Power failure monitoring system

Abstract

The invention discloses a power failure monitoring system, and relates to the technical field of failure monitoring. During operation of the system, multi-dimensional data acquisition is carried out on key equipment of a transformer substation and a power distribution room, the operation state of the equipment is comprehensively covered through real-time acquisition of data of various sensors, and data acquired by a data acquisition module is preprocessed; the intelligent gateway equipment is deployed, real-time data analysis and preliminary fault diagnosis are carried out locally, the running state of the equipment is monitored in real time through set rules and threshold values, potential faults are detected, preliminary alarm is given out, and the system automatically judges the severity of the faults and decides whether to take automatic control measures including load switching or equipment power failure or not. Meanwhile, alarm information is generated and sent to operation and maintenance personnel, data, fault diagnosis results and automatic control measures in system operation are fed back, and a deep learning model and an edge calculation strategy are periodically updated through online learning and model updating.

Description

Technical Field

[0001] The invention relates to the technical field of fault monitoring, and in particular to a power fault monitoring system. Background Art

[0002] In the operation of power fault monitoring system, substations and distribution rooms are the core links in the power network, responsible for the stable operation of power transmission and distribution. With the continuous increase of equipment operating load, the equipment in substations and distribution rooms faces more and more complex operating environment and multiple potential fault threats. Common problems include equipment overheating, abnormal current, poor contact and short circuit, which may cause equipment damage or even cause more serious accidents.

[0003] Traditional power monitoring systems mainly rely on manual inspections and regular maintenance, which is not only labor-intensive, but also cannot monitor equipment status in real time and comprehensively, and is prone to overlooking some hidden faults. Especially in key facilities such as substations and distribution rooms, equipment failures often have no obvious signs before they occur, and the types of failures are diverse, making it difficult for a single sensor to fully detect all potential problems.

[0004] In the operation of the existing power system, due to the relatively simple monitoring methods for equipment or excessive reliance on manual inspection, once a failure occurs, it often leads to equipment shutdown, power outage, and even serious safety accidents. For example, if the overheating problem is not discovered in time, it may cause the equipment to burn; poor contact or short circuit problems may cause more extensive power failures, and even cause equipment failure or fire. Traditional monitoring systems cannot efficiently integrate multiple types of sensor data, and lack sufficiently intelligent fault classification and location capabilities, resulting in the inability to identify potential problems in real time, thus affecting the overall safety and reliability of the system. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present invention provides a power fault monitoring system, which solves the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a power fault monitoring system, including a data acquisition module, a data preprocessing module, an edge computing module, a deep learning fault classification module, a fault alarm module and a feedback optimization module;

[0007] The data acquisition module is used to deploy multimodal sensors in substations and distribution rooms to collect multi-dimensional data on key equipment in substations and distribution rooms, and comprehensively cover the equipment operation status through real-time collection of multiple sensor data;

[0008] The data preprocessing module is used to preprocess the data acquired by the data acquisition module, and use multimodal data fusion technology to integrate the data from different sensors to form a complete equipment operation feature vector;

[0009] The edge computing module is used to perform real-time data analysis and preliminary fault diagnosis locally by deploying intelligent gateway devices. Through the set rules and thresholds, it monitors the operating status of the equipment in real time, detects potential faults and issues preliminary alarms. When the problem is serious, it directly transmits the data to the deep learning fault classification module through the network;

[0010] The deep learning fault classification module is used to train a deep learning model based on the historical data of the equipment, analyze the equipment operation data in real time, identify the fault type and locate the location where the fault occurs, and the model outputs the specific type and location of the fault and sends it to the fault alarm module;

[0011] The fault alarm module is responsible for automatically judging the severity of the fault according to the fault diagnosis results output by the deep learning module, and deciding whether to take automatic control measures, including load switching or equipment power off, and generating alarm information to send to the operation and maintenance personnel;

[0012] The feedback optimization module is used to provide feedback on system operation data, fault diagnosis results and automatic control measures, and regularly updates deep learning models and edge computing strategies through online learning and model updates to improve the system's intelligence level and fault diagnosis accuracy.

[0013] Preferably, the data acquisition module includes a sensor deployment unit, a data acquisition unit and a data storage unit;

[0014] The sensor deployment unit is used to deploy various types of sensors, including temperature sensors, current sensors and vibration sensors, on transformers, switchgear and busbars in substations and distribution rooms. The sensors are reasonably arranged on the surface or around the equipment according to the operating characteristics and monitoring requirements of the equipment, and various parameters of the equipment are collected in real time. The type of sensor is selected according to the operating environment and fault type of the equipment to ensure that the operating status of the equipment can be fully monitored;

[0015] The data acquisition unit is used to receive and record data from various sensors, including multi-dimensional parameters such as temperature, current, vibration and pressure, monitor and collect the operating status of the equipment in real time, ensure the real-time and accuracy of the data from each sensor, and avoid affecting the accuracy of fault diagnosis due to delayed acquisition or lost data;

[0016] The data storage unit is used to store the collected real-time data and provide a cache function. The collected data includes historical data and current equipment operation data. Through efficient storage of data, it can provide basic data support for subsequent data analysis and model training. The use of high-performance storage devices or distributed storage systems ensures that data is not lost under the requirements of high-frequency collection and fast access.

[0017] Preferably, the data acquired by the data acquisition unit includes:

[0018] The device temperature T collected by the temperature sensor, the device current I collected by the current sensor, the device voltage V collected by the voltage sensor, the device vibration intensity A collected by the vibration sensor, and the device thermal radiation value F and load condition L collected by the infrared sensor.

[0019] Preferably, the data preprocessing module includes a data cleaning unit, a data fusion unit and a feature extraction unit;

[0020] The data cleaning unit is used to clean the collected raw data, remove noise data, invalid data and outliers, automatically identify and eliminate irrelevant data through algorithms, and avoid data pollution from having a negative impact on system performance. The data cleaning algorithm is based on statistical methods and is processed through Z-score standardization and mean filtering;

[0021] The data fusion unit is used to integrate the data obtained from different sensors through multimodal data fusion technology to generate a complete equipment operation feature vector. The data fusion process uses weighted average, Kalman filtering, and principal component analysis PCA technology to combine the measurement results of different sensors to calculate and obtain the temperature current comprehensive coefficient γ1, the vibration thermal radiation comprehensive coefficient γ2, the load comprehensive coefficient γ3 and the comprehensive fault diagnosis index ξ;

[0022] The feature extraction unit is used to extract key features from the fused data to form an operation feature vector of the equipment. The feature extraction uses time series analysis and frequency domain analysis methods to extract statistical features from various sensor data, and then generate a set of concise and effective feature vectors for subsequent fault diagnosis and classification.

[0023] Preferably, the edge computing module includes a data processing unit, a fault detection unit and a data transmission unit;

[0024] The data processing unit is used to perform preliminary processing on the data transmitted to the edge computing device. The data includes temperature, current and thermal radiation device operation status data. The processing goal is to extract key information from the original data, achieve rapid response through edge computing, and reduce the load on the remote server. The data processing unit uses fast Fourier transform FFT to perform data compression and feature extraction;

[0025] The fault detection unit is used to monitor the operation status of the equipment in real time based on the set rules and thresholds, and obtain an evaluation plan by comparing the comprehensive fault diagnosis index ξ with the preset qualified threshold M. When the abnormal operation status of the equipment is monitored, an early warning is triggered through the rule engine to determine whether the equipment needs to be shut down, powered off or load switched immediately. The evaluation criteria include the type of fault, the scope of impact and the potential risk. The fault detection algorithm is based on the historical data of the equipment and combines the rule threshold to make real-time judgments, which can effectively identify potential fault problems of the equipment;

[0026] The data transmission unit is used to transmit the data processed by the edge computing module and the fault warning information to the deep learning fault classification module. The data transmission adopts a low-latency, high-bandwidth communication protocol to transmit the fault information to the cloud system in real time.

[0027] Preferably, the evaluation scheme includes:

[0028] Qualified threshold M1: The preset minimum standard for normal equipment operation status. When the fault diagnosis index ξ≤M1, it means that the equipment operation status is within the safe range and there is no major fault risk;

[0029] High-risk threshold M2: When the comprehensive fault diagnosis index ξ exceeds the qualified threshold M1 but does not reach the warning value M2 of a serious fault, the equipment status is judged to be a potential problem. When the ξ value reaches M2, the system issues a warning and focuses on monitoring and inspecting the equipment operation status.

[0030] ξ≤M1, obtain the first level assessment, the equipment has no abnormality and operates within the standard range;

[0031] M1<ξ≤M2, the second level assessment is obtained. The equipment has a potential failure risk, but it has not caused a significant impact on the operation. It is necessary to strengthen monitoring and preventive maintenance. The monitoring frequency of the equipment is increased to collect data every 10 minutes. The collection of more data points is increased to detect abnormal changes earlier. The warning value is set through the edge computing module. When ξ exceeds the set value, the operation and maintenance personnel are notified through instant messaging to ensure a quick response. Preventive inspections are carried out on the equipment regularly, focusing on temperature, current and voltage fluctuations.

[0032] ξ>M2, the third level assessment is obtained. The equipment has a serious failure risk, which leads to equipment failure. It is necessary to immediately shut down for inspection, disconnect the power supply, and start the fault handling process to avoid damage to the equipment. Through the collaborative work of the edge computing module and the deep learning fault classification module, the cause of the high-risk state is analyzed to determine whether the equipment is aging, overloaded or has poor contact. According to the fault analysis results, maintenance personnel are arranged in time to repair it, and all relevant parts of the equipment are checked to ensure that it returns to normal working condition.

[0033] Preferably, the evaluation scheme includes:

[0034] Qualified threshold M1: The preset minimum standard for normal equipment operation status. When the fault diagnosis index ξ≤M1, it means that the equipment operation status is within the safe range and there is no major fault risk;

[0035] High-risk threshold M2: When the comprehensive fault diagnosis index ξ exceeds the qualified threshold M1 but does not reach the warning value M2 of a serious fault, the equipment status is judged to be a potential problem. When the ξ value reaches M2, the system issues a warning and focuses on monitoring and inspecting the equipment operation status.

[0036] ξ≤M1, obtain the first level assessment, the equipment has no abnormality and operates within the standard range;

[0037] M1<ξ≤M2, the second level assessment is obtained. The equipment has a potential failure risk, but it has not caused a significant impact on the operation. It is necessary to strengthen monitoring and preventive maintenance. The monitoring frequency of the equipment is increased to collect data every 10 minutes. The collection of more data points is increased to detect abnormal changes earlier. The warning value is set through the edge computing module. When ξ exceeds the set value, the operation and maintenance personnel are notified through instant messaging to ensure a quick response. Preventive inspections are carried out on the equipment regularly, focusing on temperature, current and voltage fluctuations.

[0038] ξ>M2, the third level assessment is obtained. The equipment has a serious failure risk, which leads to equipment failure. It is necessary to immediately shut down for inspection, disconnect the power supply, and start the fault handling process to avoid damage to the equipment. Through the collaborative work of the edge computing module and the deep learning fault classification module, the cause of the high-risk state is analyzed to determine whether the equipment is aging, overloaded or has poor contact. According to the fault analysis results, maintenance personnel are arranged in time to repair it, and all relevant parts of the equipment are checked to ensure that it returns to normal working condition.

[0039] Preferably, the deep learning fault classification module also includes a fault location module;

[0040] The fault location unit is used to perform detailed fault location and diagnosis based on the results output by the deep learning model. The specific fault location is determined by combining the topological structure and operating status of the equipment. The results of fault location are used as a basis for decision-making to further guide operation and maintenance personnel to perform on-site maintenance.

[0041] Preferably, the fault alarm module includes an alarm generation unit and an automatic control decision unit;

[0042] The alarm generation unit is used to generate alarm information when a fault occurs, and send it to the operation and maintenance personnel in a timely manner through SMS, email and APP notification. The alarm information includes the fault type, location, severity assessment result and preliminary response measures to help the operation and maintenance personnel respond quickly;

[0043] The automatic control decision unit is used to determine whether automatic control measures need to be taken according to the fault severity assessment and alarm results, including switching loads or power off, and to avoid equipment damage or expansion of faults by executing control instructions.

[0044] Preferably, the feedback optimization module includes a data feedback unit, a model optimization unit and a strategy optimization unit;

[0045] The data feedback unit is used to feed back equipment operation data, fault diagnosis results and control measures to the system to provide a basis for system optimization;

[0046] The model optimization unit is used to regularly update the deep learning model according to the feedback results, improve the fault classification and location capabilities of the model through online learning, and continuously adapt to new equipment states and fault types through incremental learning to ensure the system's adaptability and prediction capabilities for future faults;

[0047] The strategy optimization unit is responsible for optimizing the edge computing strategy and the automatic control strategy, dynamically adjusting the parameters and threshold settings of the edge computing model based on feedback data, improving the response speed and accuracy of the system in actual operation, and ensuring the efficiency of the system under different load conditions.

[0048] The present invention provides a power failure monitoring system, which has the following beneficial effects:

[0049] (1) When the system is running, multi-dimensional data collection is performed on key equipment in the substation and distribution room. Through real-time data collection from multiple sensors, the operating status of the equipment is fully covered. The data obtained by the data collection module is pre-processed. By deploying intelligent gateway devices, real-time data analysis and preliminary fault diagnosis are performed locally. Through the set rules and thresholds, the operating status of the equipment is monitored in real time, potential faults are detected and preliminary alarms are issued. The system automatically determines the severity of the fault and decides whether to take automatic control measures, including load switching or equipment power off. At the same time, an alarm message is generated and sent to the operation and maintenance personnel, and feedback is provided on the system operation data, fault diagnosis results and automatic control measures. Through online learning and model updates, the deep learning model and edge computing strategy are regularly updated.

[0050] (2) In the fault monitoring of modern power systems, the comprehensive application of multimodal data acquisition, edge computing and deep learning technology can significantly improve the accuracy and response speed of fault detection. Through the multi-sensor deployment of the data acquisition module, the system can comprehensively monitor the operating status of the equipment and collect multi-dimensional equipment data, such as temperature, current, vibration and other key indicators. The data preprocessing module integrates data from different sensors through data cleaning and multimodal data fusion technology to generate a complete operating feature vector of the equipment, providing a solid foundation for subsequent fault analysis and classification. The edge computing module can identify potential faults in a timely manner through real-time data processing and preliminary fault diagnosis, and when serious problems occur in the equipment, it can quickly transmit data to the deep learning fault classification module to ensure accurate identification and location of the fault.

[0051] (3) Compared with traditional fault monitoring technology, the current system has achieved significant improvements in real-time, accuracy and intelligence. Traditional methods usually rely on manual inspection or simple rule-based detection, which has the problems of slow response and low diagnostic accuracy. This system trains the historical data of the equipment through a deep learning model, which can analyze the operating status of the equipment in real time, accurately determine the type of fault and its location, thereby greatly improving the efficiency and accuracy of fault diagnosis. In addition, the fault alarm module can automatically determine the severity of the fault based on the results of deep learning analysis, and notify the operation and maintenance personnel in time. At the same time, it takes automatic control measures such as load switching or equipment power off, reducing the risk of equipment damage and optimizing the fault handling process.

[0052] (4) In practical applications, this power fault monitoring system not only solves the shortcomings of traditional monitoring methods, but also realizes fully automated fault detection and processing, significantly improving the stability and intelligence level of the system. Through continuous learning and updating of the feedback optimization module, the system can adapt to changes in equipment status and new fault types, maintaining a high degree of accuracy and response speed. Compared with previous fault diagnosis systems that rely on a single sensor or static rules, this system has higher flexibility and adaptability. In a complex power network environment, it can detect potential faults earlier and intervene effectively, greatly reducing downtime and economic losses caused by equipment failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A block diagram of a power fault monitoring system of the present invention; DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Example 1

[0056] The present invention provides a power failure monitoring system, please refer to Figure 1 , including data acquisition module, data preprocessing module, edge computing module, deep learning fault classification module, fault alarm module and feedback optimization module;

[0057] The data acquisition module is used to deploy multimodal sensors in substations and distribution rooms to collect multi-dimensional data on key equipment in substations and distribution rooms, and comprehensively cover the equipment operation status through real-time collection of multiple sensor data;

[0058] The data preprocessing module is used to preprocess the data acquired by the data acquisition module, and use multimodal data fusion technology to integrate the data from different sensors to form a complete equipment operation feature vector;

[0059] The edge computing module is used to perform real-time data analysis and preliminary fault diagnosis locally by deploying intelligent gateway devices. Through the set rules and thresholds, it monitors the operating status of the equipment in real time, detects potential faults and issues preliminary alarms. When the problem is serious, it directly transmits the data to the deep learning fault classification module through the network;

[0060] The deep learning fault classification module is used to train a deep learning model based on the historical data of the equipment, analyze the equipment operation data in real time, identify the fault type and locate the location where the fault occurs, and the model outputs the specific type and location of the fault and sends it to the fault alarm module;

[0061] The fault alarm module is responsible for automatically judging the severity of the fault according to the fault diagnosis results output by the deep learning module, and deciding whether to take automatic control measures, including load switching or equipment power off, and generating alarm information to send to the operation and maintenance personnel;

[0062] The feedback optimization module is used to provide feedback on system operation data, fault diagnosis results and automatic control measures, and regularly updates deep learning models and edge computing strategies through online learning and model updates to improve the system's intelligence level and fault diagnosis accuracy.

[0063] In this embodiment, multimodal sensors are deployed in substations and distribution rooms to collect multi-dimensional data of key equipment in substations and distribution rooms. Through real-time collection of data from multiple sensors, the operating status of the equipment is fully covered. The data obtained by the data acquisition module is pre-processed, and multimodal data fusion technology is used to integrate data from different sensors to form a complete equipment operation feature vector. By deploying intelligent gateway devices, real-time data analysis and preliminary fault diagnosis are performed locally. Through the set rules and thresholds, the operating status of the equipment is monitored in real time, potential faults are detected and preliminary alarms are issued. When the problem is serious, the data is directly transmitted to the deep learning fault classification model through the network. The deep learning model is trained based on the historical data of the equipment, and the equipment operation data is analyzed in real time to identify the fault type and locate the location of the fault. The model outputs the specific category and location of the fault and sends it to the fault alarm module. The system is responsible for automatically judging the severity of the fault based on the fault diagnosis results output by the deep learning module, and deciding whether to take automatic control measures, including load switching or equipment power off. At the same time, an alarm message is generated and sent to the operation and maintenance personnel, and feedback is provided on the system operation data, fault diagnosis results and automatic control measures. Through online learning and model updates, the deep learning model and edge computing strategy are regularly updated to improve the intelligence level and fault diagnosis accuracy of the system.

[0064] Example 2

[0065] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the data acquisition module includes a sensor deployment unit, a data acquisition unit and a data storage unit;

[0066] The sensor deployment unit is used to deploy various types of sensors, including temperature sensors, current sensors and vibration sensors, on transformers, switchgear and busbars in substations and distribution rooms. The sensors are reasonably arranged on the surface or around the equipment according to the operating characteristics and monitoring requirements of the equipment, and various parameters of the equipment are collected in real time. The type of sensor is selected according to the operating environment and fault type of the equipment to ensure that the operating status of the equipment can be fully monitored;

[0067] The data acquisition unit is used to receive and record data from various sensors, including multi-dimensional parameters such as temperature, current, vibration and pressure, monitor and collect the operating status of the equipment in real time, ensure the real-time and accuracy of the data from each sensor, and avoid affecting the accuracy of fault diagnosis due to delayed acquisition or lost data;

[0068] The data storage unit is used to store the collected real-time data and provide a cache function. The collected data includes historical data and current equipment operation data. Through efficient storage of data, it can provide basic data support for subsequent data analysis and model training. The use of high-performance storage devices or distributed storage systems ensures that data is not lost under the requirements of high-frequency collection and fast access.

[0069] The data acquired by the data acquisition unit includes:

[0070] The device temperature T collected by the temperature sensor, the device current I collected by the current sensor, the device voltage V collected by the voltage sensor, the device vibration intensity A collected by the vibration sensor, and the device thermal radiation value F and load condition L collected by the infrared sensor.

[0071] In this embodiment, the data acquisition module can monitor the operating status of key equipment in the substation and distribution room in a comprehensive and real-time manner by accurately deploying a variety of sensors such as temperature, current, vibration, and pressure. The sensor deployment unit selects appropriate sensors according to the characteristics of the equipment and the type of fault, and arranges them reasonably on the surface or around the equipment, ensuring the comprehensive collection of multi-dimensional data of the equipment operation. This high-precision data acquisition provides a solid foundation for subsequent fault diagnosis, avoids diagnostic errors caused by data lag or loss, and ensures the real-time and accuracy of the system. The data acquisition unit further optimizes the real-time recording and processing of equipment operation data, ensuring that the data of various sensors are accurately and timely transmitted to the data storage unit. The efficient data storage system can not only store historical data and real-time operation data of the current equipment, but also solve the challenges of large-scale data storage and rapid access through high-performance storage devices and distributed storage architecture. Through this storage mechanism, the system can provide reliable data support for subsequent data analysis, model training, and fault diagnosis, improving the accuracy of fault diagnosis and the operation stability of the system. Through this precise multi-dimensional data acquisition, the system can deeply understand the changing laws of key parameters such as temperature, vibration, and current of the equipment, identify potential faults in advance, and thus achieve more efficient equipment monitoring and maintenance. The introduction of this module not only improves the operational safety of the equipment, but also greatly enhances the system's ability to predict and respond to sudden failures, providing important guarantees for the efficient management of smart power systems.

[0072] Example 3

[0073] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the data preprocessing module includes a data cleaning unit, a data fusion unit and a feature extraction unit;

[0074] The data cleaning unit is used to clean the collected raw data, remove noise data, invalid data and outliers, automatically identify and eliminate irrelevant data through algorithms, and avoid data pollution from having a negative impact on system performance. The data cleaning algorithm is based on statistical methods and is processed through Z-score standardization and mean filtering;

[0075] The data fusion unit is used to integrate the data obtained from different sensors through multimodal data fusion technology to generate a complete equipment operation feature vector. The data fusion process uses weighted average, Kalman filtering, and principal component analysis PCA technology to combine the measurement results of different sensors to calculate and obtain the temperature current comprehensive coefficient γ1, the vibration thermal radiation comprehensive coefficient γ2, the load comprehensive coefficient γ3 and the comprehensive fault diagnosis index ξ;

[0076] The feature extraction unit is used to extract key features from the fused data to form an operation feature vector of the equipment. The feature extraction uses time series analysis and frequency domain analysis methods to extract statistical features from various sensor data, and then generate a set of concise and effective feature vectors for subsequent fault diagnosis and classification.

[0077] The temperature current comprehensive coefficient γ1 is obtained by calculating the following formula:

[0078]

[0079] Where, T max Indicates the maximum safe temperature of the device, I rated Indicates the rated current of the device. It indicates the ratio of the device temperature to the maximum safe temperature. Indicates the ratio of measured current to rated current. It means that the combined effects of temperature and current are considered, because temperature and current interact with each other and may cause equipment to overheat or fail.

[0080] The vibration heat radiation comprehensive coefficient γ2 is obtained by calculating the following formula:

[0081]

[0082] In the formula, Arated represents the rated vibration intensity threshold of the equipment, and Frated represents the rated thermal radiation threshold of the equipment. It is a measure of the ratio of vibration intensity to the rated threshold. It indicates the ratio of the thermal radiation value to the rated threshold. It means that the combined effects of vibration and heat radiation should be considered together, because excessive vibration and heat radiation are usually signs of equipment damage;

[0083] The load comprehensive coefficient γ3 is obtained by calculating through the following formula:

[0084]

[0085] Where, L max It represents the maximum safe load percentage of the equipment, σ represents the standard deviation of the load variation range, which is used to smooth the impact of load fluctuations on the risk of failure. Indicates the percentage of load, converts the load condition into a standardized ratio, represents a Gaussian function describing the load deviation from the maximum load L max When the load is close to the maximum load, the value of the exponential term is close to 1, indicating that excessive load may bring a higher risk of failure;

[0086] The comprehensive fault diagnosis index ξ is obtained by calculating through the following formula:

[0087] ξ=α*γ1+β*γ2+χ*γ3;

[0088] In the formula, γ1 represents the comprehensive coefficient of temperature and current, γ2 represents the comprehensive coefficient of vibration and thermal radiation, γ3 represents the comprehensive coefficient of load, and α, β and χ represent the weight coefficients of the comprehensive coefficient of temperature and current γ1, the comprehensive coefficient of vibration and thermal radiation γ2 and the comprehensive coefficient of load γ3 respectively.

[0089] In this embodiment, the data preprocessing module effectively improves the accuracy of equipment fault diagnosis and the intelligence level of the system through the three core units of data cleaning, fusion and feature extraction. The data cleaning unit uses advanced statistical methods, such as Z-score standardization and mean filtering, to automatically identify and remove noise data, invalid data and outliers to ensure the high quality of the original data. This process greatly reduces the negative impact of data pollution on subsequent analysis and fault diagnosis, and provides the system with more accurate and reliable input data. The data fusion unit integrates data from different sensors by adopting multimodal data fusion technology to generate a complete equipment operation feature vector. The application of technologies such as weighted average, Kalman filtering and principal component analysis (PCA) enables the system to accurately calculate various comprehensive coefficients, such as temperature and current comprehensive coefficients, vibration and thermal radiation comprehensive coefficients, and load comprehensive coefficients. These comprehensive coefficients provide a more comprehensive perspective for fault diagnosis, taking into account the mutual influence between different factors, and can more accurately reflect the health status and potential fault risks of the equipment. The feature extraction unit uses time series analysis and frequency domain analysis methods to extract key features from the fused data to generate a concise and effective equipment operation feature vector. These feature vectors become important inputs for subsequent fault diagnosis and classification models, helping the system to accurately identify abnormal performance in equipment operation and provide high-quality feature data for deep learning models. This module provides a more scientific basis for equipment fault diagnosis by improving the accuracy and efficiency of feature extraction, further improving the overall intelligence level and fault prediction capabilities of the system.

[0090] Example 4

[0091] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the edge computing module includes a data processing unit, a fault detection unit and a data transmission unit;

[0092] The data processing unit is used to perform preliminary processing on the data transmitted to the edge computing device. The data includes temperature, current and thermal radiation device operation status data. The processing goal is to extract key information from the original data, achieve rapid response through edge computing, and reduce the load on the remote server. The data processing unit uses fast Fourier transform FFT to perform data compression and feature extraction;

[0093] The fault detection unit is used to monitor the operation status of the equipment in real time based on the set rules and thresholds, and obtain an evaluation plan by comparing the comprehensive fault diagnosis index ξ with the preset qualified threshold M. When the abnormal operation status of the equipment is monitored, an early warning is triggered through the rule engine to determine whether the equipment needs to be shut down, powered off or load switched immediately. The evaluation criteria include the type of fault, the scope of impact and the potential risk. The fault detection algorithm is based on the historical data of the equipment and combines the rule threshold to make real-time judgments, which can effectively identify potential fault problems of the equipment;

[0094] The data transmission unit is used to transmit the data processed by the edge computing module and the fault warning information to the deep learning fault classification module. The data transmission adopts a low-latency, high-bandwidth communication protocol to transmit the fault information to the cloud system in real time.

[0095] The assessment program includes:

[0096] Qualified threshold M1: The preset minimum standard for normal equipment operation status. When the fault diagnosis index ξ≤M1, it means that the equipment operation status is within the safe range and there is no major fault risk;

[0097] High-risk threshold M2: When the comprehensive fault diagnosis index ξ exceeds the qualified threshold M1 but does not reach the warning value M2 of a serious fault, the equipment status is judged to be a potential problem. When the ξ value reaches M2, the system issues a warning and focuses on monitoring and inspecting the equipment operation status.

[0098] ξ≤M1, obtain the first level assessment, the equipment has no abnormality and operates within the standard range;

[0099] M1<ξ≤M2, the second level assessment is obtained. The equipment has a potential failure risk, but it has not caused a significant impact on the operation. It is necessary to strengthen monitoring and preventive maintenance. The monitoring frequency of the equipment is increased to collect data every 10 minutes. The collection of more data points is increased to detect abnormal changes earlier. The warning value is set through the edge computing module. When ξ exceeds the set value, the operation and maintenance personnel are notified through instant messaging to ensure a quick response. Preventive inspections are carried out on the equipment regularly, focusing on temperature, current and voltage fluctuations.

[0100] ξ>M2, the third level assessment is obtained. The equipment has a serious failure risk, which leads to equipment failure. It is necessary to immediately shut down for inspection, disconnect the power supply, and start the fault handling process to avoid damage to the equipment. Through the collaborative work of the edge computing module and the deep learning fault classification module, the cause of the high-risk state is analyzed to determine whether the equipment is aging, overloaded or has poor contact. According to the fault analysis results, maintenance personnel are arranged in time to repair it, and all relevant parts of the equipment are checked to ensure that it returns to normal working condition.

[0101] The deep learning fault classification module includes a model training unit and a real-time analysis unit;

[0102] The model training unit is used to train the deep learning model based on historical data and equipment fault records. The model training uses large-scale equipment historical data, including normal operation data and fault data, and trains the neural network through a supervised learning method to optimize model parameters to ensure that the deep learning model has high classification accuracy;

[0103] The real-time analysis unit is used to receive data from the edge computing module in real time, and input the data into a trained deep learning model for analysis. Through the model output, the fault type and fault location of the equipment are judged in real time. The analysis results include the fault type, including overheating or short circuit, and the location of the fault, including transformers, busbars or switchgear.

[0104] The deep learning fault classification module also includes a fault location module;

[0105] The fault location unit is used to perform detailed fault location and diagnosis based on the results output by the deep learning model. The specific fault location is determined by combining the topological structure and operating status of the equipment. The results of fault location are used as a basis for decision-making to further guide operation and maintenance personnel to perform on-site maintenance.

[0106] In this embodiment, the edge computing module significantly improves the speed and accuracy of equipment fault warning and response by integrating data processing, fault detection and data transmission units. The data processing unit compresses and extracts features of the original data through fast Fourier transform (FFT), so that the operation data of the equipment can be quickly and accurately transmitted to the subsequent modules, reducing the load and response delay of the remote server. Real-time data processing through edge computing can ensure that the status information of the equipment is quickly fed back, and improve the responsiveness of the system, especially in the application scenarios of high-frequency data collection and processing. The fault detection unit monitors the operating status of the equipment in real time based on the set rules and thresholds, and compares and analyzes the comprehensive fault diagnosis index ξ with the preset qualified threshold M. The unit can quickly assess the fault risk when the equipment is abnormal, and trigger the early warning system through the rule engine to determine whether automatic control measures (such as shutdown, power off or load switching) need to be taken. Through the hierarchical evaluation scheme, the system can classify faults of different degrees, thereby ensuring the response speed of the equipment under different fault conditions and the rationality of the treatment measures, and ensuring the stability and safety of the equipment. The deep learning fault classification module can optimize the model parameters and improve the accuracy of fault classification through training with large-scale historical data. This module receives data transmitted by the edge computing module in real time to determine the type of equipment fault and locate the fault. Combined with the topology and operating status of the equipment, the fault location unit can accurately identify the specific location of the fault and further guide the operation and maintenance personnel to perform on-site maintenance. Through the collaborative work of deep learning and edge computing, the system can not only identify faults in real time, but also make accurate decisions in the shortest time, maximizing the operational reliability and maintenance efficiency of the equipment.

[0107] Example 5

[0108] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the fault alarm module includes an alarm generation unit and an automatic control decision unit;

[0109] The alarm generation unit is used to generate alarm information when a fault occurs, and send it to the operation and maintenance personnel in a timely manner through SMS, email and APP notification. The alarm information includes the fault type, location, severity assessment result and preliminary response measures to help the operation and maintenance personnel respond quickly;

[0110] The automatic control decision unit is used to determine whether automatic control measures need to be taken according to the fault severity assessment and alarm results, including switching loads or power off, and to avoid equipment damage or expansion of faults by executing control instructions.

[0111] The feedback optimization module includes a data feedback unit, a model optimization unit and a strategy optimization unit;

[0112] The data feedback unit is used to feed back equipment operation data, fault diagnosis results and control measures to the system to provide a basis for system optimization;

[0113] The model optimization unit is used to regularly update the deep learning model according to the feedback results, improve the fault classification and location capabilities of the model through online learning, and continuously adapt to new equipment states and fault types through incremental learning to ensure the system's adaptability and prediction capabilities for future faults;

[0114] The strategy optimization unit is responsible for optimizing the edge computing strategy and the automatic control strategy, dynamically adjusting the parameters and threshold settings of the edge computing model based on feedback data, improving the response speed and accuracy of the system in actual operation, and ensuring the efficiency of the system under different load conditions.

[0115] In this embodiment, the combination of the fault alarm module and the feedback optimization module significantly improves the responsiveness, processing efficiency and adaptability of the equipment fault. Specifically, the fault alarm module can generate accurate alarm information immediately when the equipment fails through the alarm generation unit and the automatic control decision unit, and quickly notify the operation and maintenance personnel. The alarm information includes the fault type, location, severity assessment and preliminary response measures, which helps the operation and maintenance personnel obtain key information in the shortest time and respond quickly. At the same time, the automatic control decision unit determines whether it is necessary to automatically take control measures such as load switching or power failure according to the fault severity assessment to avoid the spread of the fault and further damage to the equipment, and maximize the safety of equipment operation. The feedback optimization module continuously optimizes the system performance through the synergy of the data feedback unit, the model optimization unit and the strategy optimization unit. The data feedback unit feeds back the equipment operation data, fault diagnosis results and executed control measures to the system, providing valuable data support for subsequent optimization. The model optimization unit regularly updates the deep learning model according to the feedback information, and improves the fault classification and location capabilities of the model through incremental learning, so that the system can adapt to new equipment states and fault types, and improve the accuracy and prediction capabilities of fault detection. The strategy optimization unit dynamically adjusts the parameter settings of edge computing and automatic control strategies based on feedback data to ensure the response speed and accuracy of the system in actual operation, and to ensure efficient operation under different load conditions. Overall, through the integration of fault alarm and feedback optimization modules, the system can not only respond to faults quickly, but also improve the accuracy and efficiency of fault diagnosis, response, and control through continuous learning and optimization. This series of measures greatly enhances the system's adaptability to complex fault environments, reduces potential risks, and improves the overall reliability and safety of the equipment.

[0116] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A power fault monitoring system, characterized in that: It includes data acquisition module, data preprocessing module, edge computing module, deep learning fault classification module, fault alarm module and feedback optimization module; The data acquisition module is used to deploy multimodal sensors in substations and distribution rooms to collect multi-dimensional data on key equipment in substations and distribution rooms, and comprehensively cover the equipment operation status through real-time collection of multiple sensor data; The data preprocessing module is used to preprocess the data acquired by the data acquisition module, and use multimodal data fusion technology to integrate the data from different sensors to form a complete equipment operation feature vector; The edge computing module is used to perform real-time data analysis and preliminary fault diagnosis locally by deploying intelligent gateway devices. Through the set rules and thresholds, it monitors the operating status of the equipment in real time, detects potential faults and issues preliminary alarms. When the problem is serious, it directly transmits the data to the deep learning fault classification module through the network; The deep learning fault classification module is used to train a deep learning model based on the historical data of the equipment, analyze the equipment operation data in real time, identify the fault type and locate the location where the fault occurs, and the model outputs the specific type and location of the fault and sends it to the fault alarm module; The fault alarm module is responsible for automatically judging the severity of the fault according to the fault diagnosis results output by the deep learning module, and deciding whether to take automatic control measures, including load switching or equipment power off, and generating alarm information to send to the operation and maintenance personnel; The feedback optimization module is used to provide feedback on system operation data, fault diagnosis results and automatic control measures, and regularly updates deep learning models and edge computing strategies through online learning and model updates to improve the system's intelligence level and fault diagnosis accuracy.

2. A power failure monitoring system according to claim 1, characterized in that: The data acquisition module includes a sensor deployment unit, a data acquisition unit and a data storage unit; The sensor deployment unit is used to deploy various types of sensors, including temperature sensors, current sensors and vibration sensors, on transformers, switchgear and busbars in substations and distribution rooms. The sensors are reasonably arranged on the surface or around the equipment according to the operating characteristics and monitoring requirements of the equipment, and various parameters of the equipment are collected in real time. The type of sensor is selected according to the operating environment and fault type of the equipment to ensure that the operating status of the equipment can be fully monitored; The data acquisition unit is used to receive and record data from various sensors, including multi-dimensional parameters such as temperature, current, vibration and pressure, monitor and collect the operating status of the equipment in real time, ensure the real-time and accuracy of the data from each sensor, and avoid affecting the accuracy of fault diagnosis due to delayed acquisition or lost data; The data storage unit is used to store the collected real-time data and provide a cache function. The collected data includes historical data and current equipment operation data. Through efficient storage of data, it can provide basic data support for subsequent data analysis and model training. The use of high-performance storage devices or distributed storage systems ensures that data is not lost under the requirements of high-frequency collection and fast access.

3. A power failure monitoring system according to claim 2, characterized in that: The data acquired by the data acquisition unit includes: The device temperature T collected by the temperature sensor, the device current I collected by the current sensor, the device voltage V collected by the voltage sensor, the device vibration intensity A collected by the vibration sensor, and the device thermal radiation value F and load condition L collected by the infrared sensor.

4. A power failure monitoring system according to claim 1, characterized in that: The data preprocessing module includes a data cleaning unit, a data fusion unit and a feature extraction unit; The data cleaning unit is used to clean the collected raw data, remove noise data, invalid data and outliers, automatically identify and eliminate irrelevant data through algorithms, and avoid data pollution from having a negative impact on system performance. The data cleaning algorithm is based on statistical methods and is processed through Z-score standardization and mean filtering; The data fusion unit is used to integrate the data obtained from different sensors through multimodal data fusion technology to generate a complete equipment operation feature vector. The data fusion process uses weighted average, Kalman filtering, and principal component analysis PCA technology to combine the measurement results of different sensors to calculate and obtain the temperature current comprehensive coefficient γ1, the vibration thermal radiation comprehensive coefficient γ2, the load comprehensive coefficient γ3 and the comprehensive fault diagnosis index ξ; The feature extraction unit is used to extract key features from the fused data to form an operation feature vector of the equipment. The feature extraction uses time series analysis and frequency domain analysis methods to extract statistical features from various sensor data, and then generate a set of concise and effective feature vectors for subsequent fault diagnosis and classification.

5. A power failure monitoring system according to claim 1, characterized in that: The edge computing module includes a data processing unit, a fault detection unit and a data transmission unit; The data processing unit is used to perform preliminary processing on the data transmitted to the edge computing device. The data includes temperature, current and thermal radiation device operation status data. The processing goal is to extract key information from the original data, achieve rapid response through edge computing, and reduce the load on the remote server. The data processing unit uses fast Fourier transform FFT to perform data compression and feature extraction; The fault detection unit is used to monitor the operation status of the equipment in real time based on the set rules and thresholds, and obtain an evaluation plan by comparing the comprehensive fault diagnosis index ξ with the preset qualified threshold M. When the abnormal operation status of the equipment is monitored, an early warning is triggered through the rule engine to determine whether the equipment needs to be shut down, powered off or load switched immediately. The evaluation criteria include the type of fault, the scope of impact and the potential risk. The fault detection algorithm is based on the historical data of the equipment and combines the rule threshold to make real-time judgments, which can effectively identify potential fault problems of the equipment; The data transmission unit is used to transmit the data processed by the edge computing module and the fault warning information to the deep learning fault classification module. The data transmission adopts a low-latency, high-bandwidth communication protocol to transmit the fault information to the cloud system in real time.

6. A power failure monitoring system according to claim 5, characterized in that: The assessment program includes: Qualified threshold M1: The preset minimum standard for normal equipment operation status. When the fault diagnosis index ξ≤M1, it means that the equipment operation status is within the safe range and there is no major fault risk; High-risk threshold M2: When the comprehensive fault diagnosis index ξ exceeds the qualified threshold M1 but does not reach the warning value M2 of a serious fault, the equipment status is judged to be a potential problem. When the ξ value reaches M2, the system issues a warning and focuses on monitoring and inspecting the equipment operation status. ξ≤M1, obtain the first level assessment, the equipment has no abnormality and operates within the standard range; M1<ξ≤M2, the second level assessment is obtained. The equipment has a potential failure risk, but it has not caused a significant impact on the operation. It is necessary to strengthen monitoring and preventive maintenance. The monitoring frequency of the equipment is increased to collect data every 10 minutes. The collection of more data points is increased to detect abnormal changes earlier. The warning value is set through the edge computing module. When ξ exceeds the set value, the operation and maintenance personnel are notified through instant messaging to ensure a quick response. Preventive inspections are carried out on the equipment regularly, focusing on temperature, current and voltage fluctuations. ξ>M2, the third level assessment is obtained. The equipment has a serious failure risk, which leads to equipment failure. It is necessary to immediately shut down for inspection, disconnect the power supply, and start the fault handling process to avoid damage to the equipment. Through the collaborative work of the edge computing module and the deep learning fault classification module, the cause of the high-risk state is analyzed to determine whether the equipment is aging, overloaded or has poor contact. According to the fault analysis results, maintenance personnel are arranged in time to repair it, and all relevant parts of the equipment are checked to ensure that it returns to normal working condition.

7. A power failure monitoring system according to claim 1, characterized in that: The deep learning fault classification module includes a model training unit and a real-time analysis unit; The model training unit is used to train the deep learning model based on historical data and equipment fault records. The model training uses large-scale equipment historical data, including normal operation data and fault data, and trains the neural network through a supervised learning method to optimize model parameters to ensure that the deep learning model has high classification accuracy; The real-time analysis unit is used to receive data from the edge computing module in real time, and input the data into a trained deep learning model for analysis. Through the model output, the fault type and fault location of the equipment are judged in real time. The analysis results include the fault type, including overheating or short circuit, and the location of the fault, including transformers, busbars or switchgear.

8. The power failure monitoring system according to claim 1, characterized in that: The deep learning fault classification module also includes a fault location module; The fault location unit is used to perform detailed fault location and diagnosis based on the results output by the deep learning model. The specific fault location is determined by combining the topological structure and operating status of the equipment. The results of fault location are used as a basis for decision-making to further guide operation and maintenance personnel to perform on-site maintenance.

9. A power failure monitoring system according to claim 1, characterized in that: The fault alarm module includes an alarm generation unit and an automatic control decision unit; The alarm generation unit is used to generate alarm information when a fault occurs, and send it to the operation and maintenance personnel in a timely manner through SMS, email and APP notification. The alarm information includes the fault type, location, severity assessment result and preliminary response measures to help the operation and maintenance personnel respond quickly; The automatic control decision unit is used to determine whether automatic control measures need to be taken according to the fault severity assessment and alarm results, including switching loads or power off, and to avoid equipment damage or expansion of faults by executing control instructions.

10. The power failure monitoring system according to claim 1, characterized in that: The feedback optimization module includes a data feedback unit, a model optimization unit and a strategy optimization unit; The data feedback unit is used to feed back equipment operation data, fault diagnosis results and control measures to the system to provide a basis for system optimization; The model optimization unit is used to regularly update the deep learning model according to the feedback results, improve the fault classification and location capabilities of the model through online learning, and continuously adapt to new equipment states and fault types through incremental learning to ensure the system's adaptability and prediction capabilities for future faults; The strategy optimization unit is responsible for optimizing the edge computing strategy and the automatic control strategy, dynamically adjusting the parameters and threshold settings of the edge computing model based on feedback data, improving the response speed and accuracy of the system in actual operation, and ensuring the efficiency of the system under different load conditions.

Images