Intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment

Abstract

The invention belongs to the technical field of electrical equipment magnetic monitoring, and discloses an electrical equipment magnetic field interference intelligent monitoring and early warning sensing system, which establishes a magnetic field, working condition, environment and space four-dimensional data sensing system and matches a bimodal separation technology. Pure magnetic field interference, working condition correlation interference, environment induced interference and space correlation interference components can be effectively separated; the generated three-dimensional traceability map can visually present the position of an interference source, a propagation path and the priority of affected equipment, and then the accuracy of interference type identification and source positioning is greatly improved in combination with a feature library covering a large amount of historical normal operation data, so that fault pre-judgment is changed from passively coping with an existing fault into actively preventing and controlling risks in advance; a dual-path active suppression mechanism is formed by combining a software compensation mode of dynamically generating a magnetic field compensation current through a PID control algorithm with a tunable magnetic shielding device hardware adjustment means configured for high-priority equipment.

Description

Technical Field

[0001] This invention belongs to the field of magnetic monitoring technology for electrical equipment, specifically an intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment. Background Technology

[0002] With the deep integration of new power systems and Industry 4.0, the grid connection of new energy sources, the large-scale application of power electronic equipment, and the cluster deployment of equipment have become the mainstream in the industry. The complexity of the operating environment of electrical equipment continues to upgrade, exhibiting typical characteristics such as strong electromagnetic coupling, dynamic fluctuations in operating conditions, and superposition of multiple interference sources. Magnetic field interference is gradually evolving into a clustered risk that crosses equipment and regions, not only affecting the operating accuracy and service life of the equipment itself, but also potentially causing major safety accidents such as grid fluctuations and production interruptions. Existing magnetic field interference monitoring technologies mainly have the following technical problems: On the one hand, the coupling relationship between interference signals and equipment operating conditions and external environment has not been effectively separated. Interference signals from different sources are intertwined, making it impossible for the monitoring system to accurately identify the type of interference and locate the root cause device. The lack of traceability capability directly results in a low accuracy rate of fault prediction. Maintenance personnel can often only passively deal with faults that have already occurred, making it difficult to achieve early prevention and control.

[0003] On the other hand, existing technologies are mostly limited to the single link of monitoring and early warning. A complete closed loop of monitoring, early warning and disposal has not been formed. There is a lack of targeted active suppression measures and a mature cross-device cluster collaboration mechanism. Interference disposal often lags behind the spread of risk and cannot stop the chain reaction in time, which further amplifies the losses caused by interference. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment, the system comprising: The sensing module establishes a four-dimensional data acquisition system for magnetic field, operating conditions, environment, and space. The hardware adopts a TMR three-dimensional array, a GMR broadband sensor, a fiber optic grating operating condition sensor, and a UWB positioning module. The fiber optic grating sensor is used to capture the equipment load mutation rate and current harmonic distortion rate; and simultaneously acquires temperature and humidity spatial gradients, dust concentration distribution, and spatial information. Simultaneously, a magnetic field enhancement acquisition unit is configured: This unit is placed in the system shell, and circular fixing grooves are integrally cast on both sides of the inner wall of the shell. A coil is movably connected to the opposite side of the circular fixing groove. A magnetic rod with a diameter of 10mm, a length of 30mm, and a permeability of μ=1800±25% is fixed inside the coil. A coil with a wire diameter of 0.1mm is wound on the surface of the coil (number of turns 1500t-8000t, inductive reactance 28mH±15%-700mH±15%, DC impedance 150Ω±10%-550Ω±10%). A signal line connector is fixed on the back of the shell for transmitting the coil acquisition signal. Through FPGA timing control, the synchronization of magnetic field, operating condition fluctuations, environmental gradients and spatial data is achieved. The built-in adaptive sampling frequency adjustment mechanism increases the sampling frequency when the equipment load fluctuation exceeds 10% or the environmental temperature and humidity gradient exceeds 5℃ / meter to capture transient interference signals. The acquired four-dimensional data is transmitted to the feature extraction module.

[0006] Preferably, the feature extraction module receives four-dimensional data output by the perception module and uses a dynamically weighted PCA-mutual information entropy fusion algorithm to dynamically adjust the weights according to real-time operating conditions: when the equipment is in a high load range, a dynamic weight of 1.2 to 1.5 times is set for the current harmonic features; when the ambient temperature and humidity fluctuations exceed the normal threshold, a dynamic weight of 1.1 to 1.3 times is set for the temperature and humidity gradient features, thereby realizing the extraction of core related features. An integrated feature validity verification submodule compares the data in real time with a historical feature database of normal operating data to eliminate invalid features. On the hardware side, it is equipped with a low-power edge computing chip, a bandpass filter circuit, and a signal amplification circuit. The filter circuit includes a DC blocking circuit, an adjustable low-pass filter circuit, and a second-order high-pass filter circuit (which sequentially remove DC interference, adapt to the target frequency, and suppress high-frequency noise). The power supply module provides a stable low-voltage power supply to the filter circuit. The signal amplification circuit uses a low-noise instrumentation amplifier, integrating an adjustable gain module (gain range 10-1000 times) and an overvoltage protection circuit to amplify the filtered weak magnetic field feature signal. The raw data is cached on the edge node using the AES-256 encryption algorithm, and the purified correlation features and key anomaly data are uploaded to the separation and tracing module.

[0007] Preferably, the separation and tracing module receives the core associated features from the feature extraction module, combines the spatial and environmental data collected by the perception module, and adopts an improved LSTM-Attention and graph attention network dual-modal association separation network. The LSTM-Attention network focuses on the sensitive features of the working condition through the attention mechanism, inputting magnetic field features and dynamic correlation factors of the working condition, and separating the pure magnetic field interference component and the working condition-related interference component; the graph attention network combines the spatial topology relationship of the equipment, inputting environmental gradient features and spatial positioning information, and separating the environmentally induced interference component and the spatially related interference component. An association map of equipment cluster interference, links, and devices is established. The map has a built-in propagation attenuation coefficient and generates a three-dimensional source map of interference sources, propagation paths, and the priority of affected devices. The separated interference component data and the three-dimensional source map are transmitted to the prediction and early warning module.

[0008] Preferably, the prediction and early warning module establishes a dynamic prediction and early warning model based on the interference component data output by the separation and source tracing module, the three-dimensional source tracing map, and the operating condition and environmental parameters of the sensing module; it adopts a GPR-Transformer fusion prediction model, in which the model assigns weights to short-term and long-term predictions through an attention mechanism: the GPR model is used to process nonlinear fluctuation data and outputs a dynamic early warning baseline for fluctuations related to operating conditions and the environment; the Transformer model captures long-term time-series dependencies through multi-head attention and predicts the development trend of interference. A risk propagation prediction submodule is set up. Based on the propagation attenuation coefficient of the source map and combined with the equipment operating status, it predicts the devices to which the interference will spread and the propagation time. It outputs single device early warning information and cluster risk map. The risk map marks the device risk level with four colors: red, orange, yellow and green, and sets a propagation time axis. The early warning mechanism adopts a four-level logic and configures a dynamic adjustment rule for the early warning level. When the predicted interference propagation speed exceeds the preset threshold, the early warning level is automatically upgraded. Quantitative indicators, including interference intensity, propagation speed and the number of affected devices, are simultaneously output to the response module.

[0009] Preferably, the response module receives the warning level, risk spread path, and priority information of affected equipment output by the prediction and early warning module. Combined with the pure magnetic field interference component data from the separation and tracing module, it establishes a three-level closed-loop response system of active suppression, single-device protection, and cluster collaborative protection. Active suppression adopts two methods: software compensation and hardware adjustment. On the software side, based on the separated pure magnetic field interference component, a magnetic field compensation current is dynamically generated through a PID control algorithm to offset low-frequency magnetic field interference. On the hardware side, a tunable magnetic shielding device is configured for high-priority equipment to adapt to different intensity interference scenarios. The cluster collaboration mechanism prioritizes equipment based on load rate and power supply importance, allocates suppression resources according to priority, collects suppressed magnetic field data in real time through high-frequency magnetic field sensors, compares it with the predicted target value, and automatically adjusts compensation parameters or shielding effectiveness if the deviation exceeds the preset threshold. In the event of sudden pulse interference, the overall response time is controlled within 200 milliseconds, blocking the interference propagation path and transmitting the suppression effect data and equipment operating status to the digital twin operation and maintenance module.

[0010] Preferably, the digital twin operation and maintenance module integrates the data collected by the perception module, the feature data of the feature extraction module, the source map of the separation and tracing module, the risk information of the prediction and early warning module, and the handling effect data of the response module to build a cluster-level three-dimensional twin scene, integrating real-time data of magnetic field, working condition, environment and space in multiple dimensions. The twin scenario has a built-in virtual-real interactive simulation function to simulate the implementation effect of different suppression strategies. By comparing parameters including magnetic field distribution and equipment temperature rise, it outputs the optimal collaborative treatment plan, and the plan guides the active suppression module to adjust its strategy. It uses a time-series database to store equipment operation data, and generates the equipment magnetic field interference tolerance curve based on the fatigue damage accumulation model. Combined with real-time operation data, it predicts the remaining service life of the equipment. It supports remote collaborative operation and maintenance. Operation and maintenance personnel can click to mark faulty equipment in the twin scenario, and the system will automatically match operation and maintenance personnel with spare parts and generate the optimal driving route. The generated operation and maintenance decision data is fed back to the prediction and early warning module to optimize the prediction model parameters.

[0011] Preferably, the fault-tolerant module provides energy supply and fault tolerance guarantee, and receives real-time operating status data of each module; the energy supply adopts a three-mode scheduling scheme of solar energy, electromagnetic induction power generation and supercapacitor energy storage. Configure an energy demand forecasting submodule to dynamically allocate the proportion of three power supply modes based on equipment operation plans and meteorological data; when a fault is detected, automatically trigger the switchover backup mechanism and adjust the energy dispatch strategy.

[0012] The beneficial effects of this invention are as follows: 1. This invention establishes a four-dimensional data perception system encompassing magnetic field, operating conditions, environment, and space. Combined with a dual-modal separation technique consisting of an improved LSTM-Attention and graph attention network, it effectively separates pure magnetic field interference, operating condition-related interference, environment-induced interference, and spatially related interference components. The generated three-dimensional source map can intuitively present the location of the interference source, the propagation path, and the priority of the affected equipment. Furthermore, by combining it with a feature library covering a large amount of historical normal operation data, it significantly improves the accuracy of interference type identification and root cause localization, enabling fault prediction to shift from passively responding to existing faults to proactively preventing and controlling risks in advance, effectively avoiding the problem of cascading interference across equipment clusters.

[0013] 2. This invention establishes a closed-loop response system encompassing monitoring, early warning, suppression, verification, and iteration. On one hand, it dynamically generates a software compensation method for magnetic field compensation current through a PID control algorithm, combined with hardware adjustment of a tunable magnetic shielding device configured for high-priority equipment, forming a dual-path active suppression mechanism that can block the spread of sudden pulse interference in a very short time. On the other hand, it prioritizes equipment based on load rate and power supply importance, achieving intelligent allocation of suppression resources. Combined with dynamic early warning baselines and risk spread prediction functions, it significantly reduces false alarms, upgrading from single-device protection to collaborative protection of equipment clusters, thus solving the industry pain points of delayed interference handling and lack of collaborative capabilities in traditional technologies.

[0014] 3. This invention adopts a three-mode energy supply scheme of solar energy, electromagnetic induction power generation, and supercapacitor energy storage, coupled with a multi-level redundant design covering the sensing layer, communication layer, and data processing layer, as well as an intelligent self-diagnostic function for real-time monitoring of system status. Even in complex industrial environments such as extreme low temperatures and no light, it can still ensure continuous and stable power supply to the system, avoiding monitoring interruptions due to energy outages or single-point failures. The digital twin operation and maintenance module integrates full-link data, supports virtual-real interactive simulation testing of the effects of different interference suppression strategies, and can also realize remote collaborative operation and maintenance. Combined with fatigue damage accumulation models, it predicts the remaining service life of equipment, which not only shortens the time for fault diagnosis and handling, but also reduces unnecessary equipment replacement and lowers operation and maintenance costs. Attached Figure Description

[0015] Figure 1 This is a flowchart of the intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment according to the present invention; Figure 2 This is a flowchart of the multi-dimensional data acquisition and feature purification process of this invention; Figure 3 This is a flowchart of the interference separation and dynamic early warning process of the present invention; Figure 4 This is a flowchart of the interference suppression and operation and maintenance closed loop of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figures 1 to 4 As shown, this embodiment of the invention provides an intelligent monitoring and early warning sensing system for magnetic field interference in electrical equipment. The system includes: The core function of the sensing module is to establish a four-dimensional data acquisition system for magnetic field, operating conditions, environment, and space. The hardware adopts a TMR three-dimensional array, a GMR broadband sensor, a fiber optic grating operating condition sensor, and a UWB positioning module. The fiber optic grating sensor has anti-electromagnetic interference characteristics and is used to capture the equipment load change rate and current harmonic distortion rate. It also simultaneously acquires spatial gradients of temperature and humidity, dust concentration distribution, and spatial information output by the UWB positioning module. Through FPGA timing control, the synchronization of magnetic field, operating condition fluctuation, environmental gradient and spatial data is achieved, with the synchronization error controlled within 3 milliseconds. The built-in adaptive sampling frequency adjustment mechanism switches the sampling frequency from 1kHz to 5kHz when the equipment load fluctuation exceeds 10% or the environmental temperature and humidity gradient exceeds 5℃ / meter, ensuring that transient interference signals such as lightning strikes and equipment start-up and shutdown can be accurately captured. The 1MHz detection bandwidth of the GMR wideband sensor can cover high-frequency electromagnetic pulse interference. The flexible packaging design adapts to the non-contact installation requirements of electrical equipment of different shapes, such as high-voltage switchgear and transformer body, and the collected four-dimensional data is transmitted to the feature extraction module.

[0018] The feature extraction module receives magnetic field, operating conditions, environment, and four-dimensional spatial data output by the sensing module. It employs a dynamically weighted PCA-mutual information entropy fusion algorithm to dynamically adjust the weights based on real-time operating conditions: when the equipment is in a high-load range (load rate exceeding 80%), a dynamic weight of 1.2 to 1.5 times is set for the current harmonic features; when the ambient temperature and humidity fluctuations exceed the normal threshold (±2℃ / ±5%RH), a dynamic weight of 1.1 to 1.3 times is set for the temperature and humidity gradient features, thereby achieving accurate extraction of core related features. An integrated feature validity verification submodule compares the data in real time with a historical feature database covering 100,000 sets of normal operation data to eliminate invalid features such as instantaneous dust accumulation and accidental electromagnetic pulses, ensuring the accuracy of core related features. On the hardware side, it is equipped with a low-power edge computing chip (such as STM32H743VIT6) and a hardware-level dynamic filtering circuit, using a Butterworth filter with a dynamically adjustable cutoff frequency, and the single-frame data processing time is controlled within 0.2 milliseconds. The raw data is cached at the edge node using the AES-256 encryption algorithm, and the purified correlation features and key anomaly data are uploaded to the separation and tracing module, which greatly reduces bandwidth usage.

[0019] The separation and tracing module receives the core associated features from the feature extraction module, combines the original spatial positioning and environmental gradient data collected by the perception module, and adopts an improved LSTM-Attention and graph attention network dual-modal association separation network. The network training relies on more than 5,000 sets of historical interference case data, including types such as power grid harmonics, equipment failures, and external electromagnetic radiation. The LSTM-Attention network focuses on the sensitive features of the working conditions through the attention mechanism, inputting magnetic field features and dynamic correlation factors of the working conditions, and separating the pure magnetic field interference component and the working condition-related interference component; the graph attention network combines the spatial topology relationship of the device, inputting environmental gradient features and spatial positioning information, and separating the environmentally induced interference component and the spatially related interference component, which can effectively deal with scenarios with multiple interference sources superimposed. Each set of data in the historical interference case data includes synchronously collected four-dimensional raw data, interference type labels (pure magnetic field / operating condition related / environmentally induced / spatial related) annotated by the expert system, and interference source device identifiers, used for supervised learning; the improved LSTM-Attention network introduces an attention mechanism in the last hidden layer on the basis of the standard LSTM, calculates the attention weight of the feature at each time step, and sets the weight threshold to 0.1. Features below this threshold are regarded as noise and suppressed; the graph attention network is constructed based on the adjacency matrix of the device space. The feature vector of the node in the graph consists of the environmental gradient and spatial positioning information of the device. The edge weight is initialized to the reciprocal of the distance between devices, and feature aggregation is performed through two layers of graph attention layers.

[0020] An association map of equipment cluster interference, links, and equipment is established. The map has a built-in propagation attenuation coefficient, which is calculated by taking into account factors such as air medium, metal shield thickness, and equipment spacing. A three-dimensional source map of interference sources, propagation paths, and the priority of affected equipment is generated and displayed in the form of a heat map overlaid with an equipment topology map. The separated interference component data and the three-dimensional source map are transmitted to the prediction and early warning module.

[0021] The propagation attenuation coefficient The calculation formula is:

[0022] in, The distance between devices is in meters and is obtained from the UWB positioning module. The average thickness of the main shielding layer between equipment rooms, in millimeters, is a preset parameter. The air attenuation factor has an empirical value of 0.02. The metal shielding attenuation factor has an empirical value of 0.5; this coefficient is eventually normalized to the range of 0-1 and is used to quantify the degree of attenuation of interference propagating between devices.

[0023] The pure magnetic field interference component separated by dual-mode correlation refers to the magnetic field interference generated solely by external electromagnetic radiation, internal equipment faults, and other factors after eliminating the coupling effects of equipment operating condition fluctuations and environmental parameter changes; its calculation formula is as follows: ; In the formula: This represents the pure magnetic field interference component, with units of μT and a value range of 0~100μT; The weight matrix of the LSTM-Attention network (dimension 1xn, where n is the dimension of the input features, and n is determined according to the number of parameters collected by the sensor) is obtained by training with 5000+ sets of historical interference case data covering types such as power grid harmonics, equipment failures, and external electromagnetic radiation. The magnetic field feature vector has a dimension of nx1 and includes parameters such as the three-dimensional magnetic field strength (X / Y / Z axes) acquired by the TMR three-dimensional array and the magnetic field frequency distribution acquired by the GMR broadband sensor. The data sampling frequency is consistent with the operating conditions and environmental parameters. The operating condition feature vector has a dimension of nx1 and includes parameters such as the equipment load mutation rate and current harmonic distortion rate collected by the fiber Bragg grating sensor. The load rate monitoring range covers 0~100%. The environmental feature vector has a dimension of nx1 and includes parameters such as the spatial gradient of temperature and humidity collected by the temperature and humidity sensor and the dust concentration distribution collected by the dust sensor. The temperature and humidity gradient monitoring accuracy is ±0.1°C / m and ±0.5%RH / m. This represents the operating condition correlation factor, with a value range of 0.1-0.3, used to quantify the coupling degree of operating condition parameters to magnetic field interference. When the equipment load rate exceeds 80%, the system automatically increases it to 0.3 to enhance the separation capability of high operating condition coupling interference; when the load rate is below 30%, it drops to 0.1 to reduce over-correction in low operating condition scenarios. This represents the environmental correlation factor, with a value range of 0.05-0.2, used to quantify the coupling degree of environmental parameters to magnetic field interference. When temperature and humidity fluctuations exceed ±2°C / ±5%RH, it automatically increases to 0.2 to focus on separating environmentally induced interferences. When the fluctuations are within the normal range, it decreases to 0.05 to reduce the redundant impact of the environment on the separation results. This represents the separation bias term, in μT, with a value range of -2 to 2μT. It is corrected in real time by the graph attention network based on the spatial topology of the devices (such as the spacing between devices and the thickness of the metal shielding layer), mainly to compensate for the differences in magnetic field interference coupling caused by different spatial locations.

[0024] The prediction and early warning module establishes a dynamic prediction and early warning model based on the interference component data output by the separation and source tracing module, the three-dimensional source tracing map, and the operating condition and environmental parameters of the sensing module. It adopts a GPR-Transformer fusion prediction model, in which the model assigns weights to short-term and long-term predictions through an attention mechanism: the GPR model is used to process nonlinear fluctuation data and outputs a dynamic early warning baseline for the fluctuations related to operating conditions and environment within 10 minutes; the Transformer model captures long-term time-series dependencies through multi-head attention and predicts the development trend of interference within 1 to 2 hours. A risk propagation prediction submodule is set up. Based on the propagation attenuation coefficient of the source map and combined with the equipment operating status (such as load rate and temperature rise), it predicts the equipment to which the interference will spread and the propagation time. It outputs single-device early warning information and cluster risk map. The risk map marks the equipment risk level with four colors: red, orange, yellow and green, and sets a propagation time axis. The early warning mechanism adopts a four-level logic and configures a dynamic adjustment rule for the early warning level. When the predicted interference propagation speed exceeds 0.5 meters per minute, the early warning level is automatically upgraded by one level. If the current early warning is already at level one, the highest level is maintained and an emergency operation and maintenance dispatch command is triggered simultaneously. Quantitative indicators, including interference intensity, propagation speed and number of affected equipment, are output to the response module simultaneously.

[0025] The four-level logic of the early warning mechanism is as follows: Level 1: Interference intensity exceeds 50 μT; Level 2: Interference intensity is between 30 and 50 μT; Level 3: Interference intensity is between 10 and 30 μT; Level 4: Interference intensity is below 10μT.

[0026] GPR-Transformer Fusion Dynamic Early Warning Baseline Calculation Formula: ; In the formula: The dynamic early warning baseline at time t is expressed in μT and is updated in real time according to equipment operating conditions, environmental parameters and interference propagation status. The update cycle is synchronized with the data acquisition cycle. The weight coefficients of the Gaussian process regression model range from 0.4 to 0.6 and are used to fit the impact of short-term (within 10 minutes) coupling fluctuations of operating conditions and environment on the early warning baseline. When the load / environment fluctuates drastically, such as when the load mutation rate is >5% / s or the temperature and humidity gradient is >2°C / m, the value approaches 0.6 to enhance the short-term fitting weight. This represents the short-term warning baseline output by the GPR model, measured in μT. It excels at handling nonlinear, small-sample, coupled fluctuation data of operating conditions and the environment, with a fitting error controlled within ≤8%, ensuring the accuracy of the short-term baseline. This represents the long-term warning baseline output by the Transformer model, measured in μT. It captures long-term temporal dependencies within 1–2 hours using a multi-head attention mechanism, predicting the long-term impact of disturbance development trends on the baseline, with a prediction error ≤10%. This represents the propagation attenuation correction coefficient, with a value range of 0.02-0.05. It is derived from the propagation attenuation coefficient in the interference, link, and device association diagram (calculated by taking into account factors such as air medium and metal shield thickness), and reflects the correction effect of the attenuation of interference propagating from the source device to the current monitoring point on the baseline. This indicates the interference propagation distance, in meters (m), which is the straight-line distance between the current monitoring point and the interference source device, collected by the UWB positioning module. The greater the distance, the greater the attenuation correction, to avoid false alarms caused by propagation attenuation.

[0027] The response module receives the warning level, risk spread path, and priority information of affected equipment output by the prediction and early warning module. Combined with the pure magnetic field interference component data from the separation and tracing module, it establishes a three-level closed-loop response system of active suppression, single-device protection, and cluster collaborative protection. Active suppression adopts two methods: software compensation and hardware adjustment. On the software side, based on the separated pure magnetic field interference components, a magnetic field compensation current is dynamically generated through a PID control algorithm, with an adjustment step of 0.01A and an accuracy control within ±0.1A, to offset low-frequency magnetic field interference. On the hardware side, a tunable magnetic shielding device is configured for high-priority equipment (such as main transformers and main switches). The shielding effectiveness is precisely adjusted between 20 and 40 dB by electrically adjusting the magnetic permeability to adapt to different intensity interference scenarios. The tunable magnetic shielding device is a composite shielding structure based on a DC bias magnetic field. Its core includes a shell made of a high-permeability material (such as permalloy) and a built-in excitation coil. By adjusting the DC current input to the excitation coil, the magnetic saturation of the shield is changed, thereby achieving continuous and reversible electric adjustment of its equivalent permeability and shielding effectiveness. The system outputs a 4-20mA control signal to the current source to drive the tunable magnetic shielding device according to the warning level. The DC current adjustment range is 0-2A, and the shielding effectiveness is 20-40dB. 4mA corresponds to 0A DC current (shielding effectiveness 20dB), and 20mA corresponds to 2A DC current (shielding effectiveness 40dB). The signal and current have a linear proportional relationship. The cluster collaboration mechanism prioritizes equipment based on its load rate and power supply importance (e.g., equipment supplying power to hospitals and data centers is classified as level one), and allocates suppression resources according to priority to avoid resource waste. The response process of the cluster collaboration mechanism includes data monitoring, risk warning, interference suppression, effect verification, and parameter iteration. It collects the suppressed magnetic field data in real time through a high-frequency magnetic field sensor with a sampling rate of 10kHz and compares it with the predicted target value. If the deviation exceeds 10%, it automatically adjusts the compensation parameters or shielding effectiveness. Equipment with a load rate exceeding 80% is considered high priority. In the event of sudden pulse interference, the overall response time is controlled within 200 milliseconds, effectively blocking the interference propagation path and transmitting the suppression effect data and equipment operating status to the digital twin operation and maintenance module.

[0028] PID active suppression compensation current calculation formula: ; In the formula: This represents the compensation current in the kth control cycle, in A, which is output to the magnetic field compensation coil to cancel out pure magnetic field interference. The current accuracy is controlled within ±0.1A. This represents the proportionality coefficient, which ranges from 5 to 8 and is dynamically adjusted according to the intensity of pure magnetic field interference. This represents the integral coefficient, with a value ranging from 0.1 to 0.3, used to eliminate static interference deviations; The control period is expressed in seconds and is synchronized with the sampling period of the high-frequency magnetic field sensor to ensure that the compensation current matches the changes in disturbance in real time. This represents the differential coefficient, with a value range of 0.8-1.2. It is used to suppress the overshoot phenomenon of the compensation current and avoid secondary fluctuations in the magnetic field caused by sudden rises and falls in the current. This indicates the current adjustment step size, which is fixed at 0.01A. It is the smallest adjustment unit for the compensation current, ensuring that the current adjustment accuracy matches the response accuracy of the compensation coil. This represents the quantization value of pure magnetic field interference, indicating the quantization of pure magnetic field interference components. The result after rounding quantization with a step size of 0.1 μT; for example hour, Quantification ensures that the compensation current is consistent with... The adjustment precision is matched; The proportional coefficient of the PID control algorithm Integral coefficient Differential coefficients Using interference intensity The system uses a lookup table method for adaptive tuning; as an example tuning rule, the system sets parameters according to the warning level: when (Level 1 Warning) ; when (Level 2 warning) ; when (At level three or below warning): ; The system uses the above parameter set to achieve rapid and stable compensation for interference of different intensities.

[0029] The interference deviation in the k-th period is expressed in μT and is calculated using the following formula: ; The safe magnetic field threshold for the equipment is usually set to 10 μT, but it can be adjusted according to the type of equipment. For example, the threshold for the main transformer is set to 8 μT, and for auxiliary equipment it is set to 12 μT.

[0030] The digital twin operation and maintenance module integrates the data collected by the perception module, the feature data from the feature extraction module, the source map from the separation and tracing module, the risk information from the prediction and early warning module, the handling effect data from the response module, and the operating status and energy scheduling data from the fault tolerance module. It builds a cluster-level 3D twin scene based on Unreal Engine, with a stable scene rendering frame rate of 30fps to ensure real-time data mapping. It integrates real-time data from multiple dimensions of magnetic field, operating conditions, environment, and space, and synchronizes data with physical devices through data interfaces to ensure that the twin and the physical devices are in the same state. The twin scenario incorporates a virtual-real interactive simulation function to simulate the implementation effects of different suppression strategies (such as combined strategies like software compensation current of 0.5A + hardware shielding effectiveness of 25dB, software compensation current of 1A + hardware shielding effectiveness of 35dB, etc.). By comparing parameters such as magnetic field distribution and equipment temperature rise, it outputs the optimal collaborative treatment plan, which in turn guides the active suppression module to adjust its strategy. It uses a time-series database (such as InfluxDB) to store equipment operating data, and generates equipment magnetic field interference tolerance curves based on a fatigue damage accumulation model. Combined with real-time operating data, it predicts the remaining service life of the equipment. The fatigue damage accumulation model adopts Miner's linear accumulation rule, which describes the damage caused by a single magnetic field disturbance event. The calculation formula is: ; in, For the equipment in a magnetic field strength of Exposure time under interference, in hours; The magnetic field strength is determined by the equipment tolerance curve. The fatigue life, expressed in hours; the total cumulative damage to the equipment. ,when When the equipment reaches the end of its service life, the tolerance curve is determined. The data was obtained by fitting accelerated aging test data of the equipment, and the form is: ,in , These are characteristic constants related to the equipment's materials and structure.

[0031] It supports remote collaborative operation and maintenance. Operation and maintenance personnel can click to mark faulty equipment in the twin scenario. The system automatically matches operation and maintenance personnel with spare parts and generates the optimal driving route to improve the efficiency of fault diagnosis and operation and maintenance. The generated operation and maintenance decision data is fed back to the prediction and early warning module to optimize the prediction model parameters.

[0032] The fault-tolerant module is used to provide energy supply and fault tolerance guarantee, and at the same time receive real-time operating status data of each module; the energy supply adopts a three-mode scheduling scheme of solar energy, electromagnetic induction power generation and supercapacitor energy storage. Equipped with a 200W high-efficiency monocrystalline silicon solar panel, it prioritizes power supply to the sensing module and feature extraction module in scenarios with sufficient sunlight. The electromagnetic induction power harvesting unit obtains electrical energy by coupling the magnetic field of the equipment cable, adapting to the energy consumption requirements of the response module and digital twin operation and maintenance module when the equipment is under high load. The 500F large-capacity supercapacitor serves as an energy storage unit, with a capacity retention rate of ≥80% at extreme low temperatures (-30℃). Based on this capacity redundancy design, it can ensure that all modules can be continuously powered for no less than 72 hours in extreme low temperature (-30℃) and no-sunlight scenarios.

[0033] Configure an energy demand forecasting submodule to dynamically allocate the proportion of the three power supply methods based on the equipment operation plan and meteorological data for the next 24 hours, so as to avoid energy waste; the equipment operation plan includes start-up and shutdown times, load changes, etc. It integrates intelligent self-diagnostic function, monitors the status of each module at a frequency of 1 second / time, and automatically triggers the switching backup mechanism when a fault is detected. For example, if the sensor of the sensing module fails, the system will automatically switch to a backup sensor of the same model and the same sampling accuracy within 1 second; at the same time, it will feed back to the digital twin module to update the health status of the equipment and adjust the energy scheduling strategy, such as temporarily increasing the power supply of the backup component of the faulty module.

[0034] Three-mode energy dispatch and distribution: ; In the formula: This represents the total system power supply in watts (W). The multi-dimensional collaborative sensing array module consumes approximately 5W, the edge-end correlation feature extraction module approximately 8W, and the device cluster digital twin module approximately 15W. Total energy consumption dynamically changes based on the module's operating status. Energy consumption fluctuations need to be matched in real time; This indicates the solar energy conversion efficiency, which is fixed at ≥23% and is determined by the hardware performance of the solar panel, and is not affected by the external environment. This indicates the rated output power of the solar panel, which is fixed at 200W. It is the main power source in scenarios with sufficient sunlight, and the output power changes dynamically with the intensity of sunlight. Scenarios with sufficient sunlight are defined as sunny noon or sunlight intensity >8000 lux. This represents the light intensity coefficient, with a value ranging from 0 to 1. It is obtained by normalizing the light intensity collected in real time by the light sensor. The value is 1 when it is sunny at noon or when the light intensity is >8000 lux, and 0 when there is no light. It is used to quantify the impact of light on solar power supply. No light is defined as nighttime or cloudy / rainy weather with a light intensity <500 lux. This represents the electromagnetic induction power extraction efficiency, with a value ranging from 0.15 to 0.25. It is positively correlated with the current in the equipment cable; the higher the current, the higher the power extraction efficiency, which can reach up to 0.25; the lower the current, the lower the power extraction efficiency, which can be as low as 0.15. This indicates the rated power of the electromagnetic induction power unit, which is fixed at 100W. It is specifically designed to meet the energy consumption requirements of the device under high load scenarios. When the device is running under high load, the output power of the power unit approaches 100W. This represents the equipment load factor, with a value ranging from 0 to 1. It is obtained by normalizing the equipment load rate collected by the fiber Bragg grating condition sensor. The value is 1 when the load rate is 100% and 0.3 when the load rate is less than 30%. It is used to quantify the impact of equipment load on electromagnetic induction power extraction. This indicates the discharge efficiency of the supercapacitor, which is fixed at ≥90%. It is a core hardware parameter of the 500F high-capacity supercapacitor and is not affected by temperature. This indicates the rated discharge power of the supercapacitor, which is fixed at 150W, serving as a backup power source for extreme scenarios. This represents the low energy priority coefficient, with a value range of 0-1. It takes a value of 1 only when the solar power supply capability is weak or the electromagnetic induction power extraction capability is weak, triggering supercapacitor power supply; in other scenarios, it takes a value of 0. Positioned as having weak solar power supply capability That is, light intensity <2000 lux; weak electromagnetic induction power extraction capability is defined as... That is, the load rate is less than 30%.

[0035] The energy demand forecasting submodule dynamically allocates power supply ratios based on the following rules: Priority 1 (Solar): When Use solar energy as much as possible, and follow the instructions. Calculate the available power.

[0036] Priority 2 (Electromagnetic Induction): When the total power consumption of the system... Exceeding the power supply capacity of solar energy, and At that time, electromagnetic induction is activated to draw power to fill the gap.

[0037] Priority 3 (Supercapacitor): Only when and Only then does it mainly rely on supercapacitors for power supply.

[0038] The intelligent self-diagnostic function determines that a sensor is faulty if it detects that the sensor data is zero or exceeds the range for three consecutive sampling cycles. It then switches the data source to a preset backup sensor channel within one second and issues an alarm in the digital twin module.

[0039] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0040] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent monitoring and early warning sensing system for magnetic field interference in electrical equipment, characterized in that: The system includes: Sensing module: Employs various types of sensors to collect various parameters related to device operation and dynamically adjusts the sampling frequency; Feature extraction module: Receives data collected by the perception module, extracts the core related features from the data, removes invalid information by comparing with the historical feature library, and encrypts and caches the original data; Separation and source tracing module: Receives purified correlation feature data, separates different types of interference components through a dual-modal correlation separation network, and finally generates an interference source tracing map; Prediction and early warning module: predicts the development trend of interference through a prediction model, generates early warning information and cluster risk map; outputs quantitative indicators through four-level early warning logic and dynamic adjustment rules for early warning levels; Response module: Employs a combination of software compensation and hardware adjustment to suppress magnetic field interference; allocates suppression resources according to device priority; Digital twin operation and maintenance module: Creates a three-dimensional twin scene, with built-in virtual-real interaction simulation function to realize remote collaborative operation and maintenance; Fault-tolerant module: Provides multi-mode energy supply solutions, diagnoses faults in real time and adjusts strategies accordingly.

2. The intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment according to claim 1, characterized in that: The sensing module establishes a four-dimensional data acquisition system for magnetic field, operating conditions, environment, and space. The hardware adopts a TMR three-dimensional array, a GMR broadband sensor, a fiber optic grating operating condition sensor, and a UWB positioning module. The fiber optic grating sensor is used to capture the equipment load mutation rate and current harmonic distortion rate; and simultaneously acquires temperature and humidity spatial gradients, dust concentration distribution, and spatial information. The synchronization of four-dimensional data is achieved through timing control. The built-in adaptive sampling frequency adjustment mechanism increases the sampling frequency to capture transient interference signals when the equipment load fluctuates or the ambient temperature and humidity gradient exceeds the preset range. The collected data is then transmitted to the feature extraction module.

3. The intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment according to claim 2, characterized in that: The feature extraction module receives four-dimensional data output by the perception module, and uses a dynamically weighted PCA-mutual information entropy fusion algorithm to dynamically adjust the weights according to real-time operating conditions to extract core related features. An integrated feature validity verification submodule is used to eliminate invalid features by comparing them with the historical feature library in real time, and a low-power edge computing chip and hardware-level dynamic filtering circuit are configured. The raw data is cached at the edge nodes using an encryption algorithm, and the purified correlation features and key anomaly data are uploaded to the separation and tracing module.

4. The intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment according to claim 3, characterized in that: The separation and tracing module receives the core associated features from the feature extraction module, combines the spatial and environmental data collected by the perception module, and uses an improved LSTM-Attention and graph attention network dual-modal association separation network to separate the pure magnetic field interference component from the working condition-related interference component, and separate the environmentally induced interference component from the spatially related interference component. An interference correlation map of the device cluster is established, with a built-in propagation attenuation coefficient, and a three-dimensional source tracing map is generated that includes the interference source, propagation path, and priority of the affected devices; the separated interference component data and the three-dimensional source tracing map are transmitted to the prediction and early warning module.

5. The intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment according to claim 4, characterized in that: The prediction and early warning module establishes a dynamic prediction and early warning model based on interference component data, three-dimensional source map, and operating condition and environmental parameters of the sensing module; it adopts the GPR-Transformer fusion prediction model to assign weights to short-term and long-term predictions; it outputs a dynamic early warning baseline for fluctuations in operating conditions and environment, and predicts the development trend of interference. A risk propagation prediction submodule is set up. Based on the propagation attenuation coefficient of the source map and combined with the equipment operating status, it predicts the equipment to which the interference will spread and the propagation time, and outputs single-device early warning information and cluster risk map; the risk map uses four colors to mark the equipment risk level; The early warning mechanism adopts a four-level logic and is configured with dynamic adjustment rules for the early warning level; it also outputs quantitative indicators, including interference intensity, spread speed and number of affected devices, to the response module simultaneously.

6. The intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment according to claim 5, characterized in that: The response module receives the warning level, risk spread path, and priority information of affected equipment output by the prediction and early warning module. Combined with the pure magnetic field interference component data from the separation and tracing module, it establishes a three-level closed-loop response system of active suppression, single-device protection, and cluster collaborative protection. Active suppression adopts two methods: software compensation and hardware adjustment. The software dynamically generates magnetic field compensation current based on the separated pure magnetic field interference components. The hardware configures tunable magnetic shielding devices for high-priority equipment. The cluster collaboration mechanism prioritizes equipment based on load rate and power supply importance, allocates suppression resources according to priority, and compares the suppressed magnetic field data with the predicted target value in real time through a high-frequency magnetic field sensor. If the deviation exceeds the preset threshold, the compensation parameters or shielding effectiveness are automatically adjusted. In the event of sudden pulse interference, the interference propagation path is quickly blocked, and the suppression effect data and equipment operating status are transmitted to the digital twin operation and maintenance module.

7. The intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment according to claim 6, characterized in that: The digital twin operation and maintenance module integrates the data collected by the perception module, the feature data of the feature extraction module, the source map of the separation and tracing module, the risk information of the prediction and early warning module, and the handling effect data of the response module to build a cluster-level three-dimensional twin scene and integrate multi-dimensional real-time data. The twin scenario has a built-in virtual-real interactive simulation function to simulate the implementation effect of different suppression strategies, output the optimal collaborative handling plan, and the plan guides the active suppression module to adjust the strategy. Based on the fatigue damage accumulation model, it predicts the remaining service life of the equipment; it supports remote collaborative operation and maintenance, and the generated operation and maintenance decision data is fed back to the prediction and early warning module.

8. The intelligent monitoring and early warning sensing system for magnetic field interference of electrical equipment according to claim 7, characterized in that: The fault-tolerant module provides a three-mode energy supply solution: solar energy, electromagnetic induction power generation, and supercapacitor energy storage. Configure an energy demand forecasting submodule to dynamically allocate the proportion of power supply modes; when a fault is detected, automatically trigger the switchover backup mechanism and adjust the energy dispatch strategy.

Images