Transformer substation switch cabinet state online monitoring device and method based on multi-source data fusion

Abstract

The invention relates to the technical field of substation equipment state monitoring, in particular to a substation switch cabinet state online monitoring device and method based on multi-source data fusion, and the monitoring device comprises a multi-dimensional sensing acquisition system, an edge intelligent processing unit, a multi-mode communication system and an intelligent early warning and decision-making system. The multi-dimensional sensing acquisition system acquires various types of operation parameters, the edge intelligent processing unit processes the parameters, the multi-modal communication system transmits data, and the intelligent early warning and decision-making system performs early warning and generates a maintenance scheme; according to the monitoring method, sensor deployment and calibration, multi-source data acquisition and synchronization, data preprocessing and feature extraction, state diagnosis and health assessment, data transmission and cloud collaboration, and early warning triggering and maintenance decision making are carried out in sequence. The switch cabinet state can be comprehensively monitored, the fault diagnosis accuracy is improved, and safe and stable operation of a power grid is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of substation equipment state monitoring, and is a substation switch cabinet state online monitoring device and method based on multi-source data fusion. BACKGROUND

[0002] The substation switch cabinet is the core equipment for realizing power distribution, control and protection in the power system, and its operation state directly affects the safety and reliability of the power grid. According to statistics, about 35% of substation failures are caused by internal partial discharge, contact overheating or mechanical jamming and other hidden dangers in the switch cabinet, which can easily threaten the safe and stable operation of the power grid and the safety of the staff when a failure occurs. Therefore, it is urgent to develop an intelligent monitoring device with multi-source data fusion capability and self-learning characteristics.

[0003] Currently, the traditional switch cabinet monitoring system relies on a single sensor to collect parameters and can only obtain isolated data such as temperature or current, which cannot fully reflect the overall operation state of the equipment and cannot realize the synchronous monitoring of multiple key hidden dangers such as partial discharge and gas state. It is easy to miss potential fault hidden dangers and difficult to meet the demand for comprehensive control of the operation state of the switch cabinet.

[0004] The existing monitoring device generally uses wired transmission mode, which not only has complex wiring and high construction difficulty, increases the cost of substation site reconstruction and maintenance, but also is easily disturbed by the strong electromagnetic environment in the substation, resulting in deviation or loss of monitoring data, which cannot guarantee the real-time and accuracy of data transmission and cannot meet the requirements of the power grid for real-time monitoring of equipment state. Moreover, the existing data processing method is based on fixed threshold alarm and lacks intelligent analysis capability combined with the operation characteristics of the equipment, often resulting in false alarm or missed alarm, which cannot accurately identify the fault type and severity in time and delays the fault disposal opportunity, which threatens the safe operation of the power grid. SUMMARY

[0005] The present application provides a substation switch cabinet state online monitoring device and method based on multi-source data fusion, which overcomes the above-mentioned deficiencies of the prior art and effectively solves the problems of single parameter collection, poor transmission anti-interference and low intelligent diagnosis of the traditional monitoring system.

[0006] One of the technical solutions of the present application is realized by the following measures: a transformer substation switch cabinet state online monitoring device based on multi-source data fusion, comprising a multi-dimensional sensing acquisition system, an edge intelligent processing unit, a multi-modal communication system and an intelligent early warning and decision system; wherein the multi-dimensional sensing acquisition system is connected with the edge intelligent processing unit, and is used for acquiring multiple types of operating parameters of the transformer substation switch cabinet; wherein the edge intelligent processing unit is connected with the multi-dimensional sensing acquisition system and the multi-modal communication system respectively, and is used for receiving and processing the operating parameters collected by the multi-dimensional sensing acquisition system; wherein the multi-modal communication system is connected with the edge intelligent processing unit and the intelligent early warning and decision system respectively, and is used for transmitting the operating parameter related data processed by the edge intelligent processing unit; wherein the intelligent early warning and decision system is connected with the multi-modal communication system, and is used for making transformer substation switch cabinet state early warning and maintenance decision based on the data transmitted by the multi-modal communication system.

[0007] The following is a further optimization or / and improvement of one of the above technical solutions of the present application: The above multi-dimensional sensing acquisition system can include a distributed optical fiber sensing array, which is arranged in the busbar chamber and the circuit breaker chamber of the switch cabinet, and is used for monitoring the temperature and vibration signals of the switch cabinet.

[0008] The above multi-dimensional sensing acquisition system can include a wideband partial discharge sensor, which integrates a very high frequency antenna array and a transient ground voltage sensor, and is used for collecting the partial discharge signals of the switch cabinet.

[0009] The above multi-dimensional sensing acquisition system can include an intelligent contact monitoring module, which is used for monitoring the wear and temperature of the circuit breaker contact.

[0010] The above multi-dimensional sensing acquisition system can include a gas density monitoring unit, which integrates a MEMS pressure sensor and a laser gas analysis module, and is used for monitoring the density and micro water content of SF6 gas in the switch cabinet.

[0011] A dynamic feature extraction engine can be arranged in the above edge intelligent processing unit, which is based on a wavelet packet transform algorithm and is used for extracting time domain, frequency domain and time-frequency domain feature parameters of the operating parameters, and a device feature database is established in the edge intelligent processing unit, which is used for storing feature vectors under typical working conditions of the switch cabinet.

[0012] A hybrid diagnostic model can be constructed in the above edge intelligent processing unit, which is used for fault diagnosis of the operating state of the switch cabinet.

[0013] The second technical solution of the application is realized by the following measures: a substation switch cabinet state online monitoring method based on multi-source data fusion, comprising the following steps: step one, sensor deployment and calibration: the sensors of the multi-dimensional sensing acquisition system are planned and arranged at key positions of the switch cabinet, each sensor is calibrated online using a standard source, and a sensor-channel mapping table is established; step two, multi-source data acquisition and synchronization: the switch cabinet operation parameters are acquired through the multi-dimensional sensing acquisition system, and time synchronization of multiple sensors is realized using a hardware trigger signal; step three, data preprocessing and feature extraction: the original operation parameters are transmitted to the edge intelligent processing unit for noise reduction processing, and the multi-dimensional feature parameters of the operation parameters are extracted through a dynamic feature extraction engine; step four, state diagnosis and health assessment: the edge intelligent processing unit calls a hybrid diagnosis model to analyze the feature parameters, judges the switch cabinet fault state in combination with a device feature database, and calculates a device health index; step five, data transmission and cloud collaboration: the diagnosis results and health index are transmitted to an intelligent early warning and decision system through a multi-modal communication system, a virtual monitoring system is constructed based on digital twinning, data interaction is realized with a SCADA system, and the cloud updates the diagnosis model; step six, early warning triggering and maintenance decision: the intelligent early warning and decision system judges whether to trigger early warning through an adaptive threshold model, pushes early warning information according to a three-level early warning system, and calls a maintenance decision engine to generate a maintenance scheme.

[0014] The following is a further optimization or / and improvement of the above-mentioned second technical solution of the application: In the above-mentioned step one, an improved ant colony algorithm is used for sensor planning to ensure that the switch cabinet monitoring blind area coverage rate is less than 2%, and the temperature sensor is calibrated by two-point calibration using a constant temperature tank, and the partial discharge sensor is calibrated by gain calibration using a standard pulse generator.

[0015] In the above-mentioned step four, the device health index is calculated by a pre-trained random forest model, and the SHAP value analysis method is used to sort the feature importance, and the key features are selected for health index calculation.

[0016] This invention enables comprehensive acquisition of multi-dimensional operating parameters of substation switchgear. Through a multi-dimensional sensing system, it covers key data such as temperature, vibration, partial discharge, contact status, and gas parameters, breaking the limitations of traditional single-parameter monitoring and allowing staff to fully grasp the equipment's operating status. The edge intelligent processing unit possesses efficient data processing and intelligent diagnostic capabilities. Utilizing a dynamic feature extraction engine, it accurately extracts parameter features and, combined with a hybrid diagnostic model, accurately identifies faults, significantly reducing false alarms and missed alarms and improving fault identification reliability. A multi-modal communication system ensures stable data transmission, avoiding interference and wiring difficulties associated with wired transmission, ensuring real-time and accurate transmission of monitoring data to the intelligent early warning and decision-making system. The intelligent early warning and decision-making system can promptly trigger tiered early warnings and generate scientific maintenance plans, assisting staff in quickly responding to faults, reducing fault handling time, and mitigating the risk of grid outages. Simultaneously, the monitoring method, through standardized sensor deployment, data synchronization, and preprocessing steps, further ensures the orderliness and accuracy of the monitoring process, comprehensively improving the intelligence and comprehensiveness of substation switchgear status monitoring, effectively ensuring the safe and stable operation of the power grid, and providing strong support for reliable power supply to the power system. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the online monitoring system for substation switchgear status based on multi-source data fusion, according to an embodiment of the present invention.

[0018] Figure 2 This is a flowchart illustrating the PHM method according to an embodiment of the present invention. Detailed Implementation

[0019] The present invention is not limited to the following embodiments, and the specific implementation can be determined according to the technical solution of the present invention and the actual situation.

[0020] The present invention will be further described below with reference to embodiments: Example 1: As Figure 1As shown, this embodiment provides an online monitoring device for substation switchgear status based on multi-source data fusion, including a multi-dimensional sensing acquisition system, an edge intelligent processing unit, a multi-modal communication system, and an intelligent early warning and decision-making system. The multi-dimensional sensing acquisition system is connected to the edge intelligent processing unit to collect various operating parameters of the substation switchgear. The edge intelligent processing unit is connected to both the multi-dimensional sensing acquisition system and the multi-modal communication system to receive and process the operating parameters collected by the multi-dimensional sensing acquisition system. The multi-modal communication system is connected to both the edge intelligent processing unit and the intelligent early warning and decision-making system to transmit the processed operating parameter data. The intelligent early warning and decision-making system is connected to the multi-modal communication system to perform substation switchgear status early warning and maintenance decisions based on the data transmitted by the multi-modal communication system. In this way, the device integrates multiple systems to work collaboratively, achieving full-process monitoring from parameter acquisition to early warning decision-making. Multi-dimensional acquisition ensures parameter comprehensiveness, edge processing improves data processing efficiency, and multi-modal communication ensures stable data transmission, providing a complete solution for switchgear status monitoring and effectively supporting the safe operation of the power grid.

[0021] In this embodiment, the multi-dimensional sensing and acquisition system may include a distributed fiber optic sensor array, which is deployed in the busbar compartment and circuit breaker compartment of the switchgear to monitor the temperature and vibration signals of the switchgear. By deploying the distributed fiber optic sensor array in key locations, it can accurately capture temperature changes and vibration conditions in the busbar compartment and circuit breaker compartment, promptly detect potential faults caused by abnormal temperatures or mechanical vibrations, provide crucial data support for equipment condition assessment, and improve the targeting and effectiveness of monitoring.

[0022] In this embodiment, the multi-dimensional sensing and acquisition system may include a broadband partial discharge sensor. This broadband partial discharge sensor integrates an ultra-high frequency antenna array and a transient ground voltage sensor to acquire partial discharge signals from the switchgear. By integrating these two sensors, the broadband partial discharge sensor can more comprehensively capture different types of partial discharge signals, avoiding the monitoring blind spots of a single sensor, improving the accuracy and comprehensiveness of partial discharge signal acquisition, and providing a reliable basis for insulation fault diagnosis.

[0023] In this embodiment, the multi-dimensional sensing and acquisition system may include an intelligent contact monitoring module for monitoring the wear and temperature of the circuit breaker contacts. This allows for a dedicated monitoring module specifically designed for the circuit breaker contacts, enabling real-time monitoring of contact wear and temperature changes. As critical conductive components, the condition of the contacts directly affects the reliability of circuit continuity. This module can promptly detect excessive contact wear or overheating, preventing circuit failures caused by contact malfunctions.

[0024] In this embodiment, the multi-dimensional sensing and acquisition system may include a gas density monitoring unit. This unit integrates a MEMS pressure sensor and a laser gas analysis module to monitor the density and moisture content of SF6 gas within the switchgear. Since SF6 gas is crucial to the insulation performance of the switchgear, this monitoring unit can accurately monitor the gas density and moisture content. When the density is too low or the moisture content is too high, it can provide timely warnings, preventing safety accidents caused by a decline in gas insulation performance and ensuring the reliability of the switchgear insulation.

[0025] In this embodiment, the multi-dimensional sensing and acquisition system is equipped with the following sensors: Distributed fiber optic sensor array: Distributed fiber optic (DTS) is deployed in key locations such as busbar room and circuit breaker room to achieve temperature monitoring at the 0.1℃ level through the grating reflection principle, with a spatial resolution of 3cm and a coverage length of 200m; vibration signals are monitored synchronously using the Raman scattering principle, with a sampling frequency of 10kHz, and can detect mechanical displacement at the 0.01μm level.

[0026] Wideband partial discharge sensor: It integrates an ultra-high frequency (UHF) antenna array (300-3GHz) and a transient ground voltage (TEV) sensor, and uses array signal processing technology to achieve ±0.5ns time synchronization and a positioning accuracy of ±5cm; it is equipped with a fluorescent fiber optic sensor (sensitivity 0.1pC) for partial discharge detection in strong electromagnetic interference environments.

[0027] Intelligent contact monitoring module: It adopts the principle of non-contact electromagnetic induction and coats the surface of the circuit breaker contact with a nano-scale ferrite film. It monitors the wear of the contact in real time through the eddy current effect (resolution 0.01mm). Combined with an infrared thermal imager (640×480 pixels, temperature measurement range 20~150℃), it realizes the joint diagnosis of contact temperature and contact resistance.

[0028] Gas density monitoring unit: integrates MEMS pressure sensor (0-100 kPa range, accuracy ±0.1% FS) and laser gas analysis module to monitor SF6 gas density (1 ppm resolution) and trace moisture content (0-5000 ppm, accuracy ±2% RH) in real time.

[0029] In this embodiment, the edge intelligent processing unit may be equipped with a dynamic feature extraction engine. This engine, based on the wavelet packet transform algorithm, is used to extract time-domain, frequency-domain, and time-frequency-domain feature parameters of the operating parameters. Furthermore, the edge intelligent processing unit maintains a device feature database to store feature vectors under typical operating conditions of the switchgear. In this way, the feature extraction engine based on the wavelet packet transform algorithm can mine parameter features from multiple dimensions, providing a more comprehensive extraction method compared to traditional methods. The device feature database provides a reference for feature comparison, helping to accurately identify whether parameters deviate from normal operating conditions and improving the intelligence level of data processing.

[0030] In this embodiment, a hybrid diagnostic model can be built within the edge intelligent processing unit for fault diagnosis of the switchgear's operating status. This allows the hybrid diagnostic model to combine multiple diagnostic logics, comprehensively analyze parameter characteristics, avoid the limitations of a single model, improve the accuracy of fault identification, effectively distinguish different types of faults, provide precise direction for subsequent fault handling, and reduce unnecessary maintenance caused by misjudgments.

[0031] In this embodiment, the edge intelligence processing unit includes: Heterogeneous computing architecture: It adopts an FPGA+ARM dual-core processor. The FPGA is responsible for real-time signal processing (FFT operation rate of 1GS / s), and the ARM Cortex A72 is used to process complex algorithms. It is equipped with 8GB DDR4 memory and 128GB eMMC storage, and supports 256 channels of parallel data acquisition.

[0032] Dynamic Feature Extraction Engine: Develop a feature extraction algorithm based on wavelet packet transform, capable of simultaneously extracting multi-dimensional feature parameters such as time domain (kurtosis, impulse factor), frequency domain (bandwidth energy ratio), and time-frequency domain (wavelet entropy); establish an equipment feature database, storing 10 typical operating conditions. 6 Group feature vectors.

[0033] Hybrid diagnostic model: Construct a diagnostic framework that integrates LSTMAttention neural network and physical mechanism model; the LSTM network has 64 input layer nodes, 128 GRU units in the hidden layer, and the attention mechanism weight dimension is 32×32; the physical model includes partial differential equations such as heat conduction equation (∂T / ∂t=α∇²T) and mechanical vibration equation (mẍ+kx=F(t)).

[0034] In this embodiment, the multimodal communication system includes: Wireless self-organizing network module: adopts LoRa+NBIoT dual-mode communication, LoRa operates at 868MHz (transmission distance 3km), and NBIoT supports the CatNB2 standard (uplink rate 250kbps); deploys an adaptive routing protocol, with network latency <50ms.

[0035] Fiber optic ring network interface: Equipped with an SFP photoelectric conversion module, supporting the IEC6185092LE protocol, with a communication rate of 100Mbps, meeting the data modeling requirements of the IEC6185072 standard.

[0036] Edge-cloud collaboration: Establish a virtual monitoring system based on digital twins, and realize data interaction with SCADA system (data acquisition and monitoring control system) through OPCUA protocol; deploy TensorFlowServing inference engine in the cloud, with model update cycle ≤1 hour.

[0037] In this embodiment, the intelligent early warning and decision-making system includes: Dynamic threshold algorithm: Develop an adaptive threshold model based on the Equipment Health Index (EHI), EHI=∑(w_i·F_i), where w_i is the feature weight (determined by the entropy weight method) and F_i is the standardized feature value; the threshold calculation adopts the sliding window optimization algorithm (window length 24h, step size 1h).

[0038] Fault evolution prediction: A hybrid prediction framework based on an improved grey prediction model (GM(1,1)) and LSTM is established to predict the equipment status trend in the next 24 hours with a prediction error of <3%.

[0039] Maintenance decision engine: Construct a multi-objective optimization model with the objective function min{C_maintenance+C_downtime} (where C_maintenance is the maintenance cost and C_downtime is the downtime cost), and constraints include safety thresholds, spare parts inventory, etc.; use the NSGAII algorithm to solve the Pareto optimal solution and generate a 5-level handling plan including immediate shutdown, enhanced inspection, etc.

[0040] The specific implementation steps of this invention are as follows: (1) Deployment stage: Before the switch cabinet is installed, the sensor network topology planning is completed, and the sensor layout scheme is optimized by the improved ant colony algorithm to ensure that the monitoring blind zone coverage rate is <2%; the typical configuration is: 8 sets of fiber optic temperature measurement points (spaced 2m apart), 4 UHF sensors (arranged at the four corners of the cabinet), and 2 vibration sensors (installed on the circuit breaker base) are arranged on each cabinet. (2) Calibration stage: The sensors are calibrated online using a standard source. The temperature sensor is calibrated at two points using a constant temperature bath (accuracy 0.1℃), and the partial discharge sensor is calibrated by a standard pulse generator (amplitude error <1%). A sensor channel mapping table is established to store the calibration coefficient matrix of each channel. (3) Multi-source data synchronization: Multi-sensor time synchronization is achieved through hardware trigger signal (TTL level), with synchronization accuracy <1μs; an improved synchronization buffer queue algorithm is adopted to solve the delay problem caused by network transmission jitter. (4) Signal denoising: The acquired raw signal is processed in three levels: ① The front-end FPGA implements moving average filtering (window width 50); ② The ARM side uses improved LMS adaptive filtering (step size 0.01); ③ The cloud application uses wavelet threshold denoising (db4 wavelet basis, threshold selection VisuShrink algorithm). (5) Feature engineering: Extract 32-dimensional feature parameters such as time domain features (e.g., root mean square value, peak factor), frequency domain features (frequency domain energy ratio, centroid frequency), and time-frequency domain features (wavelet packet energy entropy); construct feature vector X∈R³² and perform normalization processing (MinMax normalization). (6) Health index calculation: Based on feature vector X, calculate the health index of each subsystem through a pre-trained random forest model (200 trees, maximum depth 15); the importance of features is ranked using SHAP value analysis to select key features with a contribution of >5%. (7) Fault classification and identification: Deploy an improved 1DCNN model (convolution kernel size 3, number of layers 12), the input is a time domain signal sequence (length 1024), and the output is the probability distribution of fault categories; the training data contains 10,000 normal / fault samples, and the transfer learning strategy is adopted. The pre-trained model uses the IEEE PES dataset. (8) Remaining life prediction: Construct a prediction model based on PHM (Prognostics and Health Management), the input is a historical health index sequence, and the output is the state prediction for the next 24 / 48 / 72 hours; the PHMTransformer architecture is adopted, the multi-head attention dimension is 512, and the position encoding adopts the sinusoidal method. (9) Multi-level early warning mechanism: Establish a three-level early warning system. The first-level early warning (yellow) is triggered when a single feature exceeds the baseline value by 15%, and is pushed to the mobile APP of the operation and maintenance personnel; the second-level early warning (orange) requires two or more features to exceed the threshold by 20%, and the local sound and light alarm is activated; the third-level early warning (red) triggers the automatic trip command and notifies the dispatch center.(10) Maintenance strategy generation: Based on the reinforcement learning framework (PPO algorithm), a maintenance decision model centered on equipment reliability is constructed; the state space contains 12 discrete states (such as normal, local overheating, insulation degradation, etc.), the action space defines 6 maintenance operations, and the reward function is designed as R=λ1·R_safety+λ2·R_economic (R_safety is the safety benefit, R_economic is the economic benefit), where λ1=0.7, λ2=0.3. (11) Remote control interface: The interaction with the intelligent terminal is realized through the IEC618507420 standard, supporting remote opening and closing operations (operation delay <200ms), energy storage motor control (start and stop accuracy ±1ms), and other functions.

[0041] During operation, the multi-dimensional sensing and acquisition system collects various operating parameters such as temperature, vibration, partial discharge, contact status, and SF6 gas parameters at key parts of the switchgear, and transmits these parameters to the edge intelligent processing unit. The edge intelligent processing unit extracts multi-dimensional features of the parameters through a dynamic feature extraction engine, combines them with the equipment feature database, and uses a hybrid diagnostic model to determine whether the equipment has a fault and assess its health status. The processed diagnostic results and health data are transmitted to the intelligent early warning and decision-making system through a multi-modal communication system. The intelligent early warning and decision-making system determines whether to trigger an early warning based on the data, pushes early warning information, and generates a maintenance plan, realizing real-time monitoring and intelligent control of the switchgear status, ensuring the reliable operation of the switchgear, and maintaining the safety and stability of the power grid.

[0042] This invention utilizes a multi-dimensional sensing and acquisition system, an edge intelligent processing unit, a multi-modal communication system, and an intelligent early warning and decision-making system. Employing a PHM (Prognostics and Health Management) predictive model, it integrates monitoring of eight parameters, including temperature, vibration, and partial discharge, increasing status information by over 300% compared to traditional solutions. The hybrid diagnostic model achieves a fault identification accuracy of 98.7%, a 22 percentage point improvement over a single model. Redundant design (dual power supply, dual communication links) results in an MTBF (Mean Time Between Failures) of 100,000 hours. This solution successfully diagnoses faults in substation switchgear, ensuring equipment reliability and the safe and stable operation of the power grid.

[0043] Example 2: Figure 2As shown, this embodiment provides a method for online monitoring of substation switchgear status based on multi-source data fusion, including the following steps: Step 1, Sensor Deployment and Calibration: The sensors of the multi-dimensional sensing acquisition system are planned and deployed in key parts of the switchgear. Standard sources are used to calibrate each sensor online, and a sensor-channel mapping table is established. Step 2, Multi-Source Data Acquisition and Synchronization: Switchgear operating parameters are acquired through the multi-dimensional sensing acquisition system, and hardware trigger signals are used to achieve multi-sensor time synchronization. Step 3, Data Preprocessing and Feature Extraction: The raw operating parameters are transmitted to the edge intelligent processing unit for noise reduction processing, and a dynamic feature extraction engine is used to extract multiple data from the operating parameters. The method involves several steps: Step 4, Status Diagnosis and Health Assessment: The edge intelligent processing unit calls a hybrid diagnostic model to analyze the feature parameters, combines the equipment feature database to determine the switchgear fault status, and calculates the equipment health index; Step 5, Data Transmission and Cloud Collaboration: The diagnostic results and health index are transmitted to the intelligent early warning and decision-making system through a multimodal communication system. A virtual monitoring system is built based on digital twins and interacts with the SCADA system. The diagnostic model is updated in the cloud; Step 6, Early Warning Triggering and Maintenance Decision: The intelligent early warning and decision-making system uses an adaptive threshold model to determine whether to trigger an early warning, pushes early warning information according to a three-level early warning system, and calls the maintenance decision engine to generate a maintenance plan. This method ensures that monitoring work is carried out in a standardized and step-by-step manner, forming a complete process from sensor deployment to early warning decision-making. Data synchronization and preprocessing ensure data quality, feature extraction and model diagnosis improve the accuracy of fault identification, and cloud collaboration and hierarchical early warning assist in efficient management and control, ensuring that monitoring work is carried out in an orderly and accurate manner.

[0044] In this embodiment, in step one, an improved ant colony algorithm is used to plan the sensor placement, ensuring that the blind zone coverage of the switchgear monitoring is less than 2%. The temperature sensor undergoes two-point calibration using a constant temperature bath, and the partial discharge sensor undergoes gain calibration using a standard pulse generator. This allows the improved ant colony algorithm to optimize sensor placement, minimize monitoring blind zones, and ensure more comprehensive parameter acquisition. The two-point calibration and gain calibration ensure the measurement accuracy of the temperature and partial discharge sensors, avoiding data deviations caused by sensor errors and providing reliable basic data for subsequent data processing and diagnosis.

[0045] In this embodiment, in step four, the equipment health index is calculated using a pre-trained random forest model, and the SHAP value analysis method is used to rank the importance of features and select key features for health index calculation. This allows the random forest model to possess strong data analysis and prediction capabilities, accurately calculating the health index based on multiple features to reflect the true health status of the equipment; the SHAP value analysis method selects key features, reducing the interference of redundant features on the calculation results, improving the efficiency and accuracy of health index calculation, and providing precise quantitative basis for equipment status assessment.

[0046] During operation, the system first plans and deploys sensor points using an improved ant colony algorithm, and establishes a mapping table by calibrating the sensors with a standard source. Then, it collects parameters through a multi-dimensional sensing acquisition system, and achieves data synchronization via hardware triggering. Next, the raw data is transmitted to an edge intelligent processing unit for noise reduction and multi-dimensional feature extraction. Subsequently, a hybrid diagnostic model is invoked in conjunction with a feature database to determine faults and calculate health indices. Data is then transmitted through a multi-modal communication system to construct a digital twin system that interacts with SCADA, and the model is updated in the cloud. Finally, an intelligent early warning and decision-making system triggers tiered early warnings and generates maintenance plans. This entire approach enables accurate and efficient monitoring of the switchgear status, ensuring reliable equipment operation.

[0047] In this invention, "multi-source data fusion" refers to integrating different types of monitoring data, such as temperature, vibration, partial discharge, contact status, and SF6 gas parameters, acquired by a multi-dimensional sensing system. This data is then comprehensively analyzed and processed by an edge intelligent processing unit to fully reflect the switchgear's operating status and improve fault diagnosis accuracy. The "edge intelligent processing unit" is a hardware unit capable of receiving, processing, and analyzing data. It can perform real-time data processing close to the data acquisition end, reducing data transmission latency and improving monitoring response speed. The "multi-modal communication system" refers to a system that simultaneously supports multiple communication methods, including wireless and wired communication, allowing for the selection of appropriate communication methods based on the actual scenario. This invention employs several methods to ensure the stability and reliability of monitoring data transmission. The "digital twin virtual monitoring system" refers to a virtual model built upon real switchgear, capable of synchronously mapping the operating status of real equipment. This allows staff to intuitively grasp equipment conditions and assists in fault analysis and maintenance decisions. The "hybrid diagnostic model" combines multiple diagnostic algorithms or logics, leveraging the advantages of different models to improve the ability to identify different types of faults and reduce the false alarm rate. The "SCADA system," or data acquisition and monitoring control system, enables remote monitoring and control of substation equipment operation. In this invention, it interacts with the digital twin system to achieve data sharing and collaborative management. This invention uses a PHM (Prognostics and Health Management) predictive model, and can be extended to develop fault diagnosis and fault location methods for different types of substation switchgear equipment.

[0048] The above technical features constitute the embodiments of the present invention, which have strong adaptability and implementation effect. Unnecessary technical features can be added or removed according to actual needs to meet the needs of different situations.

Claims

1. A substation switchgear status online monitoring device based on multi-source data fusion, characterized in that, It includes a multi-dimensional sensing and acquisition system, an edge intelligent processing unit, a multi-modal communication system, and an intelligent early warning and decision-making system; Among them, the multi-dimensional sensing and acquisition system is connected to the edge intelligent processing unit to collect various operating parameters of the substation switchgear. The edge intelligent processing unit is connected to the multi-dimensional sensing and acquisition system and the multi-modal communication system, respectively, and is used to receive and process the operating parameters collected by the multi-dimensional sensing and acquisition system. Among them, the multimodal communication system is connected to the edge intelligent processing unit and the intelligent early warning and decision-making system respectively, and is used to transmit the relevant data of the operating parameters processed by the edge intelligent processing unit; The intelligent early warning and decision-making system is connected to the multimodal communication system and is used to make early warning and maintenance decisions for substation switchgear based on the data transmitted by the multimodal communication system.

2. The online monitoring device for substation switchgear status based on multi-source data fusion according to claim 1, characterized in that, The multi-dimensional sensing and acquisition system includes a distributed fiber optic sensor array, which is arranged in the busbar compartment and circuit breaker compartment of the switchgear to monitor the temperature and vibration signals of the switchgear.

3. The online monitoring device for substation switchgear status based on multi-source data fusion according to claim 1, characterized in that, The multi-dimensional sensing and acquisition system includes a broadband partial discharge sensor, which integrates an ultra-high frequency antenna array and a transient ground voltage sensor to acquire partial discharge signals from the switchgear.

4. The online monitoring device for substation switchgear status based on multi-source data fusion according to claim 1, 2, or 3, characterized in that, The multi-dimensional sensing and acquisition system includes an intelligent contact monitoring module, which is used to monitor the wear and temperature of the circuit breaker contacts.

5. A substation switchgear status online monitoring device based on multi-source data fusion according to claim 1, 2, or 3, characterized in that, The multi-dimensional sensing and acquisition system includes a gas density monitoring unit, which integrates a MEMS pressure sensor and a laser gas analysis module to monitor the density and trace moisture content of SF6 gas inside the switch cabinet.

6. A substation switchgear status online monitoring device based on multi-source data fusion according to claim 1, 2, or 3, characterized in that, The edge intelligent processing unit is equipped with a dynamic feature extraction engine, which is based on the wavelet packet transform algorithm and is used to extract the time domain, frequency domain and time-frequency domain feature parameters of the operating parameters. In addition, the edge intelligent processing unit has established a device feature database to store feature vectors under typical operating conditions of the switchgear.

7. A substation switchgear status online monitoring device based on multi-source data fusion according to claim 1, 2, or 3, characterized in that, The edge intelligent processing unit has a hybrid diagnostic model built in it for fault diagnosis of the switch cabinet's operating status.

8. A method for online monitoring of substation switchgear status based on multi-source data fusion, characterized in that, Includes the following steps: Step 1, Sensor Deployment and Calibration: Plan the deployment of sensors for the multi-dimensional sensing acquisition system and place them in key parts of the switch cabinet. Use a standard source to calibrate each sensor online and establish a sensor-channel mapping table. Step 2, Multi-source data acquisition and synchronization: The operating parameters of the switchgear are acquired through a multi-dimensional sensing acquisition system, and the time synchronization of multiple sensors is achieved by using hardware trigger signals; Step 3: Data preprocessing and feature extraction: The raw operating parameters are transmitted to the edge intelligent processing unit for noise reduction, and multi-dimensional feature parameters of the operating parameters are extracted through the dynamic feature extraction engine; Step 4, Status Diagnosis and Health Assessment: The edge intelligent processing unit calls the hybrid diagnostic model to analyze the characteristic parameters, combines the equipment characteristic database to determine the fault status of the switchgear and calculate the equipment health index; Step 5: Data transmission and cloud collaboration: Transmit diagnostic results and health indices to the intelligent early warning and decision-making system through a multimodal communication system, build a virtual monitoring system based on digital twin and achieve data interaction with the SCADA system, and update the diagnostic model in the cloud; Step Six: Early Warning Triggering and Maintenance Decision: The intelligent early warning and decision system determines whether to trigger an early warning through an adaptive threshold model, pushes early warning information according to the three-level early warning system, and calls the maintenance decision engine to generate a maintenance plan.

9. The method for online monitoring of substation switchgear status based on multi-source data fusion according to claim 8, characterized in that, In step one, an improved ant colony algorithm is used to plan the placement of sensors to ensure that the blind spot coverage of the switch cabinet is less than 2%, and the temperature sensor is calibrated at two points using a constant temperature bath, while the partial discharge sensor is calibrated for gain using a standard pulse generator.

10. A method for online monitoring of substation switchgear status based on multi-source data fusion according to claim 8 or 9, characterized in that, In step four, the equipment health index is calculated using a pre-trained random forest model, and the SHAP value analysis method is used to rank the importance of features and select key features for health index calculation.

Images