer the deviation coordination coefficient is to 1, indicating that the deviation between the target monitoring point and adjacent monitoring points is more synchronized in time, and the higher the probability that the deviation is caused by the same physical process.

[0149] When a target monitoring point has multiple adjacent monitoring points, the deviation coordination coefficient between the target monitoring point and each adjacent monitoring point is calculated separately, and the maximum value among the calculated results is taken as the deviation coordination coefficient of the target monitoring point. The basis for taking the maximum value is that when the early deviation of a fire propagates along the airflow direction, the deviation triggering time of the target monitoring point is closest to the deviation triggering time of the upstream adjacent monitoring point, and the deviation coordination coefficient is the largest. Therefore, taking the maximum value best reflects the consistency between the deviation of the target monitoring point and the spatial propagation process of the early stage of the fire.

[0150] Furthermore, when the deviation coordination coefficient is less than the preset coordination threshold, the order of the deviation triggering time of the target monitoring point and the deviation triggering time of adjacent monitoring points is further determined. If the deviation triggering time of the target monitoring point is earlier than that of the adjacent monitoring points, the target monitoring point is determined to be a potential fire source location, and the monitoring weight is not set to zero; instead, the monitoring priority and sampling frequency of the monitoring point are increased. If the deviation triggering time of the target monitoring point is later than that of the adjacent monitoring points, and the time difference exceeds the normal range, the sensor of the target monitoring point is determined to be faulty or subject to local interference, and the monitoring weight is set to zero.

[0151] The preset coordination threshold is set to 0.3. The basis for this setting is that when the deviation coordination coefficient is lower than 0.3, the time difference between the target monitoring point and the adjacent monitoring point exceeds T1×ln(1 / 0.3)≈15×1.2=18 seconds. This time difference has significantly exceeded the normal time range for early fire deviation propagation between adjacent monitoring points. The deviation of the target monitoring point is not synchronized with the deviation of the adjacent monitoring points, and the reliability of the target monitoring point data is insufficient.

[0152] The 18-second time difference is determined as follows: when the distance between adjacent monitoring points is 40 meters and the minimum spread velocity of the fire in the early stage is 3 m / s, the maximum time difference for the deviation to spread to the adjacent monitoring point is about 13.3 seconds. After considering the difference in sensor response time and sampling period error, the upper limit of the normal time range is extended to 18 seconds.

[0153] After the monitoring weights are set to zero, the spatial propagation reliability and physical consistency reliability of the target monitoring point are not included in the maximum value calculation in the fire risk emergency index calculation, and the first correction coefficient of the target monitoring point is not included in the minimum value calculation. When the monitoring weights of all monitoring points in the monitoring area are set to zero, the fire risk emergency index F is set to 0, the warning level is judged as normal, the system maintains normal monitoring status, and sends sensor fault prompt information to the monitoring platform.

[0154] Finally, when the deviation coordination coefficient is greater than or equal to the preset coordination threshold, the sequential relationship between the deviation trigger time of the target monitoring point and the deviation trigger times of adjacent monitoring points is further determined. When the deviation trigger time of the target monitoring point is earlier than that of adjacent monitoring points, the target monitoring point is determined to be a candidate location for the fire source. The determination is based on the fact that the target monitoring point senses early fire characteristics before adjacent monitoring points, indicating that the fire source is closer to the monitoring point, and the deviation signal of the monitoring point has fire source indication significance, so the monitoring priority and sampling frequency should be increased. When the deviation trigger time of the target monitoring point is later than that of adjacent monitoring points, it is not determined to be a candidate location for the fire source, and the normal monitoring weight and regular sampling cycle are maintained. This only indicates that the deviation between the target monitoring point and adjacent monitoring points has temporal coordination, and the data reliability is normal.

[0155] The sampling frequency is increased by adjusting the sampling period of all sensors at the target monitoring point from 1 second to 0.5 seconds. After adjusting the sampling frequency to 0.5 seconds, the baseline prediction model is adapted as follows: For training data, the model was retrained using historical sensor data with a sampling period of 0.5 seconds. If no data with a 0.5-second sampling period was available in the historical data, linear interpolation was performed on the original historical data with a 1-second sampling period to generate historical data with a 0.5-second sampling period. The interpolation method was as follows: an intermediate value was inserted between two adjacent 1-second sampling points, and the intermediate value was the arithmetic mean of the values ​​of the two adjacent sampling points.

[0156] For normalization, during retraining, the maximum and minimum values ​​and normalization parameters are recalculated for historical data with a 0.5-second sampling period, instead of using the normalization parameters from the 1-second sampling period. Denormalization also uses the newly calculated parameters.

[0157] Regarding model parameters, the parameters of the original 1-second sampling period model are not inherited. Instead, the weights and biases of each layer of the Long Short-Term Memory network are reinitialized and the network is trained from scratch.

[0158] In terms of training methods, the input sequence length remains unchanged at 120 sampling points, corresponding to a physical time range of 60 seconds. The training and validation set partitioning, loss function, optimizer, learning rate, batch size, and number of training epochs are consistent with the 1-second sampling period model, with the learning rate still at 0.001, batch size at 32, and number of training epochs at 100.

[0159] Regarding online updates, after the sampling period is switched to 0.5 seconds, the original 1-second sampling period model is still used to predict the newly collected 0.5-second data before the retrained new model is put into use. The original model input is still constructed with 120 sampling points, but at this time, 120 sampling points correspond to a 60-second physical time range. The time span of the model input data is inconsistent with that during training, so this transition phase does not exceed 30 seconds. After retraining is completed, the new model immediately replaces the original model and is put into use.

[0160] The new model replaces the old model in the following way: after the new model is trained, it replaces the old model at the beginning of the next calculation cycle. Before the replacement, the old model continues to output predicted values, and after the replacement, the new model starts to output predicted values. The calculation of deviation values ​​is not interrupted during the replacement process.

[0161] The duration of the increased sampling frequency is consistent with the duration of the candidate fire source location status. When the deviation coordination coefficient falls below the preset coordination threshold or the warning level returns to normal, the sampling period returns to 1 second. When the system is at a blue warning level or above, even if the deviation coordination coefficient is lower than the preset coordination threshold, the sampling period remains at 0.5 seconds. The sampling period control priority for the blue warning level is higher than that for the fire source candidate rule.

[0162] For example, the target monitoring point is monitoring point B, and the adjacent monitoring points are monitoring points A and C. The temperature deviation trigger time for monitoring point B is at 27 seconds, for monitoring point A it is at 18 seconds, and for monitoring point C it is at 36 seconds. The time difference between monitoring point B and monitoring point A is |27-18|=9 seconds, and the time difference between monitoring point B and monitoring point C is |27-36|=9 seconds. The deviation synergy coefficient C4=e (-9 / 15)The value is approximately 0.549, which is greater than the preset collaborative threshold of 0.3. The deviation trigger time of monitoring point B is later than that of monitoring point A but earlier than that of monitoring point C. It does not meet the condition of being earlier than all adjacent monitoring points, so it is not determined as a candidate location for the fire source and the normal monitoring weight and regular sampling cycle are maintained.

[0163] In another scenario, the target monitoring point is monitoring point A, and the adjacent monitoring point is monitoring point B. The temperature deviation trigger time for monitoring point A is at second 18, while the temperature deviation trigger time for monitoring point B is at second 27. Monitoring point A's deviation triggers earlier than monitoring point B's. The deviation synergy coefficient C4 = e (-9 / 15) ≈0.549, which is greater than the preset collaborative threshold of 0.3, and the deviation of monitoring point A from the trigger time is earlier than that of the adjacent monitoring point B. Therefore, monitoring point A is determined to be a candidate location for the fire source, the monitoring priority is increased, and the sampling period is adjusted to 0.5 seconds.

[0164] The above-mentioned collaborative time constant, preset collaborative threshold, and sampling frequency adjustment values ​​are all example values. In actual applications, they should be calibrated according to the sensor layout in the monitoring area and the early propagation characteristics of the fire.

[0165] The method provided by this invention further includes: The changes in gas concentration deviation and smoke concentration deviation during the period of temperature deviation decrease were obtained, and the synchronicity coefficient between the decrease in gas concentration deviation and the decrease in smoke concentration deviation was calculated. The synchronicity coefficient was positively correlated with both the changes in gas concentration deviation and the changes in smoke concentration deviation. Obtain the wind speed change within the same time period and calculate the correlation coefficient between the wind speed change and the temperature deviation reduction. When the synchronization coefficient is greater than or equal to the preset synchronization threshold and the correlation coefficient is less than the preset correlation threshold, it is determined that the temperature deviation reduction is caused by the actual weakening of the fire, thus reducing the fire risk emergency index. When the synchronicity coefficient is less than the preset synchronicity threshold or the correlation coefficient is greater than or equal to the preset correlation threshold, it is determined that the temperature deviation decrease is caused by a temporary change in ventilation conditions, and the current fire risk emergency index is maintained.

[0166] Specifically, firstly, the changes in gas concentration deviation and smoke concentration deviation during the temperature deviation reduction period are obtained, and the synchronization coefficient between the decrease in gas concentration deviation and the decrease in smoke concentration deviation is calculated. The temperature deviation reduction period is the time period during which the temperature deviation value continuously decreases from its peak. The peak value is determined as follows: when the current warning level is blue or above, the maximum value of the temperature deviation value within the most recent 60 seconds is taken as the peak value, and the time when the peak value appears is the starting time of the temperature deviation reduction period; from the starting time, the temperature deviation value must continuously decrease for no less than 5 sampling points to confirm the establishment of the temperature deviation reduction period; otherwise, this supplementary scheme is not triggered. The change in gas concentration deviation is the decrease in the carbon monoxide deviation value during this period, and the change in smoke concentration deviation is the decrease in the smoke concentration deviation value during this period.

[0167] The synchronicity coefficient measures whether the gas concentration deviation and smoke concentration deviation decrease synchronously when the temperature deviation decreases. The formula for calculating the synchronicity coefficient is C5 = min(ΔF1 / F1, ΔF2 / F2), where ΔF1 is the decrease in gas concentration deviation, F1 is the peak value of the gas concentration deviation, ΔF2 is the decrease in smoke concentration deviation, and F2 is the peak value of the smoke concentration deviation. The synchronicity coefficient ranges from 0 to 1; a higher value indicates a greater degree of synchronous decrease in both gas and smoke concentration deviations.

[0168] Secondly, obtain the wind speed change within the same time period and calculate the correlation coefficient between the wind speed change and the temperature deviation decrease. The wind speed change is the final value of the wind speed sequence within that time period minus the initial value, and the temperature deviation decrease is the initial value of the temperature deviation within that time period minus the final value. The correlation coefficient is calculated using the formula C6=min(1,|ΔW| / W0×|ΔF3| / F3), where ΔW is the wind speed change, W0 is the wind speed normalization constant, ΔF3 is the temperature deviation decrease, and F3 is the temperature deviation normalization constant.

[0169] The wind speed normalization constant W0 is set to 1 m / s, and the temperature deviation normalization constant F3 is set to 0.5℃. This is based on the following: the normal ventilation wind speed fluctuation range in mine roadways is generally within 1 m / s, and the temperature deviation fluctuation range in the early stages of a fire is generally within 0.5℃. These values ​​are used as normalization references. The correlation coefficient ranges from 0 to 1; a larger value indicates a stronger correlation between the decrease in temperature deviation and the change in wind speed, and a higher probability that the decrease in temperature deviation is caused by changes in ventilation conditions.

[0170] Furthermore, when the synchronization coefficient is greater than or equal to the preset synchronization threshold and the correlation coefficient is less than the preset correlation threshold, it is determined that the decrease in temperature deviation is caused by the actual weakening of the fire, thereby reducing the fire risk emergency index.

[0171] The preset synchronization threshold and preset correlation threshold are determined as follows: Based on historical early fire data and normal operating condition data in the monitoring area, with the false alarm rate not exceeding the preset upper limit of the false alarm rate as a constraint and the minimum missed alarm rate as the objective, a grid search is performed with a step size of 0.05 on the preset synchronization threshold in the range of 0.3 to 0.7 and the preset correlation threshold in the range of 0.2 to 0.5, and the threshold combination that meets the constraints and has the lowest missed alarm rate is selected; if there is no historical fire data, the preset synchronization threshold of 0.5 and the preset correlation threshold of 0.3 are taken as the default values.

[0172] The above thresholds are set based on the fact that when the fire actually weakens, the temperature, carbon monoxide concentration and smoke concentration usually drop synchronously, with a high synchronicity coefficient. At the same time, the correlation between wind speed changes and temperature deviation decreases is weak, with a low correlation coefficient.

[0173] The method for reducing the fire risk emergency index is as follows: the fire risk emergency index F for the current calculation period is multiplied by an adjustment factor of 0.8. The adjustment factor 0.8 is used and calculated in the following order: after the fire risk emergency index F = S1 - S2 is calculated, if F is greater than 0, F is multiplied by 0.8 to obtain the adjusted fire risk emergency index; if F is less than or equal to 0, F remains unchanged at 0. The adjusted F value is then used in subsequent graded early warning control steps to determine the early warning level.

[0174] The adjustment coefficient of 0.8 is set based on the following: when the fire intensity truly weakens, the contribution of deviations from the early characteristics of a fire has decreased, but has not been completely eliminated. A coefficient of 0.8 is retained to reflect that although the fire risk has decreased, a certain level of monitoring intensity still needs to be maintained to avoid underreporting due to premature and significant downward adjustments to the F-value. This coefficient can be calibrated based on the average decrease in the F-value during the fire intensity weakening phase in historical fire data.

[0175] Finally, when the synchronicity coefficient is less than the preset synchronicity threshold or the correlation coefficient is greater than or equal to the preset correlation threshold, it is determined that the decrease in temperature deviation is caused by a temporary change in ventilation conditions, and the current fire risk emergency index is maintained. In this case, the F value is not adjusted.

[0176] For example, at a certain monitoring point, during the period of decreasing temperature deviation, the temperature deviation decreased from 0.9℃ to 0.5℃, with a decrease in temperature deviation ΔF3 = 0.4℃. During the same period, the carbon monoxide deviation decreased from 2.0ppm to 1.4ppm, with a decrease in gas concentration deviation ΔF1 = 0.6ppm, and the peak gas concentration deviation was 2.0ppm. The smoke concentration deviation decreased from 0.18mg / m³. 3 Reduced to 0.12 mg / m³ 3 The smoke concentration deviated from the reduction amount ΔF2 = 0.06 mg / m³ 3 The smoke concentration deviated from the peak value by 0.18 mg / m³. 3The synchronicity coefficient C5 = min(0.6 / 2.0, 0.06 / 0.18) = min(0.3, 0.333) = 0.3. Since the synchronicity coefficient of 0.3 is lower than the preset synchronicity threshold of 0.5, it is determined that the decrease in temperature deviation is caused by a temporary change in ventilation conditions, and the current fire risk emergency index remains unchanged.

[0177] In another scenario, the temperature deviation decreased from 0.9℃ to 0.5℃, with ΔF3 = 0.4℃. The carbon monoxide deviation decreased from 2.0 ppm to 0.8 ppm, with ΔF1 = 1.2 ppm, and a peak value of 2.0 ppm. The smoke concentration deviation decreased from 0.18 mg / m³. 3 Reduced to 0.06 mg / m³ 3 ΔF2 = 0.12 mg / m 3 The peak value was 0.18 mg / m³. 3 The synchronicity coefficient C5 = min(1.2 / 2.0, 0.12 / 0.18) = min(0.6, 0.667) = 0.6. The synchronicity coefficient of 0.6 is higher than the preset synchronicity threshold of 0.5. During the same period, the wind speed changes from 1.8 m / s to 1.9 m / s, with a wind speed change ΔW = 0.1 m / s. The correlation coefficient C6 = min(1, |0.1| / 1×|0.4| / 0.5) = min(1, 0.1×0.8) = min(1, 0.08) = 0.08. The correlation coefficient of 0.08 is lower than the preset correlation threshold of 0.3. Therefore, the decrease in temperature deviation is determined to be due to a genuine weakening of the fire. The fire risk emergency index F for the current calculation period is adjusted from 0.6 multiplied by 0.8 to 0.48.

[0178] The above-mentioned synchronicity coefficient, correlation coefficient, preset threshold and adjustment coefficient are all example values. In actual application, they should be calibrated according to the historical fire data of the monitoring area and engineering experience.

[0179] In this invention, determining that the temperature deviation decrease is caused by a brief change in ventilation conditions, further includes: The evolution trajectory of temperature deviation within subsequent time windows is continuously monitored, and the ratio of the rebound magnitude to the decrease magnitude of temperature deviation is calculated as the deviation rebound coefficient. Obtain the historical peak value of the temperature deviation before the reduction, and compare the rebound peak value of the current temperature deviation with the historical peak value of the temperature deviation; When the rebound peak exceeds the temperature deviating from the historical peak, it is determined that the fire is in a stage of fluctuation during the spread process, and the fire risk emergency index is increased. When the rebound peak does not exceed the historical temperature deviation peak and the subsequent temperature deviation continues to decrease, it is determined that the fire is gradually weakening, and the fire risk emergency index is reduced.

[0180] Specifically, the evolution trajectory of temperature deviation within subsequent time windows is continuously monitored, and the ratio of the rebound magnitude to the reduction magnitude of temperature deviation is calculated as the deviation rebound coefficient.

[0181] The subsequent time window is set to 60 seconds after the temperature deviation decreases due to a brief change in ventilation conditions. The rationale is that the temperature deviation caused by a brief change in ventilation conditions in a mine roadway generally completes its decline within 30 to 60 seconds. If the temperature deviation rebounds within 60 seconds, the rebound is related to the ventilation change; if it does not rebound after 60 seconds, the decrease in temperature deviation is more likely caused by a genuine weakening of the fire.

[0182] The deviation rebound coefficient is calculated using the formula R = ΔD3 / ΔD4, where ΔD3 is the magnitude of the temperature deviation rebound from its lowest point within the subsequent time window, and ΔD4 is the amount of temperature deviation reduction during the period of temperature reduction. The lowest point after the reduction is determined by taking the minimum value of the temperature deviation value sequence within the subsequent time window as the lowest point. The deviation rebound coefficient ranges from 0 to above, with a larger value indicating a more significant temperature deviation rebound. The deviation rebound coefficient is used to assist in judging the rebound intensity but is not directly used as a criterion for increasing or decreasing the fire risk emergency index; it is only used to describe the rebound characteristics.

[0183] Furthermore, the historical peak value of temperature deviation before the decrease is obtained, and the rebound peak value of the current temperature deviation is compared with the historical peak value of temperature deviation. The historical peak value of temperature deviation is the maximum temperature deviation before the start of the current temperature deviation decrease period. The rebound peak value is the maximum value reached by the temperature deviation value during the subsequent time window.

[0184] The rebound peak is determined as follows: within the subsequent time window, the temperature deviation value must be higher than the lowest point after the decrease for three consecutive sampling points and show an upward trend in order to confirm the formation of the rebound. The rebound peak value is the maximum value of the temperature deviation value within the subsequent time window.

[0185] Finally, when the rebound peak exceeds the temperature deviation from the historical peak, it is determined that the fire is in a phase of fluctuation during the spread process, and the fire risk emergency index is increased. The increase is achieved by multiplying the current calculation period's fire risk emergency index F by an adjustment coefficient of 1.2. The adjustment coefficient of 1.2 is set based on the following: during phased fluctuations in fire intensity, the contribution of deviations from the early characteristics of a fire increases again, but has not yet reached a level of sustained deterioration. A coefficient of 1.2 is retained to reflect the risk rebound without excessive amplification, avoiding false alarms due to single-point fluctuations. This coefficient can be calibrated based on the average rebound magnitude of the F value during fire intensity fluctuation phases in historical fire data.

[0186] When the rebound peak does not exceed the historical temperature deviation peak and the subsequent temperature deviation continues to decrease, the fire is judged to be gradually weakening, and the fire risk emergency index is reduced. The reduction method is: the fire risk emergency index F of the current calculation period is multiplied by an adjustment coefficient of 0.8. The adjustment coefficient of 0.8 is set on the basis that when the fire gradually weakens, the contribution of deviation from the early characteristics of the fire continues to decrease, but has not been completely eliminated. The coefficient of 0.8 is retained to reflect the risk reduction but still maintain the monitoring intensity.

[0187] The criteria for determining a continuous decrease in temperature deviation are as follows: within the subsequent time window, the temperature deviation value decreases continuously for no less than 5 sampling points from the rebound peak or the lowest point after the decrease, and the cumulative decrease is no less than 0.1℃.

[0188] When the rebound peak equals the temperature deviation from the historical peak, it is treated as if the rebound peak did not exceed the temperature deviation from the historical peak. However, the fire risk emergency index is only reduced if the subsequent temperature deviation continues to decrease. When the rebound peak does not exceed the temperature deviation from the historical peak but the subsequent temperature deviation does not continue to decrease, the current fire risk emergency index remains unchanged. If the fire risk emergency index F is multiplied by the adjustment coefficient and the F value exceeds 1, the F value is truncated to 1.

[0189] For example, at a monitoring point, after determining that the decrease in temperature deviation was caused by a brief change in ventilation conditions, the temperature deviation value rebounded from 0.5℃ to 0.9℃ within the following 60 seconds, with a rebound magnitude of 0.9 - 0.5 = 0.4℃. The decrease in temperature deviation during the reduction period was 0.4℃, and the rebound coefficient R = 0.4 / 0.4 = 1.0. The historical peak temperature deviation was 0.9℃, and the current rebound peak was also 0.9℃. The rebound peak did not exceed the historical peak temperature deviation, but the condition for a continued decrease in temperature deviation was not met, and the current fire risk emergency index remained unchanged.

[0190] In another scenario, the temperature deviation value rebounded from 0.5℃ to 1.2℃ within the following 60 seconds, with a rebound magnitude of 1.2 - 0.5 = 0.7℃. The temperature deviation decreased by 0.4℃ during the reduction period, and the rebound coefficient R = 0.7 / 0.4 = 1.75. The historical peak temperature deviation was 0.9℃, and the current rebound peak was 1.2℃. Since the rebound peak exceeded the historical peak temperature deviation, the fire was determined to be in a phase of fluctuation during its spread. The fire risk emergency index F for the current calculation period was adjusted from 0.35 multiplied by 1.2 to 0.42.

[0191] In another scenario, the temperature deviation continued to decrease from 0.5℃ to 0.2℃ within the next 60 seconds without rebounding. The temperature deviation continued to decrease, indicating that the fire was gradually weakening. The fire risk emergency index F for the current calculation period was adjusted from 0.35 multiplied by 0.8 to 0.28. The above deviation rebound coefficient, adjustment coefficient, and example values ​​are for illustrative purposes only and should be calibrated based on historical fire data of the monitored area in actual applications.

[0192] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0193] In summary, this invention effectively distinguishes between early fire characteristics and deviations caused by environmental factors by analyzing the spatiotemporal propagation physical mechanism of deviations from multiple monitoring points and multiple sensors, thereby improving the accuracy of early anomaly identification and reducing the false alarm rate. Through dynamic calculation of the fire risk emergency index and hierarchical early warning control, it achieves a closed-loop response from the quantification of deviation causes to the determination of early warning levels and the execution of on-site control actions, thereby improving the timeliness of early warning response.

[0194] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0195] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. A graded early warning control method for an automatic fire extinguishing system in a mine, characterized in that, include: The measured value sequences of multiple sensors at multiple monitoring points within the mining monitoring area are obtained, along with the predicted value sequences output by the baseline prediction model. The difference between the two is calculated to generate the time-series deviation sequence for each monitoring point. Based on the spatial and temporal propagation characteristics of the time-series deviation sequences at each monitoring point, the physical driving factors of the deviation changes are analyzed. The physical driving factors include the spatial propagation path of the deviation, the physical consistency of multiple parameters, the contribution of environmental factors, and the reversibility characteristics of the deviation. Based on the comprehensive analysis results of physical driving factors, the fire risk emergency index is dynamically calculated, and graded early warning control is implemented according to the fire risk emergency index.

2. The graded early warning control method for an automatic fire extinguishing system in a mine as described in claim 1, characterized in that, Analyze the deviated spatial propagation path, including: Obtain the time-series deviation sequence of multiple monitoring points deployed along the wind flow direction within the monitoring area, and determine the deviation trigger time of each monitoring point. The deviation trigger time is the moment when the time-series deviation sequence first exceeds the preset deviation threshold. Based on the deviation trigger time and spatial location of each monitoring point, the propagation velocity vector of the deviation is calculated, and the direction of deviation propagation is determined based on the propagation velocity vector; Obtain the airflow direction during the current monitoring period, calculate the angle between the deviation from the propagation direction and the airflow direction, and determine the propagation direction consistency coefficient based on the angle. The propagation direction consistency coefficient is negatively correlated with the angle. The reliability of spatial propagation is determined based on the consistency coefficient of propagation direction and the deviation from propagation speed. The reliability of spatial propagation is positively correlated with the consistency coefficient of propagation direction.

3. A graded early warning control method for an automatic fire extinguishing system in a mine, as described in claim 1, characterized in that... Analysis of multi-parameter physical consistency includes: Obtain temperature deviation sequences, gas concentration deviation sequences, and smoke concentration deviation sequences; Determine the trigger time for temperature deviation, and based on the trigger time and the preset time delay window, obtain the trigger status of gas concentration deviation and smoke concentration deviation within the time delay window; Calculate the first deviation between the actual trigger delay and the theoretical trigger delay of the gas concentration deviation, and calculate the second deviation between the actual trigger delay and the theoretical trigger delay of the smoke concentration deviation; The delay consistency coefficient is determined based on the first deviation and the second deviation. The delay consistency coefficient is negatively correlated with both the first deviation and the second deviation. Physical consistency reliability is determined based on the delay consistency coefficient, and physical consistency reliability is positively correlated with the delay consistency coefficient.

4. A graded early warning control method for an automatic fire extinguishing system in a mine, as described in claim 1, characterized in that... The analysis of the contribution of environmental factors includes: Acquire the mine ventilation parameter sequence and equipment operating status parameter sequence during the monitoring period. The ventilation parameter sequence shall include at least the wind speed sequence. Based on the ventilation parameter sequence and the equipment operating status parameter sequence, determine the contribution of each parameter to the normal fluctuation range of temperature, gas concentration and smoke concentration; The contribution of environmental factors is determined by the ratio of the contribution amount to the deviation amount of the time series deviation sequence; The first correction coefficient of the fire risk emergency index is determined based on the contribution of environmental factors, and the first correction coefficient is positively correlated with the contribution of environmental factors.

5. A graded early warning control method for an automatic fire extinguishing system in a mine, as described in claim 4, characterized in that... Analysis of deviations from reversibility characteristics includes: When the contribution of environmental factors exceeds the preset contribution threshold, the regression characteristics of the time-series deviation sequence of the ventilation parameter sequence after it has recovered to the normal state are obtained. The regression rate of the time series deviation sequence after the ventilation parameter sequence is recovered is calculated, and the deviation reversibility coefficient is determined based on the regression rate. The deviation reversibility coefficient is positively correlated with the regression rate. The second correction factor for the fire risk emergency index is determined based on the deviation reversibility coefficient, and the second correction factor is positively correlated with the deviation reversibility coefficient.

6. A graded early warning control method for an automatic fire extinguishing system in a mine, as described in claim 5, characterized in that... Based on the comprehensive analysis of physical driving factors, the fire risk emergency index is dynamically calculated, including: Obtain spatial propagation credibility, physical consistency credibility, first correction coefficient, and second correction coefficient; The first comprehensive value is obtained by weighted summation of the credibility of spatial propagation and the credibility of physical consistency. The environmental correction value is obtained by weighted summing of the first and second correction factors. The fire risk emergency index is obtained by subtracting the environmental correction value from the first comprehensive value.

7. A graded early warning control method for an automatic fire extinguishing system in a mine, as described in claim 2, characterized in that... Also includes: Based on the deviation trigger time and spatial spacing of each monitoring point, the propagation speed of the deviation along the airflow direction is calculated; A preset fire spread rate threshold is obtained. When the spread rate exceeds the preset fire spread rate threshold, the fire is determined to be in a rapid spread stage, and the weight coefficient of spatial spread credibility in the calculation of the first comprehensive value is increased. When the spread rate is lower than the preset fire spread rate threshold, the fire is determined to be in the smoldering stage, and the current monitoring frequency is maintained.

8. A graded early warning control method for an automatic fire extinguishing system in a mine, as described in claim 2, characterized in that... Also includes: Obtain the time-series deviation sequence of the target monitoring point and the time-series deviation sequence of adjacent monitoring points, and calculate the deviation coordination coefficient between the target monitoring point and adjacent monitoring points. The deviation coordination coefficient is negatively correlated with the time difference between the deviation trigger times of the two monitoring points. When the deviation coordination coefficient is less than the preset coordination threshold and the deviation trigger time of the target monitoring point is later than the deviation trigger time of the adjacent monitoring point, it is determined that the sensor of the target monitoring point is faulty or subject to local interference, and the monitoring weight of the target monitoring point is set to zero. When the deviation coordination coefficient is greater than or equal to the preset coordination threshold and the deviation trigger time of the target monitoring point is earlier than the deviation trigger time of the adjacent monitoring point, the target monitoring point is determined to be a candidate location of the fire source, and the monitoring priority and sampling frequency of the target monitoring point are increased.

9. A graded early warning control method for a mine automatic fire extinguishing system as described in claim 4, characterized in that, Also includes: The changes in gas concentration deviation and smoke concentration deviation during the period of temperature deviation decrease were obtained, and the synchronicity coefficient between the decrease in gas concentration deviation and the decrease in smoke concentration deviation was calculated. The synchronicity coefficient was positively correlated with both the changes in gas concentration deviation and the changes in smoke concentration deviation. Obtain the wind speed change within the same time period and calculate the correlation coefficient between the wind speed change and the temperature deviation reduction. When the synchronization coefficient is greater than or equal to the preset synchronization threshold and the correlation coefficient is less than the preset correlation threshold, it is determined that the temperature deviation reduction is caused by the actual weakening of the fire, thus reducing the fire risk emergency index. When the synchronicity coefficient is less than the preset synchronicity threshold or the correlation coefficient is greater than or equal to the preset correlation threshold, it is determined that the temperature deviation decrease is caused by a temporary change in ventilation conditions, and the current fire risk emergency index is maintained.

10. A graded early warning control method for a mine automatic fire extinguishing system as described in claim 9, characterized in that, The determination that the temperature deviation was caused by a brief change in ventilation conditions, followed by: The evolution trajectory of temperature deviation within subsequent time windows is continuously monitored, and the ratio of the rebound magnitude to the decrease magnitude of temperature deviation is calculated as the deviation rebound coefficient. Obtain the historical peak value of the temperature deviation before the reduction, and compare the rebound peak value of the current temperature deviation with the historical peak value of the temperature deviation; When the rebound peak exceeds the temperature deviating from the historical peak, it is determined that the fire is in a stage of fluctuation during the spread process, and the fire risk emergency index is increased. When the rebound peak does not exceed the historical temperature deviation peak and the subsequent temperature deviation continues to decrease, it is determined that the fire is gradually weakening, and the fire risk emergency index is reduced.

Images