An intelligent monitoring system for coal mine power supply based on Internet of Things

Abstract

The application discloses a coal mine power supply intelligent monitoring system based on Internet of Things, and belongs to the field of coal mine power supply monitoring. In order to solve the problems of the existing coal mine power supply intelligent monitoring system, such as response lag, high false alarm rate and high artificial dependence, the application is provided with an end-side global perception module, a data first analysis module, an edge data processing module, a data transmission module, a cloud data analysis and model construction module and a fault early warning and closed-loop control module. Through the deployment of multiple types of intelligent sensors, the device state is collected in real time. The local data preprocessing and abnormal prediction are carried out in combination with the edge computing node. The time series data correlation analysis algorithm is used to establish the device health degree model. Through the multi-source data fusion analysis, the overload prediction, the electric leakage positioning and the energy consumption optimization closed-loop control are realized. Finally, the three-level intelligent monitoring system of "end-side perception-edge calculation-cloud decision" is formed. The system response efficiency and accuracy are improved, and the personal labor intensity is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coal mine power supply monitoring, in particular to a coal mine power supply intelligent monitoring system based on the Internet of Things. BACKGROUND

[0002] The construction goal of the smart grid is the highest stage of coal mine power grid construction, which conforms to the construction direction of the current smart mine. Although the unattended construction of the coal mine power supply system has realized the efficient operation of the power grid, the coal mine still faces many challenges, mainly in the following aspects: system maintenance problem: the coal mine power grid is complex and changeable, and the switch equipment, mechanical and electrical equipment, etc. have migration, and the underground operation often causes damage to the power cable, mechanical and electrical equipment and control network, etc. The maintenance of the whole system becomes extremely difficult, especially under the situation of "labor shortage" and limited number of people going down the well, the annual switch equipment inspection is a challenge. Economic operation of the power grid: coal mine is also a major energy consumer, how to use the large amount of data collected by the unattended power grid system and other automation systems for further mining, calculate the energy consumption account, find energy-saving space and energy-saving methods, and make the coal mine more energy-efficient is also a problem that must be paid attention to. Therefore, it is urgent to build a mine intelligent power supply cloud network integrated management and control system platform to ensure the intelligentization of mine power supply and realize the essential safety of high-voltage power supply.

[0003] In the Chinese patent with publication number CN119298404B, an intelligent monitoring system for coal mine underground power supply is proposed. After reaching the corresponding set monitoring time interval, the running parameters of the corresponding transformer in the set time period are collected and analyzed to obtain the electrical hidden danger index, thermal state hidden danger index and abnormal sound hidden danger index of each group of transformers in the set time period. The electrical parameters, temperature parameters and sound parameters of the transformer can be comprehensively monitored, overcoming the limitation of traditional monitoring methods that only focus on basic electrical parameters, thereby more accurately evaluating the health status of the transformer, providing intuitive quantitative indicators for maintenance personnel, and helping to timely discover potential problems;

[0004] In the prior art, it is difficult to deploy multiple types of intelligent sensors, it is impossible to grasp the real-time state information of each type of equipment, it is also impossible to timely grasp whether each type of equipment is overloaded or whether there is a leakage problem, it is also difficult to optimize the energy consumption of each type of equipment, and it is impossible to form a three-level intelligent monitoring system of perception-computation-decision, resulting in problems of system response lag, high false alarm rate and high dependence on manual operation.

[0005] Therefore, we propose a coal mine power supply intelligent monitoring system based on the Internet of Things. SUMMARY

[0006] The present application aims to provide an intelligent monitoring system for coal mine power supply based on Internet of Things, which solves the problems of response lag, high false alarm rate and high artificial dependence of the intelligent monitoring system for coal mine power supply in the background art.

[0007] To achieve the above object, the present application provides the following technical solution: an intelligent monitoring system for coal mine power supply based on Internet of Things, comprising:

[0008] an end-side global perception module, configured to:

[0009] deploying current sensors, voltage sensors, temperature sensors and vibration sensors on transformers, cables, switch cabinets and motors of the coal mine power supply network to collect real-time operating current, voltage, surface temperature and vibration frequency data of the equipment;

[0010] a data-first analysis module, configured to:

[0011] controlling the timestamp error of current, voltage, temperature and vibration within 10ms through the NTP protocol of the edge gateway, performing down-sampling on high-frequency vibration data, aligning the data with temperature and current data according to the time axis, forming a device state matrix, extracting fault correlation features, and finally adopting a hybrid decision model of "rules + machine learning" to construct a fault judgment logic;

[0012] an edge data processing module, configured to:

[0013] receiving the data collected by the sensors through the edge computing node, filtering and denoising the data for preprocessing, performing abnormal prediction based on the preset normal operating parameter threshold of the equipment, and screening out abnormal data exceeding the threshold range;

[0014] a data transmission module, configured to:

[0015] uploading the normal operating data and abnormal data processed by the edge computing node to the cloud platform through a dual-channel composed of a 5G communication module and an optical fiber transmission channel;

[0016] a cloud data analysis and model construction module, configured to:

[0017] using a time series data correlation analysis algorithm to correlate and model the received equipment operating data according to the time sequence, establishing a device health degree evaluation model, wherein the health degree evaluation model takes the equipment operating time, the number of historical faults and the real-time parameter fluctuation amplitude as input parameters;

[0018] a fault early warning and closed-loop control module, configured to:

[0019] Based on the aforementioned health assessment model, and combined with the preset "fault feature map library" and "dynamic threshold early warning" dual mechanisms, the "fault feature map library" stores the current-voltage-temperature correlation feature maps corresponding to typical faults of coal mine power supply equipment, and the "dynamic threshold early warning" dynamically adjusts the parameter early warning threshold according to the cumulative running time of the equipment; through multi-source data fusion analysis, overload prediction, leakage location and energy consumption optimization are realized, and control commands are generated and fed back to the control module of the coal mine power supply equipment to form a closed-loop control.

[0020] Furthermore, the end-side global sensing module includes a transformer monitoring unit, a cable monitoring unit, a switchgear monitoring unit, and a motor monitoring unit, wherein:

[0021] The transformer monitoring unit is used to deploy partial discharge sensors and oil-immersed fiber optic grating temperature sensors to simultaneously collect winding temperature and partial discharge quantity.

[0022] The cable monitoring unit uses distributed fiber optic temperature sensors and current transformers to achieve distributed monitoring of cable sheath temperature and high-frequency pulse current acquisition.

[0023] The switchgear monitoring unit integrates an infrared thermal imager and a gas density sensor to simultaneously acquire the temperature field distribution and insulating gas status of the equipment inside the cabinet.

[0024] The motor monitoring unit is used to install a triaxial vibration acceleration sensor and a shaft current sensor to collect the motor vibration spectrum and shaft voltage signal.

[0025] Furthermore, the edge data processing module includes an adaptive multimodal filtering module and an adaptive threshold module. The adaptive multimodal filtering module automatically identifies sensor types and matches corresponding filtering strategies. It has a built-in noise feature library that stores typical noise spectra of current, temperature, and vibration acceleration sensors. By analyzing the noise spectrum of the input data through Fast Fourier Transform, it calls the matching filtering algorithm from the feature library, improving the signal-to-noise ratio of the processed data by more than 20% and effectively preserving fault characteristic signals. The adaptive threshold module identifies whether the equipment is in a startup, running, or shutdown state by analyzing the characteristics of current change rate and voltage stability. It presets threshold ranges for different operating conditions and analyzes the equipment's operating data weekly, updating the threshold range for each operating condition using the 3σ principle, thereby reducing the misjudgment rate caused by changes in operating conditions.

[0026] Furthermore, the edge data processing module also includes an edge resource dynamic scheduling module and an edge node redundancy backup module, wherein:

[0027] The edge resource dynamic scheduling module is used to classify device parameters according to their importance. High-priority data will occupy computing power first, while medium-priority data will have its processing frequency reduced. When the node CPU utilization rate is greater than 80%, some non-critical data will be migrated to idle edge nodes for processing, thereby optimizing the allocation of computing power and storage resources of edge nodes.

[0028] The edge node redundancy backup module is used to deploy two edge nodes in each region, one master node and one slave node. The slave node synchronizes the configuration and data of the master node in real time. When the master node fails, the slave node automatically takes over the sensor data reception and processing tasks. After the node recovers, it automatically synchronizes the data during the failure period from the backup node to ensure data continuity.

[0029] Furthermore, the data transmission module includes an intelligent switching control module, a dynamic bandwidth adaptation module, and a fiber optic link self-healing module, wherein:

[0030] The intelligent switching control module monitors the dual-channel status in real time, enabling seamless and rapid switching to avoid data loss. It collects signal strength and signal-to-noise ratio data for the 5G channel and optical power and bit error rate data for the fiber optic channel every second, establishing a channel health scoring model with a maximum score of 100. 30 points are deducted when signal strength is < -100dBm and bit error rate is > 0. Deduct 40 points. When the health of a certain channel is less than 60 points, start the backup channel to warm up, shorten the switching preparation time, and adopt the "connect first and then disconnect" mechanism. During the switching, the current data is transmitted through the backup channel first. After confirming that the data packets are continuously received, the original channel is disconnected. The switching delay is controlled within 50ms.

[0031] The dynamic bandwidth adaptation module is used to adjust the data transmission strategy according to the real-time bandwidth of 5G to avoid congestion. The 5G communication module reports the current available bandwidth every second. The edge node establishes a bandwidth prediction model and divides the data into three levels: emergency data, important data and ordinary data. When the predicted bandwidth is less than the current transmission demand, priority is given to ensuring the transmission of emergency and important data. Ordinary data is temporarily stored at the edge node and uploaded in batches after the bandwidth is restored.

[0032] The fiber optic link self-healing module has a built-in optical power detection circuit. When the received optical power is less than -25dBm (normal range -1 to -20dBm), it is determined to be a link fault. At the same time, the fault point is located. If the fault point is located in a branch link, it automatically switches to the backup fiber. When the fiber loss suddenly increases by more than 5dB, it is a sign of impending breakage and immediately pushes an early warning to the operation and maintenance terminal to prompt manual inspection.

[0033] Furthermore, the cloud-based data analysis and model building module includes a dynamic weight optimization module, a personalized model fine-tuning module, and an evaluation result tracing and visualization module, wherein:

[0034] The dynamic weight optimization module is used to adaptively adjust the weight of input parameters according to the equipment type and operation stage. It presets the base weight for different equipment, reduces the weight of "historical failure count" and increases the weight of "real-time parameter fluctuation" for new equipment, and increases the weight of "running time" for older equipment. The module also calculates the consistency between the model evaluation results and actual failures every month and fine-tunes the weights accordingly.

[0035] The personalized model fine-tuning module fine-tunes the general model based on individual device data to improve the accuracy of individual assessments. Specifically, it establishes an initial health baseline for each device as an assessment benchmark, uses the general model as a basis, and performs incremental training with the device's operating data from the past year. While retaining general features, it learns individual characteristics. When the assessment result of an individual device deviates from the actual state by more than 15%, fine-tuning is automatically triggered.

[0036] The assessment results traceability and visualization module is used to analyze the constituent factors of the health score, display key influencing factors in a visual way, break down the health score according to the input parameters, clarify the contribution of each factor, generate a linkage chart of "health trend + key parameter curve", and output suggestions based on the dominant influencing factors to improve the efficiency of operation and maintenance personnel in understanding the health assessment results.

[0037] Furthermore, the fault early warning and closed-loop control module includes a dual-mechanism confidence fusion module and an instruction priority scheduling and execution monitoring module, wherein:

[0038] The dual-mechanism confidence fusion module is used to resolve conflicts by quantifying confidence and output a unique decision. It calculates the confidence for the two mechanisms separately: the matching degree of the map library and the threshold exceedance range. Then, it introduces the historical accuracy weight. The comprehensive confidence = (map confidence × map library accuracy weight) + (threshold confidence × dynamic threshold accuracy weight). ≥80% triggers an alert, 50%-80% enters the observation period, and <50% is judged as a false alarm.

[0039] The instruction priority scheduling and execution monitoring module is used to optimize instruction execution efficiency, avoid excessive control, and classify instructions by priority.

[0040] Furthermore, the adaptive threshold module also includes a load intensity classification judgment module. Based on the identification of the equipment start-up / running / stopping state, it further subdivides the "running state" into three load intensity levels: light load, full load, and overload. By calculating the ratio of the effective value of the current to the rated value in real time, it triggers the threshold range of the corresponding load level. In the light load state, the temperature threshold is relaxed by 15%, and in the overload state, the vibration threshold is tightened by 20%. A load-temperature-vibration correlation matrix is ​​established. When the current suddenly increases by 10% and the temperature rise rate is >2℃ / min, it automatically switches from the current load level threshold range to a more stringent temporary threshold.

[0041] Furthermore, the dynamic weight optimization module also includes a fault mode association weight module, which adds a fault mode association factor to the input parameters of equipment health assessment. When the proportion of "overheating fault" in historical fault records is >60%, the weight of real-time parameter fluctuation amplitude is increased by 15%; when the proportion of "mechanical fault" is >60%, the weight of runtime is increased by 20%. By constructing a fault mode-parameter influence matrix and using the analytic hierarchy process to calculate the influence coefficient, the weight is dynamically adapted to the evolution of equipment fault characteristics. The matrix coefficients are recalculated monthly based on newly added fault data.

[0042] Furthermore, the intelligent switching control module also includes a service scenario priority adaptation module, which introduces service scenario correction parameters into the channel health scoring model. When transmitting fault warning data, the signal strength weight of the 5G channel is reduced by 10% and the latency weight is increased by 20%. When transmitting historical data backup, the bandwidth stability weight of the fiber optic channel is increased by 30%. A scenario-parameter mapping table is established, which includes three scenarios: fault warning, daily monitoring, and data backup. By identifying the service identifier field in the data frame, the corresponding weight calculation strategy is automatically called, and the scenario switching response time is ≤10ms.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] This invention proposes an IoT-based intelligent monitoring system for coal mine power supply. Existing intelligent monitoring systems for coal mine power supply suffer from slow response, high false alarm rates, and heavy reliance on manual intervention. This invention, however, utilizes a multi-sensor-based system, comprising a global edge sensing module, a data pre-analysis module, an edge data processing module, a data transmission module, a cloud-based data analysis and model building module, and a fault early warning and closed-loop control module. It achieves real-time data acquisition of equipment status through the deployment of multiple types of intelligent sensors, performs local data preprocessing and anomaly prediction using edge computing nodes, and uploads the data to the cloud platform via 5G / fiber optic dual channels. A time-series data correlation analysis algorithm is used to establish an equipment health model. The system focuses on developing a dual mechanism of "fault feature map library" and "dynamic threshold early warning." Through multi-source data fusion analysis, it achieves closed-loop control for overload prediction, leakage current location, and energy consumption optimization, ultimately forming a three-level intelligent monitoring system of "edge sensing - edge computing - cloud decision-making," improving system response efficiency and accuracy while reducing individual workload. Attached Figure Description

[0045] Figure 1 This is the overall program flowchart of the IoT-based intelligent monitoring system for coal mine power supply of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] To address the technical challenges of improving system response efficiency and accuracy, such as... Figure 1 As shown, the following preferred technical solutions are provided:

[0048] An IoT-based intelligent monitoring system for coal mine power supply includes:

[0049] The edge-side global perception module is used for:

[0050] Current sensors, voltage sensors, temperature sensors, and vibration sensors are deployed on transformers, cables, switchgear, and motors in the coal mine power supply network to collect real-time operating current, voltage, surface temperature, and vibration frequency data of the equipment.

[0051] The data-first analysis module is used for:

[0052] By using the NTP protocol of the edge gateway, the timestamp error of current, voltage, temperature and vibration is controlled within 10ms. High-frequency vibration data is downsampled and aligned with temperature and current data along the time axis to form a device state matrix. Then, fault correlation features are extracted. Finally, a hybrid decision model of "rules + machine learning" is used to construct fault judgment logic.

[0053] Specifically, taking electric motors as an example, for common motor faults (bearing wear, rotor bar breakage, overload), the correlation features of multiple sensors are extracted:

[0054] Vibration characteristics: When the bearing fails, the vibration signal shows a peak in the 100-500Hz frequency band; when the rotor fails, a sideband appears near twice the power supply frequency.

[0055] Current characteristics: Rotor bar breakage will cause a 2sf harmonic component in the current signal, where sf is the slip frequency. Under overload, the effective value of the current will continuously exceed 1.2 times the rated value.

[0056] Temperature characteristics: Mechanical failures, such as bearing jamming, can cause the temperature to rise in a "stepwise" manner, increasing by more than 2°C every 5 minutes; electrical failures, such as inter-turn short circuits, can cause the temperature to rise a "sudden" manner, increasing by more than 5°C within 1 minute.

[0057] The above feature values ​​are calculated using a sliding window (such as a 10-second window) to form a feature vector;

[0058] Pre-defined multi-sensor association rules can quickly eliminate obvious false alarms.

[0059] If the vibration sensor detects an abnormality, but the current and temperature are within the normal range, and the vibration signal has no obvious frequency characteristics, such as random noise, it is determined that the sensor is loosely installed and is a false alarm.

[0060] If the temperature suddenly rises by more than 5°C, but the current is normal and the vibration is stable, it is determined that the temperature sensor is faulty.

[0061] If the current continues to exceed the limit, >1.2 times the rated value, and the temperature rises linearly, increasing by 3°C every 10 minutes, but the vibration is normal, it is determined that the equipment is overloaded and is a true fault.

[0062] For complex scenarios that the rule base cannot determine, such as multiple coupled faults, use historical fault data to train a classification model:

[0063] Input: Vibration characteristics of peak values ​​in 3 frequency bands, harmonic component current characteristics, temperature characteristics of temperature rise rate, a total of 6 dimensions;

[0064] Model: Random forest is used, and the fault type and confidence level are output.

[0065] Threshold determination: When the model output confidence level is greater than 85% and at least two sensor features conform to the fault pattern, it is determined to be a true fault.

[0066] This multi-layered fusion strategy can reduce the false alarm rate of motor faults by more than 60%, while keeping the false alarm rate below 0.5%, thereby improving the accuracy of system operation.

[0067] Edge data processing module, used for:

[0068] The edge computing node receives data collected by the sensor, performs filtering and noise reduction preprocessing on the data, and makes anomaly prediction based on preset normal operating parameter thresholds of the device, filtering out abnormal data that exceeds the threshold range.

[0069] Data transmission module, used for:

[0070] Normal operation data and abnormal data processed by edge computing nodes are uploaded to the cloud platform through a dual channel consisting of a 5G communication module and an optical fiber transmission channel.

[0071] The cloud-based data analysis and model building module is used for:

[0072] Using a cloud platform and a time-series data association analysis algorithm, the received equipment operation data is correlated and modeled according to the time series to establish an equipment health assessment model. The health assessment model takes equipment runtime, historical failure count, and real-time parameter fluctuation amplitude as input parameters.

[0073] The fault warning and closed-loop control module is used for:

[0074] Based on the aforementioned health assessment model, and combined with the preset "fault feature map library" and "dynamic threshold early warning" dual mechanisms, the "fault feature map library" stores the current-voltage-temperature correlation feature maps corresponding to typical faults of coal mine power supply equipment, and the "dynamic threshold early warning" dynamically adjusts the parameter early warning threshold according to the cumulative running time of the equipment; through multi-source data fusion analysis, overload prediction, leakage location and energy consumption optimization are realized, and control commands are generated and fed back to the control module of the coal mine power supply equipment to form a closed-loop control.

[0075] The end-side global sensing module includes a transformer monitoring unit, a cable monitoring unit, a switchgear monitoring unit, and a motor monitoring unit, among which:

[0076] The transformer monitoring unit is used to deploy partial discharge sensors and oil-immersed fiber Bragg grating temperature sensors to simultaneously collect winding temperature and partial discharge quantity. The partial discharge sensor has a detection sensitivity of ≤5pC, and the oil-immersed fiber Bragg grating temperature sensor has a temperature measurement range of -50~120℃ and an accuracy of ±0.5℃.

[0077] The cable monitoring unit uses a distributed fiber optic temperature sensor and a current transformer to achieve distributed monitoring of cable sheath temperature and high-frequency pulse current acquisition. The temperature sensor has a spatial resolution of 1m and a sampling interval of 0.5s, while the current transformer has a bandwidth of 10kHz~1MHz.

[0078] The switchgear monitoring unit integrates an infrared thermal imager and a gas density sensor to simultaneously acquire the temperature field distribution and insulating gas status of the equipment inside the cabinet. The integrated infrared thermal imager has a temperature measurement accuracy of ±2℃, and the gas density sensor has a measurement range of 0~1.0MPa.

[0079] The motor monitoring unit is used to install a triaxial vibration acceleration sensor and a shaft current sensor to collect the motor vibration spectrum and shaft voltage signal. The vibration acceleration sensor has a range of ±50g and a frequency range of 0.1Hz~10kHz. A waterproof sealing box is added to the sensor on the cable, and a magnetic oil-proof shell is used for the triaxial vibration acceleration sensor to avoid direct contact with oil.

[0080] The edge data processing module includes an adaptive multimodal filtering module and an adaptive threshold module. The adaptive multimodal filtering module automatically identifies sensor types and matches corresponding filtering strategies. It has a built-in noise feature library that stores typical noise spectra of current, temperature, and vibration acceleration sensors. By analyzing the noise spectrum of the input data through Fast Fourier Transform, it calls the matching filtering algorithm from the feature library, improving the signal-to-noise ratio of the processed data by more than 20% and effectively preserving fault characteristic signals. The adaptive threshold module identifies whether the equipment is in the start-up, running, or shutdown state by analyzing the characteristics of current change rate and voltage stability. It presets threshold ranges for different operating conditions and analyzes the equipment's operating data weekly, updating the threshold range for each operating condition using the 3σ principle, thereby reducing the misjudgment rate caused by changes in operating conditions.

[0081] The edge data processing module also includes an edge resource dynamic scheduling module and an edge node redundancy backup module, wherein:

[0082] The edge resource dynamic scheduling module is used to classify device parameters according to their importance. High-priority data will occupy computing power first, while medium-priority data will have its processing frequency reduced. When the node CPU utilization rate is greater than 80%, some non-critical data will be migrated to idle edge nodes for processing, thereby optimizing the allocation of computing power and storage resources of edge nodes.

[0083] The edge node redundancy backup module is used to deploy two edge nodes in each region, one master node and one slave node. The slave node synchronizes the configuration and data of the master node in real time. When the master node fails, the slave node automatically takes over the sensor data reception and processing tasks. After the node recovers, it automatically synchronizes the data during the failure period from the backup node to ensure data continuity and ensure that data processing is not interrupted when the edge node fails.

[0084] The data transmission module includes an intelligent switching control module, a dynamic bandwidth adaptation module, and a fiber optic link self-healing module, wherein:

[0085] The intelligent switching control module monitors the dual-channel status in real time, enabling seamless and rapid switching to avoid data loss. It collects signal strength and signal-to-noise ratio data for the 5G channel and optical power and bit error rate data for the fiber optic channel every second, establishing a channel health scoring model with a maximum score of 100. 30 points are deducted when signal strength is < -100dBm and bit error rate is > 0. Deduct 40 points. When the health of a certain channel is less than 60 points, start the backup channel to warm up, shorten the switching preparation time, and adopt the "connect first and then disconnect" mechanism. During the switching, the current data is transmitted through the backup channel first. After confirming that the data packets are continuously received, the original channel is disconnected. The switching delay is controlled within 50ms.

[0086] The dynamic bandwidth adaptation module is used to adjust the data transmission strategy according to the real-time bandwidth of 5G to avoid congestion. The 5G communication module reports the current available bandwidth every second. The edge node establishes a bandwidth prediction model and divides the data into three levels: emergency data, important data, and ordinary data. When the predicted bandwidth is less than the current transmission demand, priority is given to ensuring the transmission of emergency and important data. Ordinary data is temporarily stored at the edge node and uploaded in batches after the bandwidth is restored, thereby reducing the data congestion rate of the 5G channel and reducing the latency of emergency data transmission.

[0087] The fiber optic link self-healing module has a built-in optical power detection circuit. When the received optical power is less than -25dBm (normal range -1 to -20dBm), it is determined to be a link fault. At the same time, the fault point is located. If the fault point is located in a branch link, it automatically switches to the backup fiber. When the fiber loss suddenly increases by more than 5dB, it is a sign of impending breakage and immediately pushes an early warning to the operation and maintenance terminal to prompt manual inspection, thereby improving the automatic repair time of fiber optic faults.

[0088] The cloud-based data analysis and model building module includes a dynamic weight optimization module, a personalized model fine-tuning module, and an evaluation result traceability and visualization module, among which:

[0089] The dynamic weight optimization module is used to adaptively adjust the weight of input parameters according to the equipment type and operation stage. It presets the base weight of different equipment. For new equipment, the weight of "historical failure count" is reduced and the weight of "real-time parameter fluctuation" is increased. For older equipment, the weight of "running time" is increased. The module also calculates the consistency between the model evaluation results and actual failures every month and fine-tunes the weights to improve the accuracy of health assessment.

[0090] The personalized model fine-tuning module fine-tunes the general model based on individual device data to improve the accuracy of individual assessments. Specifically, it establishes an initial health baseline for each device as an assessment benchmark. Based on the general model, it incrementally trains the model using the device's operating data from the past year, retaining general features while learning individual characteristics. When the assessment result of an individual device deviates from the actual state by more than 15%, fine-tuning is automatically triggered to further improve the accuracy of health prediction.

[0091] The assessment results traceability and visualization module is used to analyze the constituent factors of the health score, display key influencing factors in a visual way, break down the health score according to the input parameters, clarify the contribution of each factor, generate a linkage chart of "health trend + key parameter curve", and output suggestions based on the dominant influencing factors to improve the efficiency of operation and maintenance personnel in understanding the health assessment results.

[0092] The fault early warning and closed-loop control module includes a dual-mechanism confidence fusion module and an instruction priority scheduling and execution monitoring module, wherein:

[0093] The dual-mechanism confidence fusion module is used to resolve conflicts by quantifying confidence and output a unique decision. It calculates confidence for the two mechanisms separately: the matching degree of the map library and the threshold exceedance range. Then, it introduces the historical accuracy weight. The comprehensive confidence = (map confidence × map library accuracy weight) + (threshold confidence × dynamic threshold accuracy weight). ≥80% triggers an alert, 50%-80% enters the observation period, and <50% is judged as a false alarm, thus reducing the false alarm rate.

[0094] The instruction priority scheduling and execution monitoring module is used to optimize instruction execution efficiency and avoid excessive control. Instructions are prioritized and classified as follows: Level 1 instructions (such as leakage current, short circuit): executed immediately, and transmitted using the 5G emergency channel; Level 2 instructions (such as minor overload): executed with a 3-second delay, allowing manual intervention; Level 3 instructions (such as energy consumption optimization): non-urgent, executed in batches, and for the same type of instruction on the same device, a "cooling-off period" is set (e.g., Level 1 instructions have an interval of ≥30 seconds, Level 2 instructions have an interval of ≥5 minutes). During the cooling-off period, repeated warnings only update the status and do not issue instructions. The device returns an "execution success / failure" signal within 1 second after execution. If it fails, it automatically switches to a backup instruction.

[0095] The adaptive threshold module also includes a load intensity classification judgment module. Based on the identification of the equipment's start-up / running / stopping status, it further subdivides the "running status" into three load intensity levels: light load, full load, and overload. By calculating the ratio of the effective current value to the rated value in real time, it triggers the corresponding load level threshold range. In the light load state, the temperature threshold is relaxed by 15%, and in the overload state, the vibration threshold is tightened by 20%. A load-temperature-vibration correlation matrix is ​​established. When the current suddenly increases by 10% and the temperature rise rate is >2℃ / min, it automatically switches from the current load level threshold range to a more stringent temporary threshold: light load (current ≤30% of rated value), full load (30% < current ≤80% of rated value), and overload (current >80% of rated value). This solves the problem of insufficient threshold adaptability caused by load fluctuations in a single operating state, improves the accuracy of anomaly judgment under different load intensities, and especially reduces the misjudgment rate of the overload transition process after motor start-up, realizing refined threshold management in two dimensions: operating condition and load.

[0096] The dynamic weight optimization module also includes a fault mode association weight module, which adds a fault mode association factor to the input parameters of equipment health assessment. When the proportion of "overheating faults" in historical fault records is >60%, the weight of real-time parameter fluctuation amplitude is increased by 15%; when the proportion of "mechanical faults" is >60%, the weight of runtime is increased by 20%. By constructing a fault mode-parameter influence matrix and using the analytic hierarchy process to calculate the influence coefficient, the weights are dynamically adapted to the evolution of equipment fault characteristics. The matrix coefficients are recalculated monthly based on newly added fault data, which solves the limitation of traditional dynamic weights that only rely on the degree of equipment aging. This improves the sensitivity of the health assessment model to equipment-specific faults, reduces the health prediction error of old equipment to within ±5%, and provides a more accurate quantitative basis for targeted maintenance.

[0097] The intelligent switching control module also includes a service scenario priority adaptation module, which introduces service scenario correction parameters into the channel health scoring model. When transmitting fault warning data, the signal strength weight of the 5G channel is reduced by 10%, and the latency weight is increased by 20%. When transmitting historical data backup, the bandwidth stability weight of the fiber optic channel is increased by 30%. A scenario-parameter mapping table is established, which includes three scenarios: fault warning, daily monitoring, and data backup. By identifying the service identifier field in the data frame, the corresponding weight calculation strategy is automatically called. The scenario switching response time is ≤10ms, realizing deep coupling between channel switching logic and service requirements, improving the transmission integrity of fault warning data, improving the transmission efficiency of historical data backup, and avoiding the problem of insufficient service adaptability caused by general health scoring.

[0098] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A smart monitoring system for coal mine power supply based on the Internet of Things, characterized in that, include: The edge-side global perception module is used for: Current sensors, voltage sensors, temperature sensors, and vibration sensors are deployed on transformers, cables, switchgear, and motors in the coal mine power supply network to collect real-time operating current, voltage, surface temperature, and vibration frequency data of the equipment. The data-first analysis module is used for: By using the NTP protocol of the edge gateway, the timestamp error of current, voltage, temperature and vibration is controlled within 10ms. High-frequency vibration data is downsampled and aligned with temperature and current data along the time axis to form a device state matrix. Then, fault correlation features are extracted. Finally, a hybrid decision model of "rules + machine learning" is used to construct fault judgment logic. Edge data processing module, used for: The edge computing node receives data collected by the sensor, performs filtering and noise reduction preprocessing on the data, and makes anomaly prediction based on preset normal operating parameter thresholds of the device, filtering out abnormal data that exceeds the threshold range. Data transmission module, used for: Normal operation data and abnormal data processed by edge computing nodes are uploaded to the cloud platform through a dual channel consisting of a 5G communication module and an optical fiber transmission channel. The cloud-based data analysis and model building module is used for: Using a cloud platform and a time-series data association analysis algorithm, the received equipment operation data is correlated and modeled according to the time series to establish an equipment health assessment model. The health assessment model takes equipment runtime, historical failure count, and real-time parameter fluctuation amplitude as input parameters. The fault early warning and closed-loop control module is used for: Based on the aforementioned health assessment model, and combined with a pre-set dual mechanism of "fault feature map library" and "dynamic threshold early warning", the "fault feature map library" stores current-voltage-temperature correlation feature maps corresponding to typical faults of coal mine power supply equipment, and the "dynamic threshold early warning" dynamically adjusts the parameter early warning threshold according to the cumulative running time of the equipment; through multi-source data fusion analysis, overload prediction, leakage current location and energy consumption optimization are realized, and control commands are generated and fed back to the control module of coal mine power supply equipment to form a closed-loop control.

2. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 1, characterized in that: The end-side global sensing module includes a transformer monitoring unit, a cable monitoring unit, a switchgear monitoring unit, and a motor monitoring unit, among which: The transformer monitoring unit is used to deploy partial discharge sensors and oil-immersed fiber optic grating temperature sensors to simultaneously collect winding temperature and partial discharge quantity. The cable monitoring unit uses distributed fiber optic temperature sensors and current transformers to achieve distributed monitoring of cable sheath temperature and high-frequency pulse current acquisition. The switchgear monitoring unit integrates an infrared thermal imager and a gas density sensor to simultaneously acquire the temperature field distribution and insulating gas status of the equipment inside the cabinet. The motor monitoring unit is used to install a triaxial vibration acceleration sensor and a shaft current sensor to collect the motor vibration spectrum and shaft voltage signal.

3. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 2, characterized in that: The edge data processing module includes an adaptive multimodal filtering module and an adaptive threshold module. The adaptive multimodal filtering module is used to automatically identify the sensor type and match the corresponding filtering strategy. It has a built-in noise feature library that stores typical noise spectra of current, temperature, and vibration acceleration sensors. It analyzes the noise spectrum of the input data through fast Fourier transform and calls the matching filtering algorithm from the feature library. The signal-to-noise ratio of the processed data is improved by more than 20%, and fault feature signals are effectively preserved. The adaptive threshold module is used to identify whether the equipment is in the start-up, running, or shutdown state by analyzing the characteristics of current change rate and voltage stability. It presets threshold ranges for different operating conditions and analyzes the equipment operation data weekly, updating the threshold range for each operating condition using the 3σ principle, thereby reducing the misjudgment rate caused by changes in operating conditions.

4. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 3, characterized in that: The edge data processing module further includes an edge resource dynamic scheduling module and an edge node redundancy backup module, wherein: The edge resource dynamic scheduling module is used to classify device parameters according to their importance. High-priority data will occupy computing power first, while medium-priority data will have its processing frequency reduced. When the node CPU utilization rate is greater than 80%, some non-critical data will be migrated to idle edge nodes for processing, thereby optimizing the allocation of computing power and storage resources of edge nodes. The edge node redundancy backup module is used to deploy two edge nodes in each region, one master node and one slave node. The slave node synchronizes the configuration and data of the master node in real time. When the master node fails, the slave node automatically takes over the sensor data reception and processing tasks. After the node recovers, it automatically synchronizes the data during the failure period from the backup node to ensure data continuity.

5. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 1, characterized in that: The data transmission module includes an intelligent switching control module, a dynamic bandwidth adaptation module, and a fiber optic link self-healing module, wherein: The intelligent switching control module monitors the dual-channel status in real time, enabling seamless and rapid switching to avoid data loss. It collects signal strength and signal-to-noise ratio data for the 5G channel and optical power and bit error rate data for the fiber optic channel every second, establishing a channel health scoring model with a maximum score of 100. 30 points are deducted when signal strength is < -100dBm and bit error rate is > 0. Deduct 40 points. When the health of a certain channel is less than 60 points, start the backup channel to warm up, shorten the switching preparation time, and adopt the "connect first and then disconnect" mechanism. During the switching, the current data is transmitted through the backup channel first. After confirming that the data packets are continuously received, the original channel is disconnected. The switching delay is controlled within 50ms. The dynamic bandwidth adaptation module is used to adjust the data transmission strategy according to the real-time bandwidth of 5G to avoid congestion. The 5G communication module reports the current available bandwidth every second. The edge node establishes a bandwidth prediction model and divides the data into three levels: emergency data, important data and ordinary data. When the predicted bandwidth is less than the current transmission demand, priority is given to ensuring the transmission of emergency and important data. Ordinary data is temporarily stored at the edge node and uploaded in batches after the bandwidth is restored. The fiber optic link self-healing module has a built-in optical power detection circuit. When the received optical power is less than -25dBm (normal range -1 to -20dBm), it is determined to be a link fault. At the same time, the fault point is located. If the fault point is located in a branch link, it automatically switches to the backup fiber. When the fiber loss suddenly increases by more than 5dB, it is a sign of impending breakage and immediately pushes an early warning to the operation and maintenance terminal to prompt manual inspection.

6. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 1, characterized in that: The cloud-based data analysis and model building module includes a dynamic weight optimization module, a personalized model fine-tuning module, and an evaluation result traceability and visualization module, wherein: The dynamic weight optimization module is used to adaptively adjust the weight of input parameters according to the equipment type and operation stage. It presets the base weight of different equipment, reduces the weight of "historical failure count" and increases the weight of "real-time parameter fluctuation" for new equipment, and increases the weight of "running time" for older equipment. The module also calculates the consistency between the model evaluation results and actual failures every month and fine-tunes the weights accordingly. The personalized model fine-tuning module fine-tunes the general model based on individual device data to improve the accuracy of individual assessments. Specifically, it establishes an initial health baseline for each device as an assessment benchmark, uses the general model as a basis, and performs incremental training with the device's operating data from the past year. While retaining general features, it learns individual characteristics. When the assessment result of an individual device deviates from the actual state by more than 15%, fine-tuning is automatically triggered. The assessment results traceability and visualization module is used to analyze the constituent factors of the health score, display key influencing factors in a visual way, break down the health score according to the input parameters, clarify the contribution of each factor, generate a linkage chart of "health trend + key parameter curve", and output suggestions based on the dominant influencing factors to improve the efficiency of operation and maintenance personnel in understanding the health assessment results.

7. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 1, characterized in that: The fault early warning and closed-loop control module includes a dual-mechanism confidence fusion module and an instruction priority scheduling and execution monitoring module, wherein: The dual-mechanism confidence fusion module is used to resolve conflicts by quantifying confidence and output a unique decision. It calculates the confidence for the two mechanisms separately: the matching degree of the map library and the threshold exceedance range. Then, it introduces the historical accuracy weight. The comprehensive confidence = (map confidence × map library accuracy weight) + (threshold confidence × dynamic threshold accuracy weight). ≥80% triggers an alert, 50%-80% enters the observation period, and <50% is judged as a false alarm. The instruction priority scheduling and execution monitoring module is used to optimize instruction execution efficiency, avoid excessive control, and classify instructions by priority.

8. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 3, characterized in that: The adaptive threshold module also includes a load intensity classification judgment module. Based on the identification of the equipment start-up / running / shutdown status, it further subdivides the "running status" into three load intensity levels: light load, full load, and overload. By calculating the ratio of the effective value of the current to the rated value in real time, it triggers the threshold range of the corresponding load level. In the light load state, the temperature threshold is relaxed by 15%, and in the overload state, the vibration threshold is tightened by 20%. A load-temperature-vibration correlation matrix is ​​established. When the current suddenly increases by 10% and the temperature rise rate is >2℃ / min, it automatically switches from the current load level threshold range to a more stringent temporary threshold.

9. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 6, characterized in that: The dynamic weight optimization module also includes a fault mode association weight module, which adds a fault mode association factor to the input parameters of equipment health assessment. When the proportion of "overheating fault" in historical fault records is >60%, the weight of real-time parameter fluctuation amplitude is increased by 15%; when the proportion of "mechanical fault" is >60%, the weight of runtime is increased by 20%. By constructing a fault mode-parameter influence matrix and using the analytic hierarchy process to calculate the influence coefficient, the weight is dynamically adapted to the evolution of equipment fault characteristics. The matrix coefficients are recalculated monthly based on newly added fault data.

10. The intelligent monitoring system for coal mine power supply based on the Internet of Things as described in claim 5, characterized in that: The intelligent switching control module also includes a service scenario priority adaptation module, which introduces service scenario correction parameters into the channel health scoring model. When transmitting fault warning data, the signal strength weight of the 5G channel is reduced by 10% and the latency weight is increased by 20%. When transmitting historical data backups, the bandwidth stability weight of the fiber channel is increased by 30%, and a scenario-parameter mapping table is established. The parameter mapping table includes three scenarios: fault warning, daily monitoring, and data backup. By identifying the service identifier field in the data frame, the corresponding weight calculation strategy is automatically called, and the scenario switching response time is ≤10ms.

Images