A coal mine gas concentration monitoring and early warning system

Abstract

This invention discloses a coal mine gas concentration monitoring and early warning system, belonging to the field of safety early warning technology. It aims to solve the technical problems of insufficient monitoring accuracy and untimely early warning in traditional coal mine gas concentration monitoring and early warning systems. The system includes: a monitoring module for real-time monitoring of environmental temperature, humidity, oxygen concentration, and gas concentration values, and transmitting these values, along with the location data of the monitoring nodes, to a data processing module; an adaptive compensation module for receiving the real-time temperature, humidity, oxygen concentration, and gas concentration values ​​transmitted from the data processing module, correcting them to obtain standard values ​​for temperature, humidity, oxygen concentration, and gas concentration; and the data processing module itself. This invention has the advantages of accurate and comprehensive real-time monitoring and effective early warning of underground coal mine gas concentration.

Description

Technical Field

[0001] This invention relates to the field of safety early warning technology, and more specifically, to a coal mine gas concentration monitoring and early warning system. Background Technology

[0002] Accurate monitoring and timely early warning of methane concentration are crucial in coal mining. Currently, traditional coal mine methane concentration monitoring and early warning systems face challenges in practical applications. They struggle to accurately and comprehensively monitor and provide effective early warnings of methane concentration in the complex underground environment, failing to fully guarantee the safety and reliability of coal mining. Specifically, traditional systems lack sufficient accuracy in monitoring parameters such as methane concentration when faced with fluctuations in environmental parameters like temperature and humidity underground, as well as the influence of sensor performance. Furthermore, they lack effective multi-parameter collaborative analysis and dynamic early warning mechanisms, making it difficult to accurately determine dangerous methane concentration conditions and issue timely warning signals. Therefore, we propose a coal mine methane concentration monitoring and early warning system. Summary of the Invention

[0003] The purpose of this invention is to provide a coal mine gas concentration monitoring and early warning system to solve the technical problems of insufficient monitoring accuracy and untimely early warning in traditional coal mine gas concentration monitoring and early warning systems.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a coal mine gas concentration monitoring and early warning system, comprising:

[0005] The monitoring module is used to monitor the temperature, humidity, oxygen concentration, and gas concentration of the environment in real time, and transmit the temperature, humidity, oxygen concentration, and gas concentration values, as well as the location data of the monitoring node, to the data processing module.

[0006] The adaptive compensation module is used to receive the real-time monitoring values ​​of temperature, humidity, oxygen concentration and gas concentration transmitted by the data processing module from the monitoring node, and to correct them to obtain the standard values ​​of temperature, humidity, oxygen concentration and gas concentration.

[0007] The data processing module is used to determine whether at least two of the standard values ​​of temperature, humidity, oxygen concentration, and gas concentration are abnormal. If so, it controls the alarm communication module to issue a warning signal and sends the warning signal to the receiving end; if not, it further determines whether there is a gas concentration danger signal.

[0008] The alarm communication module is used to transmit warning signals, hazard level signals, abnormal concentration signals, or normal gas concentration signals when the corresponding conditions are met.

[0009] The receiving end is used to receive signals sent by the alarm communication module and take corresponding measures based on the content of the signals.

[0010] Preferably, when the data processing module determines whether there is a gas concentration danger signal, it compares the real-time monitored temperature, humidity, oxygen concentration and gas concentration values ​​with the normal temperature, normal humidity, normal oxygen concentration and normal gas concentration values.

[0011] The system automatically determines the normal temperature, humidity, oxygen concentration, and gas concentration values ​​as follows: The data processing module uses a big data analysis algorithm that combines time series decomposition and machine learning based on historical monitoring data to decompose the monitoring data into time dimensions of hours, days, and months, and extract the variation characteristics of temperature, humidity, oxygen concentration, and gas concentration at different time scales.

[0012] Preferably, the specific method by which the data processing module determines the gas concentration danger signal is as follows:

[0013] Determine whether the temperature, humidity, oxygen concentration, and methane concentration exceed the normal values ​​for temperature, humidity, oxygen concentration, and methane concentration, respectively.

[0014] 3 is used for judgment. In principle, when the real-time monitored value exceeds the range of "the mean of the corresponding parameter within the normal range ± 3 times the standard deviation", the monitored value is determined to be outside the normal range.

[0015] If the standard values ​​for temperature, humidity, oxygen concentration, or gas concentration exceed the normal values ​​for temperature, humidity, oxygen concentration, or gas concentration, then the concentration at the time of detection is a gas concentration danger signal.

[0016] Otherwise, by further calculating the product of temperature and humidity and comparing it with a preset value, it can be determined whether there is an abnormality in sensor data caused by the combined effect of temperature and humidity.

[0017] The gas concentration hazard signal indicates that the detection location at the time of detection is a hazardous area.

[0018] Preferably, the data processing module counts the number of abnormal sensor output data caused by the combined effect of temperature and humidity that constitute a gas concentration hazard signal. If the count is greater than or equal to 2, the concentration at the detection time is determined to be a gas concentration hazard signal.

[0019] The gas concentration hazard signal is divided into three levels: minor hazard, moderate hazard, and major hazard.

[0020] Minor danger refers to a situation where only one parameter exceeds the normal range and no combined temperature and humidity anomalies are triggered;

[0021] Moderate risk refers to the presence of 2-3 parameters exceeding the normal range or the triggering of combined abnormalities in temperature and humidity;

[0022] Major danger refers to all four parameters exceeding the normal range or two consecutive moderate danger signals appearing in the same area within a short period of time;

[0023] Different levels of danger correspond to different graded early warning and response strategies, with audible and visual warnings used for minor dangers;

[0024] Initiate enhanced local ventilation when the risk level is moderate.

[0025] A major hazard triggers an emergency evacuation order for the entire area, and at the same time sends evacuation route information to all personnel positioning equipment underground. The evacuation routes are dynamically generated based on the real-time monitoring of the hazard situation in each area.

[0026] Preferably, the method by which the adaptive compensation module corrects the real-time monitoring values ​​of temperature, humidity, oxygen concentration, and gas concentration received by the data processing module from the monitoring nodes is as follows:

[0027] The calibration procedure is set according to the type of sensor. For temperature sensors, the Steinhart-Hart equation based on platinum resistance is used for nonlinear compensation.

[0028] For humidity sensors, a polynomial fitting algorithm is used to establish a compensation model based on the temperature-humidity cross-sensitivity characteristics;

[0029] For the oxygen concentration sensor, an electrochemical sensor zero-point drift compensation algorithm is used, and dynamic correction is performed using a reference gas calibration coefficient.

[0030] For the gas concentration sensor, a combined algorithm of temperature compensation and aging correction for catalytic combustion sensor is adopted. Temperature compensation is performed linearly based on the temperature sensor data built into the sensor.

[0031] The original data on temperature, humidity, oxygen concentration, and gas concentration are corrected using the aforementioned correction procedure to obtain the corresponding standard values.

[0032] Preferably, the monitoring module uses an edge computing node to perform local data preprocessing. The edge computing node has a built-in FPGA chip and ARM processor collaborative processing architecture. The FPGA chip is responsible for high-speed data acquisition and preliminary processing, while the ARM processor performs complex algorithm calculations to reduce noise, normalize, and compress the data volume of the acquired raw data.

[0033] Preferably, the data processing module adopts a distributed computing architecture, including multiple data processing sub-nodes and a management node. The management node is based on Kubernetes container orchestration technology and dynamically allocates tasks to each sub-node according to the data type and computing workload.

[0034] Preferably, a multi-parameter time series correlation model is established. The multi-parameter time series correlation model has a built-in environmental perception module. The environmental perception module is connected to a ground-penetrating radar, vibration sensor, and equipment status monitor via an RS485 bus to collect environmental information such as geological structure data, mining equipment operating parameters, and ventilation system status of the monitoring area in real time as model input variables.

[0035] The collected data is first converted and validated to ensure its integrity and accuracy before being input into the model;

[0036] An online learning mechanism is set up to automatically trigger the model update process when the model prediction results deviate from the actual monitoring data;

[0037] Using the latest environmental information and monitoring data, the stochastic gradient descent algorithm is employed to dynamically adjust the model parameters, with the adjustment step size adaptively adjusted according to the rate of change of the data.

[0038] Preferably, the online learning mechanism employs a transfer learning algorithm, which rapidly adjusts the parameters of the training model based on existing similar scenarios when a new scenario or working condition arises.

[0039] The domain adaptation algorithm aligns the feature distributions of the source and target domains, reducing model training time and data requirements.

[0040] Preferably, the data processing module uses a long short-term memory network combined with an attention mechanism to construct a prediction model to predict the abnormal development trend of gas concentration.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] 1. This invention, through the design of a monitoring module, an adaptive compensation module, a data processing module, an alarm communication module, and a receiving end, achieves real-time monitoring, precise correction, multi-parameter collaborative analysis, and graded early warning of parameters such as temperature, humidity, oxygen concentration, and gas concentration in coal mines. It effectively solves the problems of insufficient monitoring accuracy and untimely early warning in traditional systems, enabling more accurate and comprehensive real-time monitoring and effective early warning of gas concentration in coal mines. This greatly improves the accuracy and reliability of gas concentration monitoring and early warning during coal mining, providing strong protection for safe coal mine production.

[0043] 2. This invention employs a big data analysis algorithm combining time-series decomposition and machine learning based on historical monitoring data in its data processing module to automatically determine normal temperature, humidity, oxygen concentration, and gas concentration values, and utilizes 3 The principle of comparing the product of temperature and humidity to determine the dangerous signal of gas concentration has enabled accurate judgment and graded early warning of dangerous gas concentration. This further solves the problem of inaccurate judgment of dangerous gas concentration in traditional systems, making the early warning more scientific and reasonable. It can take corresponding early warning response strategies according to different danger levels, thereby improving the ability of coal mines to cope with gas hazards.

[0044] 3. This invention uses an adaptive compensation module to set different calibration programs according to the type of sensor, correcting the raw data of temperature, humidity, oxygen concentration and gas concentration, thereby improving the accuracy and reliability of sensor data. It further solves the problem that the performance of the sensor itself affects the monitoring accuracy, providing a more reliable data foundation for subsequent data processing and early warning, and ensuring that the monitoring and early warning functions of the entire system are more stable and reliable. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0046] To facilitate understanding of the technical solution of the present invention by those skilled in the art, the technical solution of the present invention will now be further described in conjunction with the accompanying drawings.

[0047] Example 1, such as Figure 1 As shown, the present invention provides a coal mine gas concentration monitoring and early warning system, comprising:

[0048] The monitoring module is used to monitor the temperature, humidity, oxygen concentration, and gas concentration of the environment in real time, and transmit the temperature, humidity, oxygen concentration, and gas concentration values, as well as the location data of the monitoring nodes, to the data processing module.

[0049] The adaptive compensation module is used to receive the real-time monitoring values ​​of temperature, humidity, oxygen concentration and gas concentration transmitted by the data processing module from the monitoring node, and to correct them to obtain the standard values ​​of temperature, humidity, oxygen concentration and gas concentration.

[0050] The data processing module is used to determine whether at least two of the standard values ​​of temperature, humidity, oxygen concentration, and gas concentration are abnormal. If so, it controls the alarm communication module to issue a warning signal and sends the warning signal to the receiving end; if not, it further determines whether there is a gas concentration danger signal.

[0051] The alarm communication module is used to transmit warning signals, hazard level signals, abnormal concentration signals, or normal gas concentration signals when the corresponding conditions are met.

[0052] The receiving end is used to receive signals sent by the alarm communication module and take corresponding measures based on the content of the signals.

[0053] In an embodiment of the present invention, when the data processing module determines whether there is a gas concentration danger signal, it compares the real-time monitored temperature, humidity, oxygen concentration and gas concentration values ​​with the normal temperature, normal humidity, normal oxygen concentration and normal gas concentration values.

[0054] The system automatically determines normal temperature, humidity, oxygen concentration, and gas concentration values ​​as follows: The data processing module uses historical monitoring data from the coal mine over the past 3-5 years, employing a big data analysis algorithm combining time-series decomposition and machine learning. It decomposes the monitoring data by hour, day, and month, extracting the variation characteristics of temperature, humidity, oxygen concentration, and gas concentration at different time scales. Specifically:

[0055] For hourly data, analyze its fluctuation range and rate of change;

[0056] For daily data, calculate the daily mean and standard deviation;

[0057] For monthly data, calculate the slope of the monthly trend change;

[0058] Combining geological structure data (including coal seam thickness, rock permeability, etc.), mining progress data (such as the advance distance of the mining face, mining intensity, etc.), and ventilation system status data (such as wind speed, wind pressure, ventilation volume, etc.) of the monitoring sites, a random forest regression model is trained.

[0059] In a random forest regression model, the output of each decision tree Determined by the following formula:

[0060] ;

[0061] in, For the number of nodes in the decision tree, For nodes The weight, For nodes Input features;

[0062] The random forest model contains 500 decision trees, each with a maximum depth of 10. By averaging the outputs of all decision trees, the normal temperature for the corresponding region over the next 72 hours is calculated. Normal humidity Normal oxygen concentration and normal gas concentration value:

[0063] ;

[0064] ;

[0065] ;

[0066] ;

[0067] in, , , and They represent the first The predicted values ​​of temperature, humidity, oxygen concentration, and gas concentration output by the decision tree

[0068] The above values ​​are updated every 6 hours. During the update, an incremental learning approach is used, which only processes newly collected data and incorporates it into the model.

[0069] In an embodiment of the present invention, the specific method by which the data processing module determines the gas concentration danger signal is as follows:

[0070] Determine whether the temperature, humidity, oxygen concentration, and methane concentration exceed the normal values ​​for temperature, humidity, oxygen concentration, and methane concentration, respectively.

[0071] The 3σ principle is used for judgment. When the real-time monitoring value exceeds the range of "mean ± 3 times standard deviation of the corresponding parameter's normal range", it is determined that the monitoring value exceeds the normal range and may be abnormal or dangerous.

[0072] Among them, standard deviation The calculation is based on rolling historical data from the past 7 days, reflecting the data dispersion. The 3σ principle defines a statistically normal fluctuation range using "mean ± 3 standard deviations". Monitoring values ​​exceeding this range are considered low-probability events and thus identified as abnormal. Taking temperature data as an example, the calculation formula is:

[0073] ;

[0074] In the formula, The standard deviation of temperature is used to quantify the dispersion of temperature data. This represents the sample size for temperature data from the past 7 days. For the first Each temperature sample value, This is the average of the temperature data over the past 7 days, i.e. ;

[0075] The standard deviation calculated using the above formula can reflect data fluctuations in real time.

[0076] If the standard values ​​for temperature, humidity, oxygen concentration, or gas concentration exceed the normal values ​​for temperature, humidity, oxygen concentration, or gas concentration, then the concentration at the time of detection is a gas concentration danger signal.

[0077] Otherwise, by further calculating the product of temperature and humidity and comparing it with a preset value, it can be determined whether there is an abnormality in sensor data caused by the combined effect of temperature and humidity.

[0078] A gas concentration hazard signal indicates that the detection location at the time of detection is a hazardous area;

[0079] The preset values ​​are determined using the 95% confidence interval method based on sensor type and historical data statistics.

[0080] The specific determination process involves collecting 1000 sets of historical data for this type of sensor under different temperature and humidity combinations. Calculate product data ,calculate mean and standard deviation :

[0081] ;

[0082] ;

[0083] in, This represents the standard deviation of the temperature-humidity product, used to quantify the dispersion of data under the combined effects of temperature and humidity. Indicates the first The product of temperature and humidity in a set of historical data is used to analyze the combined effect of temperature and humidity on sensor data. Indicates the first Temperature values ​​from a set of historical data. Indicates the first Humidity values ​​from a set of historical data;

[0084] Then the preset value for:

[0085] ;

[0086] in, This indicates the threshold for determining whether the combined effect of temperature and humidity causes abnormal sensor data; exceeding this value indicates an abnormality.

[0087] In the formula, 1.645 is the quantile of the 95% confidence interval of the standard normal distribution;

[0088] If the current calculated value Greater than the preset value If the sensor data is abnormal, it is determined that there is a dangerous or abnormal signal of gas concentration.

[0089] In an embodiment of the present invention, the data processing module counts the number of abnormal sensor output data caused by the combined effect of temperature and humidity that constitute a gas concentration hazard signal. If the count is greater than or equal to 2, the concentration at the detection time is determined to be a gas concentration hazard signal.

[0090] Gas concentration hazard signals are divided into three levels: minor hazard, moderate hazard, and major hazard.

[0091] Minor danger refers to a situation where only one parameter exceeds the normal range and no combined temperature and humidity anomalies are triggered;

[0092] Moderate risk refers to the presence of 2-3 parameters exceeding the normal range or the triggering of combined abnormalities in temperature and humidity;

[0093] Major danger refers to all four parameters exceeding the normal range or two consecutive moderate danger signals appearing in the same area within a short period of time;

[0094] Different levels of danger correspond to different graded early warning and response strategies. Minor dangers are given an audible and visual warning, with the sound intensity of the audible and visual alarm set to 80-90 decibels and the flashing frequency to 2-3 times per second.

[0095] For moderate hazard situations, local ventilation will be enhanced, with the ventilation equipment's airflow increased to 1.5 times the normal level.

[0096] A major hazard triggers an emergency evacuation order for the entire area, and at the same time sends evacuation route information to all personnel positioning equipment underground. The evacuation routes are dynamically generated based on the real-time monitoring of the hazard situation in each area.

[0097] In an embodiment of the present invention, the method by which the adaptive compensation module corrects the temperature, humidity, oxygen concentration, and gas concentration values ​​monitored in real time by the monitoring node received by the data processing module is as follows:

[0098] The calibration procedure is set according to the type of sensor. For temperature sensors, the Steinhart-Hart equation based on platinum resistance is used for nonlinear compensation. The equation is as follows:

[0099] ;

[0100] in, This is the temperature value. The resistance value of the platinum resistance thermometer is... , , The equation parameters were obtained through three calibrations at standard temperature environments (0℃, 25℃, 100℃).

[0101] A second correction is then made based on environmental pressure data; the correction formula is as follows:

[0102] ;

[0103] in, For the corrected temperature value, To measure the temperature value, The pressure correction factor, determined experimentally to be 0.0001 / Pa, This is the pressure deviation value;

[0104] For humidity sensors, a polynomial fitting algorithm is used to establish a compensation model based on the temperature-humidity cross-sensitivity characteristics. The model is a quadratic polynomial:

[0105] ;

[0106] in, This is the corrected humidity value. This is the temperature value. To measure humidity, , , The humidity data was obtained by collecting humidity data at different temperatures and then fitting the data using the least squares method.

[0107] When using least squares fitting, the objective function for:

[0108] ;

[0109] in, Indicates the first The actual humidity values ​​in the set of data, Indicates the first The humidity sensor readings in the data set;

[0110] By minimizing the objective function Solve , , ;

[0111] For the oxygen concentration sensor, an electrochemical sensor zero-point drift compensation algorithm is used, which is dynamically corrected using a reference gas calibration coefficient. Zero-point calibration is performed every 24 hours using standard oxygen concentration gas (20.9%), and the calibration coefficient is updated after calibration. The corrected formula is:

[0112] ;

[0113] in, This is the corrected oxygen concentration value. To measure oxygen concentration, This is the zero-point drift value;

[0114] For gas concentration sensors, a combined algorithm of temperature compensation and aging correction is used for catalytic combustion sensors. Temperature compensation is performed linearly based on data from the sensor's built-in temperature sensor, as shown in the formula:

[0115] ;

[0116] in, This is the gas concentration value after temperature compensation. To measure the gas concentration value, This is the temperature compensation coefficient. This is a reference temperature value.

[0117] Aging correction is achieved by recording sensor usage time. and cumulative number of tests An aging curve model is established for correction, and the aging correction formula is as follows:

[0118] ;

[0119] in, This is the gas concentration value after aging correction. This is the aging correction factor. Design the usage time for the sensor. Design the cumulative number of detections for the sensor;

[0120] The original data on temperature, humidity, oxygen concentration, and gas concentration were corrected using a calibration procedure to obtain the corresponding standard values;

[0121] The calibration procedure involves staff fine-tuning parameters based on factors such as dust concentration and electromagnetic interference intensity in the gas monitoring environment. When the dust concentration exceeds 50 mg / m³... 3 At that time, the humidity sensor compensation model parameter 'a' increases by 10%.

[0122] In an embodiment of the present invention, the monitoring module uses an edge computing node to perform local data preprocessing. The edge computing node has a built-in FPGA chip and ARM processor collaborative processing architecture. The FPGA chip is responsible for high-speed data acquisition and preliminary processing, while the ARM processor performs complex algorithm calculations.

[0123] The collected raw data is denoised, normalized, and the data volume is compressed.

[0124] The noise reduction uses a wavelet thresholding algorithm, selects the db4 wavelet basis function, sets the number of decomposition layers to 3, and uses a soft thresholding function for processing.

[0125] For noisy signals Wavelet coefficients are obtained through wavelet transform. The soft thresholding formula is:

[0126] ;

[0127] in, These are the wavelet coefficients after thresholding. For symbolic functions, The threshold is determined using the general threshold formula. calculate, The noise standard deviation can be estimated using the median method. , The signal length;

[0128] Normalization uses the min-max normalization method to normalize the data to the [0,1] interval, as shown in the formula:

[0129] ;

[0130] in, For normalized data, The original data, and These are the minimum and maximum values ​​in the original data, respectively.

[0131] Data compression uses the LZ77 lossless compression algorithm. During the compression process, the sliding window size is set to 4096 bytes and the forward search buffer size is set to 256 bytes, compressing the data volume to less than 30% of the original, and only transmitting key data features to the data processing module.

[0132] The system uses 5G-U or industrial-grade wireless mesh network to build downhole data transmission channels, and allocates dedicated transmission channels for monitoring data through network slicing technology, with a bandwidth of no less than 100Mbps.

[0133] It also sets up a three-layer redundant transmission path, including a primary path, a backup path, and an emergency path;

[0134] The system monitors network transmission status in real time. When network latency exceeds 50ms or packet loss rate is greater than 5%, it automatically switches to redundant paths and initiates a fast retransmission mechanism based on the TCP protocol.

[0135] When switching paths, a weighted round-robin algorithm is used to select the optimal backup path, with path weights... The calculation formula is:

[0136] ;

[0137] in, For path Historical delay, For path Available bandwidth, This represents the total number of backup paths.

[0138] In an embodiment of the present invention, the data processing module adopts a distributed computing architecture, including multiple data processing sub-nodes and a management node. The management node is based on Kubernetes container orchestration technology and dynamically allocates tasks to each sub-node according to the data type and the amount of computing tasks.

[0139] The specific allocation strategy is to prioritize allocating computational tasks with high real-time requirements to child nodes with lower load and higher performance.

[0140] For batch processing tasks, task sharding is used to distribute the tasks evenly across multiple child nodes.

[0141] Each sub-node uses the Spark Streaming framework for parallel processing. Data processing sub-nodes interact with each other through the ZeroMQ message queue and send the results back to the management node for integration and analysis.

[0142] Using stream computing technology and based on the Flink real-time computing engine, duration sequence association calculations are performed on real-time data streams;

[0143] Data segmentation is performed using the sliding window algorithm, and the similarity of time series with different parameters is calculated by combining the dynamic time warping algorithm.

[0144] When calculating similarity, Euclidean distance is used as the metric, and in order to reduce the amount of computation, the time series is downsampled, and the sampling interval is dynamically adjusted according to the frequency of data change.

[0145] Suppose two time series and After downsampling, it becomes and The DTW algorithm finds the optimal time series alignment path to minimize the sum of the aligned Euclidean distances.

[0146] Define the local distance matrix ,in Cumulative distance matrix The calculation is as follows:

[0147] ;

[0148] in, , ,and , ;

[0149] final This is the DTW distance between two downsampled time series; the smaller the distance, the higher the similarity between the two time series.

[0150] Simultaneously, a water level mechanism is set up to handle out-of-order data, ensuring the accuracy of data processing, and the water level timestamp is included. Based on the maximum delay of event time in the data stream Confirmed, the formula is: ,in This is the current processing time.

[0151] In an embodiment of the present invention, a multi-parameter time series correlation model is established. The multi-parameter time series correlation model has a built-in environmental perception module. The environmental perception module is connected to a ground radar, vibration sensor and equipment status monitor via an RS485 bus to collect environmental information such as geological structure data, mining equipment operating parameters and ventilation system status of the monitored area in real time as model input variables.

[0152] The collected data is first converted and validated to ensure its integrity and accuracy before being input into the model;

[0153] An online learning mechanism is set up to automatically trigger the model update process when the model prediction results deviate from the actual monitoring data;

[0154] Using the latest environmental information and monitoring data, the stochastic gradient descent algorithm is used to dynamically adjust the model parameters, and the adjustment step size is adaptively adjusted according to the rate of change of the data.

[0155] The specific adjustment method is as follows: assuming the current data change rate is... Step size is The step size adjustment formula is:

[0156] ;

[0157] in, The adjusted step size;

[0158] In the stochastic gradient descent algorithm, for model parameters Its update formula is ,in loss function Regarding parameters gradient, For input data, This is a real label.

[0159] In an embodiment of the present invention, the online learning mechanism employs a transfer learning algorithm, which rapidly adjusts the parameters of the training model based on existing similar scenarios when a new scenario or working condition occurs.

[0160] The feature distributions of the source domain (existing scene) and the target domain (new scene) are aligned by the domain adaptation algorithm, reducing model training time and data requirements;

[0161] When aligning feature distributions, the maximum mean difference (MMD) is used as a metric, and distribution alignment is achieved by minimizing the MMD values ​​of the source and target domain features.

[0162] Let the source domain feature set be The target domain feature set is The kernel function is The formula for calculating MMD is:

[0163] ;

[0164] By optimizing the objective function To adjust model parameters ,in Let the loss function be the loss function over the source domain. This is the balance coefficient;

[0165] Even when the amount of data in the target domain is less than 20% of the original training data, the model can still guarantee a prediction accuracy of no less than 85%, thus improving the model's generalization ability in different environments.

[0166] In an embodiment of the present invention, the data processing module uses a long short-term memory network combined with an attention mechanism to construct a prediction model to predict the abnormal development trend of gas concentration.

[0167] The LSTM network contains three hidden layers, each with 128 neurons.

[0168] Forget gate in LSTM unit Input gate Output gate and cell state The update formula is as follows:

[0169] ;

[0170] ;

[0171] ;

[0172] ;

[0173] ;

[0174] ;

[0175] in, For the Sigmoid function, It is the hyperbolic tangent function. , , , This is the weight matrix. , , , For bias vectors, For element-wise multiplication, This indicates that the previous state will be hidden. and current input splicing;

[0176] The attention mechanism highlights the impact of key time steps on the prediction results by calculating the attention weights at different time steps;

[0177] Set attention weights The calculation method is as follows:

[0178] ;

[0179] ;

[0180] in, This is the weight matrix. It is the bias vector;

[0181] The final output y is:

[0182] ;

[0183] If the gas concentration is predicted to exceed 80% of the normal upper limit, the alarm module will send a pre-warning signal to the communication module. The pre-warning signal includes information such as the predicted concentration value, the expected time of exceeding the standard, and the potential hazard level, reminding staff to take precautions in advance. At the same time, the system will automatically generate prevention suggestions, such as increasing the ventilation frequency of the area and arranging personnel to conduct on-site inspections. The prevention suggestions are generated based on effective handling measures for similar situations in historical data.

[0184] The embodiments disclosed in this invention are preferred embodiments, but are not limited thereto. Those skilled in the art can easily understand the spirit of this invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of this invention, they are all within the protection scope of this invention.

Claims

1. A coal mine gas concentration monitoring and early warning system, characterized in that, include: The monitoring module is used to monitor the temperature, humidity, oxygen concentration, and gas concentration of the environment in real time, and transmit the temperature, humidity, oxygen concentration, and gas concentration values, as well as the location data of the monitoring node, to the data processing module. The adaptive compensation module is used to receive the real-time monitoring values ​​of temperature, humidity, oxygen concentration and gas concentration transmitted by the data processing module from the monitoring node, and to correct them to obtain the standard values ​​of temperature, humidity, oxygen concentration and gas concentration. The data processing module is used to determine whether at least two of the standard values ​​for temperature, humidity, oxygen concentration, and methane concentration are abnormal. If so, it controls the alarm communication module to issue a warning signal and sends the warning signal to the receiving end; if not, it further determines whether there is a methane concentration hazard signal. The specific method used by the data processing module to determine the methane concentration hazard signal is as follows: Determine whether the temperature, humidity, oxygen concentration, and methane concentration exceed the normal values ​​for temperature, humidity, oxygen concentration, and methane concentration, respectively. The 3σ principle is used for judgment. When the real-time monitoring value exceeds the range of "mean ± 3 times standard deviation of the corresponding parameter in normal range", the monitoring value is judged to be outside the normal range. If the standard values ​​for temperature, humidity, oxygen concentration, or gas concentration exceed the normal values ​​for temperature, humidity, oxygen concentration, or gas concentration, then the concentration at the time of detection is a gas concentration danger signal. Otherwise, by further calculating the product of temperature and humidity and comparing it with a preset value, it can be determined whether there is an abnormality in sensor data caused by the combined effect of temperature and humidity. The gas concentration hazard signal indicates that the detection location at the time of detection is a hazardous area; The alarm communication module is used to transmit warning signals, hazard level signals, abnormal concentration signals, or normal gas concentration signals when the corresponding conditions are met. The receiving end is used to receive signals sent by the alarm communication module and take corresponding measures based on the signal content; The adaptive compensation module corrects the temperature, humidity, oxygen concentration, and gas concentration values ​​received by the data processing module from the monitoring nodes in real time as follows: The calibration procedure is set according to the type of sensor. For temperature sensors, the Steinhart-Hart equation based on platinum resistance is used for nonlinear compensation. The equation is as follows: ； in, This is the temperature value. The resistance value of the platinum resistance thermometer is... , , These are the equation parameters; A second correction is then made based on environmental pressure data; the correction formula is as follows: ； in, For the corrected temperature value, To measure the temperature value, This is the pressure correction factor. This is the pressure deviation value; For humidity sensors, a polynomial fitting algorithm is used to establish a compensation model based on the temperature-humidity cross-sensitivity characteristics. The model is a quadratic polynomial: ； in, This is the corrected humidity value. This is the temperature value. To measure humidity values, , , The humidity data was obtained by collecting humidity data at different temperatures and then fitting the data using the least squares method. When using least squares fitting, the objective function for: ； in, Indicates the first The actual humidity values ​​in the set of data, Indicates the first The humidity sensor readings in the data set; By minimizing the objective function Solve , , ; For the oxygen concentration sensor, an electrochemical sensor zero-point drift compensation algorithm is used, which dynamically corrects the zero point using a reference gas calibration coefficient. Zero-point calibration is performed every 24 hours using standard oxygen concentration gas, and the calibration coefficient is updated after calibration. The corrected formula is: ； in, This is the corrected oxygen concentration value. To measure oxygen concentration, This is the zero-point drift value; For gas concentration sensors, a combined algorithm of temperature compensation and aging correction is used for catalytic combustion sensors. Temperature compensation is performed linearly based on data from the sensor's built-in temperature sensor, as shown in the formula: ； in, This is the gas concentration value after temperature compensation. To measure the gas concentration value, This is the temperature compensation coefficient. This is a reference temperature value. Aging correction is achieved by recording sensor usage time. and cumulative number of tests An aging curve model is established for correction, and the aging correction formula is as follows: ； in, This is the gas concentration value after aging correction. This is the aging correction factor. Design the usage time for the sensor. Design the cumulative number of detections for the sensor; The original data on temperature, humidity, oxygen concentration, and gas concentration are corrected using the aforementioned correction procedure to obtain the corresponding standard values.

2. The coal mine gas concentration monitoring and early warning system according to claim 1, characterized in that, When the data processing module determines whether there is a gas concentration danger signal, it compares the real-time monitored temperature, humidity, oxygen concentration and gas concentration values ​​with the normal temperature, normal humidity, normal oxygen concentration and normal gas concentration values. The system automatically determines the normal temperature, humidity, oxygen concentration, and gas concentration values ​​as follows: The data processing module uses a big data analysis algorithm that combines time series decomposition and machine learning based on historical monitoring data to decompose the monitoring data into time dimensions of hours, days, and months, and extract the variation characteristics of temperature, humidity, oxygen concentration, and gas concentration at different time scales.

3. The coal mine gas concentration monitoring and early warning system according to claim 1, characterized in that, The data processing module counts the number of abnormal sensor output data caused by the combined effects of temperature and humidity that constitute a gas concentration hazard signal. If the count is greater than or equal to 2, the concentration at the detection time is determined to be a gas concentration hazard signal. The gas concentration hazard signal is divided into three levels: minor hazard, moderate hazard, and major hazard. Minor danger refers to a situation where only one parameter exceeds the normal range and no combined temperature and humidity anomalies are triggered; Moderate risk refers to the presence of 2-3 parameters exceeding the normal range or the triggering of combined abnormalities in temperature and humidity; Major danger refers to all four parameters exceeding the normal range or two consecutive moderate danger signals appearing in the same area within a short period of time; Different levels of danger correspond to different graded early warning and response strategies, with audible and visual warnings used for minor dangers; Initiate enhanced local ventilation when the risk level is moderate. A major hazard triggers an emergency evacuation order for the entire area, and at the same time sends evacuation route information to all personnel positioning equipment underground. The evacuation routes are dynamically generated based on the real-time monitoring of the hazard situation in each area.

4. The coal mine gas concentration monitoring and early warning system according to claim 1, characterized in that, The monitoring module uses edge computing nodes to perform local data preprocessing. The edge computing nodes have a built-in FPGA chip and ARM processor collaborative processing architecture. The FPGA chip is responsible for high-speed data acquisition and preliminary processing, while the ARM processor performs complex algorithm calculations to reduce noise, normalize, and compress the data volume of the acquired raw data.

5. The coal mine gas concentration monitoring and early warning system according to claim 1, characterized in that, The data processing module adopts a distributed computing architecture, including multiple data processing sub-nodes and a management node. The management node is based on Kubernetes container orchestration technology and dynamically allocates tasks to each sub-node according to the data type and computing workload.

6. The coal mine gas concentration monitoring and early warning system according to claim 1, characterized in that, A multi-parameter time series correlation model is established. The multi-parameter time series correlation model has a built-in environmental perception module. The environmental perception module is connected to a ground radar, vibration sensor and equipment status monitor through an RS485 bus to collect environmental information such as geological structure data, mining equipment operating parameters and ventilation system status of the monitored area in real time as model input variables. The collected data is first converted and validated to ensure its integrity and accuracy before being input into the model; An online learning mechanism is set up to automatically trigger the model update process when the model prediction results deviate from the actual monitoring data; Using the latest environmental information and monitoring data, the stochastic gradient descent algorithm is employed to dynamically adjust the model parameters, with the adjustment step size adaptively adjusted according to the rate of change of the data.

7. A coal mine gas concentration monitoring and early warning system according to claim 6, characterized in that, The online learning mechanism employs a transfer learning algorithm, which rapidly adjusts the parameters of the training model based on existing similar scenarios when new scenarios or working conditions emerge. The domain adaptation algorithm aligns the feature distributions of the source and target domains, reducing model training time and data requirements.

8. The coal mine gas concentration monitoring and early warning system according to claim 1, characterized in that, The data processing module uses a long short-term memory network combined with an attention mechanism to construct a prediction model to predict the development trend of abnormal gas concentration.

Images