Low-voltage switch cabinet intelligent monitoring system and method based on Internet of Things

Abstract

The invention discloses a low-voltage switch cabinet intelligent monitoring system and method based on the Internet of Things, particularly relates to the technical field of power equipment state monitoring, and is used for solving the problem of monitoring data distortion caused by accumulated damage of a sensor due to arc electromagnetic pulses in the prior art. A current and voltage waveform is collected and an arc event is identified; calculating an electromagnetic coupling weight coefficient according to the arc position and the sensor coordinate; capturing working parameters of the sensor, comparing the working parameters with historical reference values, and calculating offset of key parameters; information entropy mutation of the offset of the same type of sensor groups is analyzed, and a nonlinear damage mark is triggered; based on the offset and the damage mark, establishing a segmentation mapping relation and outputting an accumulated damage evaluation value; correcting monitoring data according to the evaluation value and inputting the monitoring data into a fault diagnosis model to execute early warning; quantitative evaluation and data self-correction of the hidden damage of the sensor are realized, and the accuracy of fault early warning is guaranteed from the source.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment status monitoring, and more specifically, to an intelligent monitoring system and method for a low-voltage switchgear based on the Internet of Things. Background Art

[0002] Low-voltage switchgear, as a critical component of the power distribution system, has a direct impact on power supply reliability through its operational status. With the widespread adoption of IoT technology, remote intelligent monitoring achieved through the deployment of multiple sensors within the switchgear (such as temperature, current, and partial discharge monitoring devices) has become a mainstream solution in the industry. Existing technologies typically use wired or common IoT communication protocols (such as Modbus and MQTT) to transmit sensor data to cloud platforms, combining data analysis algorithms to provide fault warnings. Especially for transient, high-risk events such as arc faults, the system relies on high-precision sensors to capture transient characteristics and trigger protection mechanisms. Current technical solutions focus on optimizing arc detection algorithms, improving sampling frequency, and enhancing real-time communication to ensure a rapid response to switchgear faults.

[0003] However, existing monitoring solutions overlook a flaw: the strong electromagnetic pulse (EMP) generated by arc faults can penetrate the sensor's internal circuitry through conduction or radiation coupling, causing irreversible, cumulative damage to microelectronic devices. While this damage doesn't initially cause device failure, it can cause sensor measurement parameters (such as temperature readings and current waveforms) to slowly drift, and conventional calibration cannot detect this latent degradation. The direct consequence is that the monitoring system's fault diagnosis and early warning, based on distorted data, will gradually deviate from the actual operating conditions, ultimately leading to missed alarms or misjudgments. The risk of monitoring equipment failure caused by the monitored object (arc fault) cannot be eliminated at the source of the data. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a low-voltage switchgear intelligent monitoring system and method based on the Internet of Things to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: The intelligent monitoring method of low-voltage switchgear based on the Internet of Things includes: S1. Collect current and voltage waveform data. When the current mutation rate exceeds a preset mutation threshold and is accompanied by high-frequency oscillation characteristics, an arc event is determined to have occurred. S2. Based on the arc location and sensor installation coordinates, extract the cabinet skin effect characteristics and the parasitic capacitance of the sensor power supply circuit, and calculate the coupling weight coefficient of the electromagnetic pulse to each sensor; S3. Capturing the operating parameters of the corresponding sensor when the arc event occurs based on the coupling weight coefficient, and comparing the operating parameters with the historical reference parameters of the corresponding sensor to calculate the key parameter offset; S4. Calculate the information entropy value of the offset of the key parameters of the same type of sensor group. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, trigger the nonlinear damage mark. S5. Establish a segmented mapping relationship based on the key parameter offset and the nonlinear damage marker, and output the cumulative damage assessment value of the sensor group; S6. Correct the sensor monitoring data according to the cumulative damage assessment value and input it into the fault diagnosis model to execute early warning.

[0006] Furthermore, current and voltage waveform data are collected. When the current mutation rate exceeds a preset mutation threshold and is accompanied by high-frequency oscillation characteristics, an arc event is determined to have occurred, including: Calculating a current differential sequence based on current waveform data, and identifying intervals in the current differential sequence that continuously exceed a preset mutation threshold as arc event determination intervals; Analyze the spectral characteristics of the voltage waveform within the arc event determination interval to confirm the presence of high-frequency oscillation components that meet the arc characteristics; The spatial coordinates of the arc occurrence position are determined according to the time domain center position of the arc event judgment interval.

[0007] Furthermore, based on the arc location and sensor installation coordinates, the cabinet skin effect characteristics and the parasitic capacitance of the sensor power supply circuit are extracted, and the coupling weight coefficient of the electromagnetic pulse to each sensor is calculated, including: Based on the spatial coordinates of the arc occurrence location and the spatial distance vectors of the sensor installation coordinates, the skin depth distribution characteristic values ​​of the electromagnetic wave propagation path on the cabinet surface are calculated; Measure the parasitic capacitance parameters of each sensor power supply circuit using an impedance analyzer; The skin depth distribution characteristic value and parasitic capacitance parameters of each sensor are input into the preset electromagnetic coupling equation, and the coupling weight coefficient of each sensor is output.

[0008] Furthermore, the operating parameters of the corresponding sensor when the arc event occurs are captured based on the coupling weight coefficient, and the operating parameters are compared with the historical benchmark parameters of the corresponding sensor to calculate the key parameter offset, including: Screening sensors whose coupling weight coefficients exceed a preset weight threshold as target sensors; When an arc event occurs, the real-time values ​​of the target sensor's operating parameters are recorded synchronously; Retrieve the corresponding parameter reference value from the historical reference parameter database of the target sensor; The relative offset between the real-time value of the working parameter and the parameter reference value is calculated as the key parameter offset.

[0009] Furthermore, the information entropy value of the offset of the key parameters of the same type of sensor group is calculated. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, the nonlinear damage mark is triggered, including: Normalize the key parameter offset of each sensor in the same type of sensor group to generate a normalized offset probability distribution; The Shannon information entropy value is calculated based on the normalized offset probability distribution as the information entropy value at the current moment; Obtain the information entropy value of the same group at the previous moment, and calculate the absolute difference between the current information entropy value and the previous information entropy value as the entropy mutation amplitude; When the entropy value mutation amplitude exceeds the preset entropy change threshold, a nonlinear damage mark trigger signal is generated.

[0010] Furthermore, a segmented mapping relationship is established based on the key parameter offset and the nonlinear damage marker, and the cumulative damage assessment value of the sensor group is output, including: Selecting a linear mapping mode or a nonlinear mapping mode according to the state of a nonlinear damage marker trigger signal; In the linear mapping mode, the cumulative damage assessment value is obtained by performing a weighted sum operation on the offsets of key parameters of the same type of sensor group; In the nonlinear mapping mode, the maximum key parameter offset in the same type of sensor group is extracted and then subjected to exponential function transformation to obtain the cumulative damage assessment value.

[0011] Furthermore, the sensor monitoring data is corrected according to the cumulative damage assessment value and input into the fault diagnosis model to perform early warning, including: Obtaining the cumulative damage assessment value of the sensor group and the original monitoring data of the corresponding sensor group; The cumulative damage assessment value is converted into a data correction coefficient through the material damage transfer function; Multiplying the data correction coefficient with the original monitoring data to generate the corrected monitoring data; The corrected monitoring data is input into the pre-trained fault diagnosis model to generate early warning signals.

[0012] On the other hand, the present invention provides an intelligent monitoring system for low-voltage switchgear based on the Internet of Things, comprising: The arc detection module is used to collect current and voltage waveform data. When the current mutation rate exceeds the preset mutation threshold and is accompanied by high-frequency oscillation characteristics, an arc event is determined to have occurred. The coupling weight module is used to extract the skin effect characteristics of the cabinet and the parasitic capacitance of the sensor power supply circuit based on the arc location and the sensor installation coordinates, and calculate the coupling weight coefficient of the electromagnetic pulse to each sensor; An offset calculation module is used to capture the operating parameters of the corresponding sensor when an arc event occurs based on the coupling weight coefficient, and compare the operating parameters with the historical reference parameters of the corresponding sensor to calculate the offset of the key parameters; The entropy change monitoring module is used to calculate the information entropy value of the offset of key parameters of groups of sensors of the same type. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, the nonlinear damage mark is triggered; A segmented mapping module is used to establish a segmented mapping relationship based on the key parameter offset and nonlinear damage marker, and output the cumulative damage assessment value of the sensor group; The early warning correction module is used to correct the sensor monitoring data according to the cumulative damage assessment value and input it into the fault diagnosis model to execute the early warning.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. Through the electromagnetic damage perception and data self-correction mechanism, the monitoring distortion problem caused by the implicit degradation of sensors is solved; a dynamic correlation model between arc events and sensor health status is established: the electromagnetic coupling weight is calculated based on the physical space characteristics, and the interference intensity of the electromagnetic pulse on each sensor is accurately quantified, breaking through the limitation of traditional solutions that ignore the electromagnetic propagation characteristics of the equipment structure; relative offset analysis is used to effectively identify the parameter drift caused by cumulative damage of sensors by comparing the dynamic deviation between real-time working parameters and long-term baseline values; an information entropy change detection mechanism is introduced to capture the critical point of nonlinear damage by using the sudden change in the distribution characteristics of the sensor group offset. This feature has unique sensitivity to early implicit degradation.

[0014] 2. Realize real-time assessment of the health status of the monitoring equipment itself, upgrade the sensor from a simple data acquisition unit to an intelligent node with self-diagnosis function, and eliminate the risk of misjudgment caused by failure of the monitoring equipment at the source; convert the accumulated damage into a data correction coefficient through a segmented mapping relationship, and dynamically correct the monitoring data input into the fault diagnosis model to ensure that early warning decisions are always based on real working conditions; while maintaining the existing IoT architecture unchanged, significantly improve the long-term reliability of the system in strong electromagnetic environments, especially suitable for harsh working conditions where high-frequency arcing occurs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of the low-voltage switchgear intelligent monitoring method based on the Internet of Things of the present invention; Figure 2 This is a structural diagram of the low-voltage switchgear intelligent monitoring system based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] Example 1: Figure 1 The present invention provides an intelligent monitoring method for low-voltage switchgear based on the Internet of Things, comprising: S1. Collect current and voltage waveform data. When the current mutation rate exceeds a preset mutation threshold and is accompanied by high-frequency oscillation characteristics, an arc event is determined to have occurred. S2. Based on the arc location and sensor installation coordinates, extract the cabinet skin effect characteristics and the parasitic capacitance of the sensor power supply circuit, and calculate the coupling weight coefficient of the electromagnetic pulse to each sensor; S3. Capturing the operating parameters of the corresponding sensor when the arc event occurs based on the coupling weight coefficient, and comparing the operating parameters with the historical reference parameters of the corresponding sensor to calculate the key parameter offset; S4. Calculate the information entropy value of the offset of the key parameters of the same type of sensor group. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, trigger the nonlinear damage mark. S5. Establish a segmented mapping relationship based on the key parameter offset and the nonlinear damage marker, and output the cumulative damage assessment value of the sensor group; S6. Correct the sensor monitoring data according to the cumulative damage assessment value and input it into the fault diagnosis model to execute early warning.

[0018] S1. Collect current and voltage waveform data. When the current mutation rate exceeds the preset mutation threshold and is accompanied by high-frequency oscillation characteristics, it is determined that an arc event has occurred. The specific implementation is as follows: When collecting current and voltage waveform data, the raw current waveform data is obtained through the Rogowski coil sensor installed on the switchgear busbar, and the raw voltage waveform data is collected through the capacitive voltage divider. The sampling frequency of the raw current waveform data is set to 1 million sampling points per second, and the raw current waveform data is differentially calculated to generate a current differential sequence. The current differential sequence represents the numerical value of the current change rate within a unit time window, and its dimension is amperes per microsecond. The preset mutation threshold is set according to the ratio of the rated current of the switchgear. For example, for a switchgear with a rated current of 1000 amperes, the preset mutation threshold is set to 50 amperes per microsecond. The interval in the current differential sequence that lasts for more than a specific time and continuously exceeds the preset mutation threshold is identified as the arc event judgment interval. The start and end timestamps of the arc event judgment interval are recorded by the time synchronization unit of the data acquisition system.

[0019] Within the defined arc event determination interval, the raw voltage waveform data for the corresponding time period is extracted, and the time-domain voltage waveform is converted to a frequency-domain representation using a spectrum conversion algorithm. The energy distribution of the frequency-domain data within a specific high-frequency range is analyzed. For example, within the 300 kHz to 30 MHz frequency range, when a high-frequency oscillation component with an amplitude significantly exceeding the background noise and a duration that meets the minimum requirements is present, the presence of a high-frequency oscillation component that meets the arc characteristics is confirmed. The determination criteria for the high-frequency oscillation component are based on the physical characteristics of the arc, and the background noise level is obtained by statistically analyzing the voltage waveform data continuously collected by the system under no-load conditions.

[0020] The time domain center position is calculated based on the start and end timestamps of the arc event determination interval. This calculation is performed by adding the start and end timestamp values ​​and dividing by 2. The calculated time domain center position is entered into a spatial position mapping database, which stores the correspondence between time information and spatial location. By querying the physical location information that matches the timestamp, the spatial coordinates of the arc occurrence location are output. The spatial coordinates are represented in a three-dimensional rectangular coordinate system.

[0021] The current differential sequence generation process involves performing a differential operation on adjacent sampling points of the raw current waveform data, with a fixed differential step size of one sampling interval. The preset mutation threshold is set based on the equipment's technical specifications and dynamically adjusted for different switchgear specifications. The adjusted threshold is stored in the device's memory. The criteria for determining the high-frequency oscillation component include a specific high-frequency range determined by the physical characteristics of arc discharge, a background noise level obtained through actual measurements, and a minimum duration requirement based on the power system's power frequency cycle.

[0022] The mapping of the time-domain center position to spatial coordinates relies on a discretized model of the switchgear's physical structure. This model divides the conductive components into multiple physical segments, with the coordinates of each segment's center point pre-measured and stored in a database. The spatial coordinates are determined by first matching the physical segment closest to the time-domain center position and then outputting the coordinates of that segment's center point as the spatial coordinates of the arc's location. To improve positioning accuracy, a coordinate correction mechanism can be added, such as calculating a distance correction based on the electromagnetic wave propagation speed and time difference.

[0023] The duration of the high-frequency oscillation component is verified by detecting the continuous period of time during which the amplitude of the high-frequency oscillation component exceeds a specific decibel value of the background noise within the arc event determination interval. Spatial coordinate error compensation can be achieved through physical models, for example, by considering the propagation speed of electromagnetic waves. All calculations are based on actual, measurable physical parameters, and the set values ​​for each parameter can be calibrated and determined during the equipment commissioning process. The preset mutation threshold is dynamically adjusted by calculating the specific threshold value based on the rated current value marked on the switchgear nameplate using a fixed proportional coefficient.

[0024] S2. Based on the arc location and the sensor installation coordinates, extract the cabinet skin effect characteristics and the parasitic capacitance of the sensor power supply circuit, and calculate the coupling weight coefficient of the electromagnetic pulse to each sensor. The specific implementation is as follows: After obtaining the spatial coordinates of the arc location and the installation coordinates of each sensor, the spatial distance vector is calculated in a three-dimensional rectangular coordinate system. For each sensor, the X-axis component of the arc location's spatial coordinates is subtracted from the sensor's installation coordinates to obtain the ΔX value. The Y-axis components are subtracted to obtain the ΔY value, and the Z-axis components are subtracted to obtain the ΔZ value. The modulus of the spatial distance vector is calculated using the formula: the square root of the square of the ΔX value, the square of the ΔY value, and the square of the ΔZ value. The result is expressed in meters. This modulus represents the straight-line path length of the electromagnetic wave from the arc location to the sensor.

[0025] The calculation of the skin depth distribution characteristic value of the electromagnetic wave propagation path on the cabinet surface is implemented based on electromagnetic field propagation theory. The method for determining the skin depth distribution characteristic value is as follows: First, the skin depth reference value is calculated. This reference value is equal to the reciprocal of the square root of the product of the constant term multiplied by pi, the electromagnetic wave frequency value, the magnetic permeability of the cabinet material, and the electrical conductivity of the cabinet material. The electromagnetic wave frequency value is the center frequency value of the high-frequency oscillation component detected during the arc event, and the magnetic permeability and electrical conductivity of the cabinet material are obtained from the material performance test report. The skin depth distribution characteristic value is ultimately equal to the skin depth reference value multiplied by the distance correction factor, which is adjusted according to the modulus of the spatial distance vector.

[0026] The distance correction coefficient is set as follows: when the modulus of the spatial distance vector is less than a specific distance threshold, the distance correction coefficient is 1. When the modulus of the spatial distance vector is greater than or equal to the specific distance threshold, the distance correction coefficient decreases linearly as the modulus increases. The specific distance threshold is set based on the cabinet size; for example, 1 meter is used for standard switchgear. The linear reduction ratio is: for every 1-meter increase in the modulus of the spatial distance vector, the distance correction coefficient decreases by 0.1, with a minimum of 0.7.

[0027] To measure the parasitic capacitance parameters of each sensor's power supply circuit using an impedance analyzer, perform the following: Connect the impedance analyzer's test terminals between the positive and negative terminals of the sensor's power supply circuit. Set the measurement frequency to the center frequency of the arc's high-frequency oscillation component, for example, 300 kHz. Maintain the measurement temperature between 20°C and 30°C, and ensure the power supply circuit load current does not exceed 10% of the sensor's rated current. The impedance analyzer outputs a complex admittance value. The parasitic capacitance parameter is equal to the imaginary component of the admittance value divided by the angular frequency, which is 2 times pi times the measured frequency. The measurement results are stored in farads.

[0028] The preset electromagnetic coupling equation uses a multiplication relationship structure, specifically expressed as follows: the coupling weight coefficient is equal to the inverse of the skin depth distribution characteristic value multiplied by the parasitic capacitance parameter multiplied by the proportional coefficient. The proportional coefficient is determined through an experimental calibration process. The calibration method is: in a standard electromagnetic pulse test environment, the sensor output signal amplitude is recorded. The coupling weight coefficient is equal to the output signal amplitude divided by the standard electromagnetic pulse field strength. The proportional coefficient is obtained by the statistical average of multiple sets of test data. The proportional coefficient value range is set between 0.5 and 2.0. The specific value is determined by the sensor type, such as 1.2 for current transformer sensors and 0.8 for voltage divider sensors.

[0029] The electromagnetic wave frequency value required for calculating the skin depth distribution characteristic value is derived from the center frequency of the high-frequency oscillation component determined during the arc event detection process. This frequency value is obtained through fast Fourier transform spectrum analysis. The magnetic permeability of the cabinet material adopts the vacuum magnetic permeability constant, which is 4π×10 -7 Henry per meter. The electrical conductivity of the cabinet material is determined based on the material test report provided by the switch cabinet manufacturer. For example, the conductivity of the copper cabinet is 5.8×10 7 Siemens per meter. The spatial distance vector modulus is calculated based on 3D coordinate measurement data, and the coordinate measurement accuracy reaches the millimeter level.

[0030] Quality control during parasitic capacitance parameter measurement includes: performing zero-point calibration on the impedance analyzer before measurement; performing three repeated measurements on each sensor and taking the arithmetic mean; and recording measurement results to three significant digits. The experimental calibration environment for the scale factor requires: a standard electromagnetic pulse field strength of 1 volt per meter; a sensor output signal sampling rate of 10 MHz; and statistical averaging of at least 10 data samples. The calibration process is completed during the initial installation and commissioning of the equipment.

[0031] The coupling weight coefficient for each sensor is calculated independently. Data validity is verified during the calculation process: the skin depth distribution characteristic value is checked to be greater than 0; the parasitic capacitance parameter is verified to be within the acceptable range of 1 picofarad to 100 nanofarads; and the scaling factor is confirmed to be within the preset range. If any parameter is outside the acceptable range, a remeasurement process is triggered. The final output coupling weight coefficient is dimensionless and rounded to four significant figures.

[0032] The spatial coordinate measurement method includes: using a laser rangefinder to measure the horizontal and vertical distances of the sensor installation location relative to the cabinet reference point; using an angle measuring instrument to determine the angular offset in the installation height direction; and converting the absolute coordinates of the reference point marked on the cabinet design drawing into spatial coordinates in a three-dimensional rectangular coordinate system. All coordinate data is stored in the device configuration database, and the coordinate update period is the equipment overhaul and maintenance period.

[0033] The electrical conductivity of the cabinet material can be obtained by reading it directly from the switchgear factory inspection report or by using a portable conductivity meter for on-site measurement. For on-site measurements, measurements are taken at three different locations on the cabinet surface and the average value is taken. The linear adjustment rule for the distance correction coefficient is implemented in the software algorithm. Specifically, when the spatial distance vector modulus is between 1 and 3 meters, the difference between the spatial distance vector modulus and 1 is first calculated. The product of this difference and 0.1 is then calculated. Finally, this product is subtracted from 1.0 to obtain the distance correction coefficient. When the spatial distance vector modulus is greater than 3 meters, the distance correction coefficient is fixed at 0.7.

[0034] A regular calibration mechanism has been established for the proportionality coefficient in the electromagnetic coupling equation, with a calibration cycle set every 12 months. The calibration process involves generating a 300 kHz test signal using a standard signal generator; measuring the actual electromagnetic field strength using a standard field strength probe; recording the sensor output signal and recalculating the proportionality coefficient. If the newly calculated proportionality coefficient deviates from the previously stored value by more than 5%, the stored value is automatically updated. All calculations are performed using the International System of Units, intermediate variables are rounded to six significant figures, and the final result is rounded.

[0035] When calculating the spatial distance vector modulus, the units of ΔX, ΔY, and ΔZ values ​​are meters, and the square root result is rounded to three decimal places. The constant term in the skin depth baseline calculation is the reciprocal of the square root of 2, approximately 0.707. When measuring the electrical conductivity of the cabinet material on-site, the measurement points are selected on flat areas that avoid seams and coatings. Five readings are repeated at each measurement point, and the data is averaged after removing outliers. Zero-point calibration of the impedance analyzer is performed before each measurement. During calibration, the test terminals are short-circuited and an automatic zeroing procedure is performed.

[0036] The data verification mechanism for coupling weight coefficient calculation includes range and consistency checks. The range check ensures that the calculated result is within the theoretical range of 0.01 to 5.0. The consistency check compares the current result with the historical record, triggering an alarm if the deviation exceeds 10% for three consecutive calculations. Ambient temperature monitoring during measurement is performed in real time using a temperature sensor. Measurements are automatically paused when the temperature exceeds the range of 20 to 30 degrees Celsius. Spatial coordinate mapping data includes timestamps, and coordinate data version verification is performed after each device maintenance.

[0037] The linear adjustment rule for the distance correction coefficient is implemented in the software as a conditional judgment function: the spatial distance vector modulus is input, and when the modulus is less than 1 meter, the output is 1.0; when the modulus is between 1 and 3 meters, a three-step calculation process is performed; and when the modulus is greater than 3 meters, the output is 0.7. For scale factor calibration, the standard field strength probe is placed 10 cm from the sensor under test, perpendicular to the sensor's sensing surface. The final coupling weight coefficient is normalized before being used in subsequent calculations, so that the sum of the coupling weight coefficients for the same sensor group is equal to 1.

[0038] S3. Capture the operating parameters of the corresponding sensor when the arc event occurs based on the coupling weight coefficient, compare the operating parameters with the historical reference parameters of the corresponding sensor, and calculate the key parameter offset. The specific implementation is as follows: When selecting target sensors based on coupling weight coefficients, the following steps are performed: The coupling weight coefficients calculated in step S2 for all sensors are read and compared to a preset weight threshold. This threshold is set empirically between 0.1 and 0.3, for example, 0.2. Only sensors with coupling weight coefficients exceeding this threshold are selected as target sensors; sensors below this threshold are excluded from this step. The target sensor selection results are stored in the target device list register, which is updated every 100 milliseconds.

[0039] At the precise moment an arc event is detected, the target sensor's operating parameters are captured synchronously. A high-speed data acquisition unit (DAQ) acquires real-time values ​​of the target sensor's operating parameters, including but not limited to electrical characteristics such as current RMS value, voltage peak value, and signal-to-noise ratio. The data acquisition window is aligned with the arc event triggering moment, with time synchronization error controlled within 50 microseconds. The real-time values ​​are stored in a temporary buffer as 32-bit floating-point numbers, with a millisecond timestamp and sensor ID tag.

[0040] The process for retrieving historical baseline parameters is as follows: The historical baseline parameter database is retrieved from non-volatile memory based on the sensor number. The database structure consists of three core fields: sensor number, parameter type, and baseline value. The baseline value is determined through statistical analysis: 30 consecutive days of monitoring data for the sensor, with the device in a fault-free state, is obtained. After removing the maximum and minimum values, the arithmetic mean is taken as the baseline value. The baseline value is automatically recalculated on the first day of each month, and an alarm is triggered if the new baseline value deviates by more than 5% from the original value.

[0041] The calculation process for key parameter offsets is specifically implemented as follows: the real-time value of the operating parameter is read from a temporary buffer and the corresponding parameter baseline value is obtained from a database. The mathematical relationship for calculating the relative offset is: subtract the baseline value from the real-time value to obtain the difference, then divide this difference by the baseline value, and finally multiply by 100 to convert it into a percentage value. The calculation result is a signed real number, with positive values ​​indicating a positive parameter offset and negative values ​​indicating a negative offset. The calculation formula performs data verification: when the baseline value is zero, it automatically switches to absolute difference calculation; when the calculation result exceeds the ±200% range, it is forcibly clamped to the boundary value. The calculation process is completed in the digital signal processor, and a single calculation time does not exceed 10 microseconds.

[0042] When communication with the target sensor is interrupted, the moving average of the five most recent valid real-time values ​​is automatically called. If a database search fails, the factory-set default baseline value is used. If a division-by-zero error occurs during calculations, the special error code 0xFFFF is output and system diagnostics are triggered. All key parameter offset results are stored in dedicated result registers, which contain four fields: a timestamp (accurate to milliseconds), a sensor number, a parameter type, and an offset value.

[0043] The preset weight thresholds are based on an electromagnetic interference propagation attenuation model. The lower threshold of 0.1 corresponds to the maximum coupling effect value of the sensor farthest from the arcing point. The historical benchmark parameter database uses a circular storage structure, retaining the benchmark values ​​for the last 12 months for trend analysis. The percentage conversion step in the relative offset calculation can be disabled, directly outputting the ratio as a decimal. Offset data is transmitted to the central processing unit via the industrial bus with a transmission delay of no more than 2 milliseconds.

[0044] S4. Calculate the information entropy value of the offset of the key parameters of the same type of sensor group. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, trigger the nonlinear damage mark. The specific implementation is as follows: After obtaining the key parameter offsets of the same type of sensor group, normalization processing is performed to generate a normalized offset probability distribution. The specific implementation method of this normalization processing is as follows: first, the sum of the absolute values ​​of all key parameter offsets of the sensor group is calculated, and this absolute value sum is used as the normalization denominator; then, the key parameter offset of each sensor is divided by the normalization denominator to obtain the normalized offset of each sensor; the normalized offsets of all sensors constitute a probability distribution, and this probability distribution satisfies the condition that the sum of all normalized offsets is equal to 1. The numerical range of the normalized offset is between 0 and 1, and it is a dimensionless real number. The key parameter offset is derived from the sensor state analysis result output by step S3. The offset is a signed physical quantity offset value.

[0045] When calculating the Shannon information entropy value based on the normalized offset probability distribution, the calculation process is implemented according to the basic principles of information theory. The specific calculation method of the Shannon information entropy value is as follows: for each normalized offset in the probability distribution, the natural logarithm of the offset is first calculated; then the offset value is multiplied by the corresponding natural logarithm; finally, the product of these products for all sensors is accumulated and multiplied by negative 1 to obtain the final entropy value. When the normalized offset value is 0, the product term corresponding to this offset is defined as 0. The calculated Shannon information entropy value is output as the information entropy value at the current moment. This value is a dimensionless positive real number, and the calculation result is rounded to four decimal places.

[0046] To obtain the information entropy value for the same group at the previous moment, the most recent valid record is retrieved from the device's history database. The history database stores information entropy data using timestamps as indexes. Retrieval criteria include: an exact match of the sensor group number; a timestamp that is closest to and before the current moment; and a time difference of no more than 72 hours. The entropy mutation amplitude is calculated by taking the current information entropy value and the retrieved information entropy value from the previous moment and calculating the absolute value of the difference between the two. This absolute value is the entropy mutation amplitude. The entropy mutation amplitude is also a dimensionless positive real number.

[0047] The preset entropy change threshold is set in two modes: In the initial setting mode, the threshold is set based on empirical values ​​provided by the equipment manufacturer, for example, the initial threshold for the current sensor group is set to 0.3. In the adaptive adjustment mode, the system automatically collects at least 100 consecutive records of entropy mutation amplitude data under normal historical operating conditions, calculates the arithmetic mean and standard deviation of these data, and sets the preset entropy change threshold equal to the arithmetic mean plus three times the standard deviation. When the entropy mutation amplitude exceeds the preset entropy change threshold, a nonlinear damage flag trigger signal with a Boolean value of true is generated; otherwise, a Boolean value of false is generated.

[0048] The data verification mechanism during normalization includes: verifying that all key parameter offsets have been correctly entered; checking that the normalization denominator is greater than 0.000001 (if it is less than this value, all sensors are deemed to have no significant offset); and confirming that the sum of all normalization offsets is within a reasonable error range of 0.999 to 1.001. If verification fails, the raw data re-collection process is triggered to re-obtain the output results of step S3.

[0049] The logarithmic operation in the Shannon entropy calculation is implemented using the natural logarithm function, and the calculation process uses double-precision floating-point arithmetic. The historical entropy retrieval process includes exception handling: when there is no matching record in the database, the current entropy value is stored as the first record in the database, and the mutation amplitude calculation step is skipped. When multiple valid records are retrieved, the single record with the timestamp closest to the current time is selected.

[0050] The dynamic adjustment rules for the preset entropy change threshold are as follows: after every 50 valid arc event records, the statistical characteristic values ​​of the historical data set are recalculated; the stored value is updated when the new threshold differs from the previous threshold by more than 10%; the lower threshold limit is 0.1, and the upper threshold limit is 1.0. When a nonlinear damage marker trigger signal is generated, the corresponding flag bit in the device status register is set. This flag bit remains in effect until manually reset or automatically reset after a 72-hour timeout.

[0051] Quality control for entropy mutation calculations includes checking that both input entropy values ​​are within the theoretical range of 0 to 10; confirming that the time interval is valid between 1 minute and 72 hours; and triggering a data review process when the calculated result is abnormally large. All numerical calculations use the IEEE 754 floating-point standard, and overflow protection mechanisms are implemented in key calculation steps.

[0052] The data structure for the normalized offset probability distribution is stored in a dynamic array, with the array length strictly matching the number of sensor group members. The historical database storage format consists of four fields: a millisecond-accurate timestamp, a sensor group ID, an information entropy value, and a data validity flag. The preset entropy threshold is initialized during the initial operation of the device. Initial values ​​are entered through the human-machine interface, and the adaptive adjustment function is enabled.

[0053] The specific rules for handling edge cases are as follows: when the sum of the absolute values ​​of the key parameter offsets is less than 0.000001, an equal normalized offset (i.e., 1 divided by the number of sensors) is assigned to each sensor in the group; when the normalized offset is less than 0.0000000001, the product term corresponding to the offset is forced to be zero in the Shannon entropy calculation. The natural logarithm calculation module has a built-in input value lower limit protection. When the input value is lower than 10⁻¹ 0 , which returns the logarithm of -23.025851.

[0054] The application logic for the nonlinear damage marker trigger signal includes: blocking repeated triggers within the signal's validity period; initiating a dedicated detection program for associated sensors after signal generation; and displaying corresponding alarm indicators on the system status panel. Historical entropy data is stored in a circular manner, with a maximum of 1000 records stored, automatically overwriting the oldest record if the number exceeds the limit. A hysteresis interval of 0.001 is set for the threshold comparison operation, maintaining the previous judgment result when the entropy value mutation amplitude falls within the threshold ±0.001.

[0055] The statistical calculation process for the preset entropy change threshold is as follows: extract the entropy value mutation amplitude data for the last 200 valid records from the database; remove the extreme values ​​of the largest and smallest 5%; calculate the arithmetic mean μ and standard deviation σ of the remaining data; and set the final threshold to μ + 3σ. This statistical calculation is automatically performed every 90 days. Sensor groups are identified as similar based on the model matching rules defined in the device configuration file. Sensors of the same model installed in different locations are grouped together.

[0056] The key parameter offset is obtained by reading the relative offset values ​​of each sensor from the data buffer within one second after step S3. The normalization module's execution cycle is synchronized with the arc event trigger, generating a separate probability distribution dataset for each arc event. The Shannon information entropy calculation unit initiates operations within 200 milliseconds of generating the probability distribution, ensuring real-time performance. All time parameters are derived from the device's high-precision real-time clock, with a clock synchronization error of less than one millisecond.

[0057] S5. Establish a segmented mapping relationship based on the key parameter offset and the nonlinear damage marker, and output the cumulative damage assessment value of the sensor group. The specific implementation is as follows: After obtaining the key parameter offset and the nonlinear damage marker trigger signal, a segmented mapping relationship is established to output the cumulative damage assessment value of the sensor group. In specific implementation, the mapping mode is selected based on the Boolean state of the nonlinear damage marker trigger signal: when the signal is false, the linear mapping mode is enabled; when the signal is true, the nonlinear mapping mode is enabled. Mode selection is implemented through conditional judgment logic, and the judgment result is stored in the mode status register. The key parameter offset is derived from the signed real value output in step S3, and the nonlinear damage marker trigger signal is derived from the Boolean signal output in step S4. Both are synchronously transmitted to the processing unit via the data bus.

[0058] The specific operation process of the linear mapping mode is as follows: read the key parameter offset values ​​one by one from the same type of sensor group, and at the same time obtain the coupling weight coefficient pre-calculated and stored in step S2 from the non-volatile memory. Perform weighted calculation on each sensor: multiply the key parameter offset value by the corresponding coupling weight coefficient to obtain the weighted offset of the sensor. The cumulative damage assessment value is equal to the algebraic sum of all weighted offsets. This calculation process must meet the weight coefficient normalization condition, that is, the sum of the coupling weight coefficients of all sensors in the same group is equal to 1. Perform data validity verification during calculation: check whether each coupling weight coefficient is in the closed interval of 0 to 1; confirm whether the key parameter offset is within the relative variation range of ±200% allowed by the equipment. The calculation result is a signed real value, and its dimension is consistent with the original key parameter offset.

[0059] The specific operational process of the nonlinear mapping mode is as follows: All key parameter offsets for a group of sensors of the same type are traversed, and the value with the largest absolute value is extracted as the base input value. This base input value is then subjected to an exponential transformation: first, the base input value is multiplied by a scaling factor to obtain an intermediate variable; then, the base of the natural logarithm is raised to the power of the intermediate variable; then, a constant of 1 is subtracted from the result; and finally, the difference is multiplied by a scaling factor to obtain the cumulative damage assessment value. This calculation result is a dimensionless real number with an output range limited to 0 to 1.0. The scaling factor is obtained through calibration of the device tolerance characteristics test. The calibration method includes conducting a gradient test on at least 10 sensor samples of the same model in a laboratory environment, gradually increasing the key parameter offset in 10% increments until the sensor output deviation exceeds 20% of the range, and recording the critical failure point data. The scaling factor is equal to the natural logarithm function value divided by the critical offset value. Typical values ​​for the scaling factor range from 0.5 to 2.0, for example, and 1.2 for current sensors is used in practical implementation. The scaling factor is fixed at 0.632, which corresponds to a statistical characteristic point with a failure probability of 63.2%.

[0060] The time series control mechanism for mode switching and data processing is as follows: within 5 milliseconds of the arc event trigger signal arriving, the state of the nonlinear damage marker trigger signal output in step S4 is obtained; a mapping mode is selected based on this signal state; and the cumulative damage assessment value for the selected mode is calculated within 20 milliseconds. The output values ​​of both modes are converted to dimensionless damage indices ranging from 0 to 1.0: the linear mapping mode result is divided by a preset baseline value and multiplied by a conversion factor of 0.5; the nonlinear mapping mode result is used directly as the output value. The preset baseline value is 90% of the maximum key parameter deviation in the historical records of a group of sensors of the same type, and this baseline value is automatically updated monthly.

[0061] Boundary condition handling rules include: in linear mapping mode, when a negative weight coefficient is detected, the absolute value is automatically taken and a system alarm is triggered; when the weighted sum exceeds the ±300% range, the output value is clamped to the ±3.0 boundary value. In nonlinear mapping mode, when the base input value is less than 0.01, it is forced to 0; when it is greater than 0.8, the output is fixed to 1.0. The exception handling mechanism covers the following scenarios: when input data is missing, the previous valid value is automatically recalled; when the mode status register is abnormal, the linear mapping mode is defaulted; and when the calculation timeout occurs, a simplified algorithm is used (the arithmetic mean is used in linear mode, and the maximum offset is directly output in nonlinear mode).

[0062] The sources and precision control measures for key parameters are as follows: Key parameter offsets are updated within 100 milliseconds after an arc event is triggered and stored as 32-bit floating-point numbers. Calibration data for the scaling factor is stored in the device's secure storage area, access requires password authentication. Exponential function calculations are implemented using a table lookup combined with linear interpolation, with a step size of 0.001, ensuring a maximum error of less than 0.1%. The final cumulative damage assessment output is rounded to three decimal places, stored in a dedicated register, and simultaneously uploaded to the monitoring system.

[0063] The dynamic parameter adjustment mechanism includes: The preset baseline value is automatically updated on the first day of each month. The new baseline value is set to 90% of the moving average value in the previous 30 days of data. The scaling factor is calibrated annually during equipment maintenance. New values ​​are obtained through calibration tests of spare sensors. The database is updated when the difference between the old and new values ​​exceeds 10%. The weighting factor is remeasured and re-entered after equipment hardware modifications. A detailed log is generated for all adjustment operations, including the time of operation, the values ​​before and after the modification, and the operator information.

[0064] The application logic for the output results is as follows: After the cumulative damage assessment value is output, three status flags are triggered based on the value range: 0 to 0.3 indicates a green normal state, 0.3 to 0.7 indicates a yellow warning state, and 0.7 to 1.0 indicates a red alarm state. The status flags are persistently stored in the equipment operation database and used to generate lifespan trend analysis reports. If the red alarm state is detected three times in a row, a maintenance work order is automatically generated in the maintenance system. Data transmission uses an industrial bus protocol to ensure delivery to the central monitoring platform within 100 milliseconds.

[0065] S6. Correct the sensor monitoring data based on the cumulative damage assessment value and input it into the fault diagnosis model to execute early warning. The specific implementation is as follows: After the cumulative damage assessment value is calculated, the monitoring data correction and fault warning process is executed. First, the cumulative damage assessment value and the corresponding raw monitoring data of the same type of sensor group are obtained. The cumulative damage assessment value is derived from the dimensionless real number in the range of 0 to 1.0 output in step S5. The raw monitoring data is the physical quantity data collected by the sensor in real time, such as current values ​​in amperes or voltage values ​​in volts, with a sampling rate of 10 kHz and data formatted as 32-bit floating point numbers. The two types of data are aligned and matched using millisecond timestamps, and the time synchronization error is controlled within 1 millisecond.

[0066] The method for constructing a material damage transfer function specifically involves characterizing the quantitative relationship between the degree of material damage and the distortion of monitoring data, and is obtained through calibration of accelerated material aging tests. The calibration process is as follows: In a laboratory environment, at least 30 sensor samples of the same model are subjected to a stepwise increasing mechanical stress, with a stress gradient of 10% of the device's rated value per step. Microscopic crack growth rates in the material are observed using an electron microscope. When the crack growth rate reaches 1 micron per hour, the sensor output deviation is simultaneously recorded. The least squares method is used to fit the experimental data, establishing a mapping relationship between the cumulative damage assessment value and the data correction factor. The specific conversion rule is: when the cumulative damage assessment value is between 0 and 0.3, the correction factor is output as 1.0; when the cumulative damage assessment value is between 0.3 and 0.7, the correction factor decreases linearly from 1.0 to 0.7; and when the cumulative damage assessment value is greater than 0.7, the correction factor decays exponentially to a lower limit of 0.5. The correction factor is a dimensionless real number, calculated to four decimal places.

[0067] The data correction process is as follows: The raw monitoring data array at the current moment is read and the corresponding data correction coefficient is obtained. A multiplication operation is performed on each sample point of the raw monitoring data: the sample point value is multiplied by the data correction coefficient to obtain the corrected monitoring data. This calculation is performed in real time in the digital signal processor, processing 1000 data points per millisecond. The corrected monitoring data retains its original physical dimensions; for example, current data remains in amperes, and the data storage format remains the same as the original data.

[0068] The input and output rules of the pre-trained fault diagnosis model are as follows: the model input is a modified multi-sensor monitoring data set, which contains waveform data from all sensors in the same group within a 200-millisecond time window, with a fixed total of 2,000 sampling points. The model output is a three-level warning signal: a value of 0 indicates a normal state, a value of 1 indicates a warning state, and a value of 2 indicates an alarm state. The model architecture uses a convolutional neural network consisting of three convolutional layers and two fully connected layers. It is trained on a historical fault data set containing 10,000 sets of labeled samples. The model execution cycle is synchronized with the data window, with the output updated every 200 milliseconds.

[0069] Key parameter sources and processing details include: the cumulative damage assessment value is read from the result register in step S5, with an update frequency synchronized with the arc event trigger. The parameters of the material damage transfer function are stored in the device's read-only memory and include five characteristic mapping points. For example, a cumulative damage assessment value of 0.3 corresponds to a correction factor of 1.0, a cumulative damage assessment value of 0.5 corresponds to a correction factor of 0.85, a cumulative damage assessment value of 0.7 corresponds to a correction factor of 0.7, a cumulative damage assessment value of 0.8 corresponds to a correction factor of 0.6, and a cumulative damage assessment value of 1.0 corresponds to a correction factor of 0.5. Correction calculations are performed in parallel using a hardware multiplier array, with a computational latency of less than 10 microseconds.

[0070] The boundary conditions and exception handling mechanism cover the following scenarios: when the cumulative damage assessment value exceeds the range of 0 to 1.0, it is forcibly set to the nearest boundary value (that is, if it is less than 0, it is treated as 0, and if it is greater than 1.0, it is treated as 1.0); when the transfer function output coefficient is lower than 0.5, 0.5 is fixed as the correction coefficient; when part of the original monitoring data is missing, the arithmetic mean of the valid data of the previous 5 milliseconds is used instead; when the model input data is abnormal, the current window data is discarded and the next window data is enabled.

[0071] The training and updating rules for the fault diagnosis model are as follows: The training dataset covers normal operating conditions and 12 typical fault modes, with a total of 10,000 sample sets. Model parameters are updated online every 90 days, with the most recent 100 arc events added to retrain the fully connected layer parameters. A confidence check is added to the model output layer: if the maximum probability value of the output layer falls below 0.6, the warning signal is marked as invalid.

[0072] The technical effectiveness of data correction is verified by injecting a standard test signal during equipment maintenance and comparing the total harmonic distortion (THD) of the waveforms before and after correction. When the cumulative damage assessment value exceeds 0.7, the harmonic content of the corrected signal should be reduced by at least 30%. The engineering application rules for early warning signals are as follows: an output value of 1 triggers the device status indicator to flash yellow; an output value of 2 triggers a solid red light and sends a text message notification; and three consecutive output values ​​of 2 trigger the device's automatic power-off protection program.

[0073] Real-time guarantee measures include: setting up two-level buffer registers in the data correction pipeline to ensure continuous and uninterrupted data processing; using a fixed-point optimization algorithm in the model inference process, and controlling the single execution time within 180 milliseconds; establishing a timeout fuse mechanism, and directly outputting the previous valid result when the calculation timeout exceeds 250 milliseconds.

[0074] Historical data management is implemented by storing all pre- and post-correction monitoring data in a circular buffer, complete with millisecond timestamps and cumulative damage assessment values. A complete record of the last 72 hours is retained. The early warning decision log contains five fields: timestamp, cumulative damage assessment, correction coefficient, model input data fingerprint, output result, and confidence level, which are used for post-fault source tracing.

[0075] Correction coefficients are applied only to analog sensor data, such as current, voltage, and temperature; digital signals are not corrected. For multi-range sensors, raw data is automatically converted to percentages of the range before correction calculations for unified processing. The final output warning signal is transmitted to the monitoring center via Industrial Ethernet, with an end-to-end transmission delay of no more than 50 milliseconds.

[0076] The technical solution of this embodiment achieves a significant improvement in the accuracy of equipment fault warning through the synergistic effect of multiple steps. When establishing the coupling weight system in step S1, different from the conventional average weighting method, the weight is dynamically allocated based on the spatial relationship between the physical position of the sensor and the arc occurrence point. This allocation logic needs to be implemented in combination with the equipment structure topology; step S3 uses relative offset analysis instead of absolute value comparison to effectively eliminate the influence of individual differences in equipment; step S4 introduces an entropy mutation detection mechanism to capture early signs of nonlinear material damage through probability distribution characteristics. This feature is essentially different from the traditional linear cumulative damage model; the segmented mapping relationship established in step S5 is not a simple mode switch, but is based on the study of material failure mechanism. In the nonlinear stage, exponential transformation is used to highlight the dominant effect of local damage; the material damage transfer function in step S6 is constructed through accelerated aging test evidence. The mapping relationship that converts the structural damage amount into the electrical signal correction coefficient must be strictly calibrated and verified. The technical closed loop formed by each step breaks through the limitation of the traditional monitoring system that only focuses on electrical parameters, and realizes the advancement of fault warning through the mechanical-electrical damage correlation model.

[0077] Example 2: Figure 2 The present invention provides a structural diagram of a low-voltage switchgear intelligent monitoring system based on the Internet of Things, which includes: The arc detection module is used to collect current and voltage waveform data. When the current mutation rate exceeds the preset mutation threshold and is accompanied by high-frequency oscillation characteristics, an arc event is determined to have occurred. The coupling weight module is used to extract the skin effect characteristics of the cabinet and the parasitic capacitance of the sensor power supply circuit based on the arc location and the sensor installation coordinates, and calculate the coupling weight coefficient of the electromagnetic pulse to each sensor; An offset calculation module is used to capture the operating parameters of the corresponding sensor when an arc event occurs based on the coupling weight coefficient, and compare the operating parameters with the historical reference parameters of the corresponding sensor to calculate the offset of the key parameters; The entropy change monitoring module is used to calculate the information entropy value of the offset of key parameters of groups of sensors of the same type. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, the nonlinear damage mark is triggered; A segmented mapping module is used to establish a segmented mapping relationship based on the key parameter offset and nonlinear damage marker, and output the cumulative damage assessment value of the sensor group; The early warning correction module is used to correct the sensor monitoring data according to the cumulative damage assessment value and input it into the fault diagnosis model to execute the early warning.

[0078] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to actual conditions.

[0079] It should be noted that the present invention can be deployed on the device itself to implement embedded applications, and can also be run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0080] The above embodiments can be implemented in whole or in part via software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in the embodiments of this application are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission. Wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission methods include infrared, microwave, etc. The computer-readable storage medium can be any available medium accessible by a computer, or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0081] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0082] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0083] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0084] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0085] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0086] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

[0087] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent monitoring method for low-voltage switchgear based on the Internet of Things, characterized in that: include: S1. Collect current and voltage waveform data. When the current mutation rate exceeds a preset mutation threshold and is accompanied by high-frequency oscillation characteristics, an arc event is determined to have occurred. S2. Based on the arc location and sensor installation coordinates, extract the cabinet skin effect characteristics and the parasitic capacitance of the sensor power supply circuit, and calculate the coupling weight coefficient of the electromagnetic pulse to each sensor; S3. Capturing the operating parameters of the corresponding sensor when the arc event occurs based on the coupling weight coefficient, and comparing the operating parameters with the historical reference parameters of the corresponding sensor to calculate the key parameter offset; S4. Calculate the information entropy value of the offset of the key parameters of the same type of sensor group. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, trigger the nonlinear damage mark. S5. Establish a segmented mapping relationship based on the key parameter offset and the nonlinear damage marker, and output the cumulative damage assessment value of the sensor group; S6. Correct the sensor monitoring data according to the cumulative damage assessment value and input it into the fault diagnosis model to execute early warning.

2. The method for intelligent monitoring of low-voltage switchgear based on the Internet of Things according to claim 1 is characterized in that: Collect current and voltage waveform data. When the current mutation rate exceeds the preset mutation threshold and is accompanied by high-frequency oscillation characteristics, an arc event is determined to have occurred, including: Calculating a current differential sequence based on current waveform data, and identifying intervals in the current differential sequence that continuously exceed a preset mutation threshold as arc event determination intervals; Analyze the spectral characteristics of the voltage waveform within the arc event determination interval to confirm the presence of high-frequency oscillation components that meet the arc characteristics; The spatial coordinates of the arc occurrence position are determined according to the time domain center position of the arc event judgment interval.

3. The method for intelligent monitoring of low-voltage switchgear based on the Internet of Things according to claim 2, characterized in that: Based on the arc location and sensor installation coordinates, the cabinet skin effect characteristics and the parasitic capacitance of the sensor power supply circuit are extracted, and the coupling weight coefficient of the electromagnetic pulse to each sensor is calculated, including: Based on the spatial coordinates of the arc occurrence location and the spatial distance vectors of the sensor installation coordinates, the skin depth distribution characteristic values ​​of the electromagnetic wave propagation path on the cabinet surface are calculated; Measure the parasitic capacitance parameters of each sensor power supply circuit using an impedance analyzer; The skin depth distribution characteristic value and parasitic capacitance parameters of each sensor are input into the preset electromagnetic coupling equation, and the coupling weight coefficient of each sensor is output.

4. The method for intelligent monitoring of low-voltage switchgear based on the Internet of Things according to claim 3 is characterized in that: The operating parameters of the corresponding sensor when the arc event occurs are captured based on the coupling weight coefficient, and the operating parameters are compared with the historical benchmark parameters of the corresponding sensor to calculate the key parameter offset, including: Screening sensors whose coupling weight coefficients exceed a preset weight threshold as target sensors; When an arc event occurs, the real-time values ​​of the target sensor's operating parameters are recorded synchronously; Retrieve the corresponding parameter reference value from the historical reference parameter database of the target sensor; The relative offset between the real-time value of the working parameter and the parameter reference value is calculated as the key parameter offset.

5. The method for intelligent monitoring of low-voltage switchgear based on the Internet of Things according to claim 4 is characterized in that: Calculate the information entropy value of the offset of key parameters of the same type of sensor group. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, trigger the nonlinear damage mark, including: Normalize the key parameter offset of each sensor in the same type of sensor group to generate a normalized offset probability distribution; The Shannon information entropy value is calculated based on the normalized offset probability distribution as the information entropy value at the current moment; Obtain the information entropy value of the same group at the previous moment, and calculate the absolute difference between the current information entropy value and the previous information entropy value as the entropy mutation amplitude; When the entropy value mutation amplitude exceeds the preset entropy change threshold, a nonlinear damage mark trigger signal is generated.

6. The method for intelligent monitoring of low-voltage switchgear based on the Internet of Things according to claim 5, characterized in that: A segmented mapping relationship is established based on the key parameter offset and nonlinear damage markers, and the cumulative damage assessment value of the sensor group is output, including: Selecting a linear mapping mode or a nonlinear mapping mode according to the state of a nonlinear damage marker trigger signal; In the linear mapping mode, the cumulative damage assessment value is obtained by performing a weighted sum operation on the offsets of key parameters of the same type of sensor group; In the nonlinear mapping mode, the maximum key parameter offset in the same type of sensor group is extracted and then transformed into an exponential function to obtain the cumulative damage assessment value.

7. The method for intelligent monitoring of low-voltage switchgear based on the Internet of Things according to claim 6, characterized in that: The sensor monitoring data is corrected based on the cumulative damage assessment value and input into the fault diagnosis model to perform early warning, including: Obtaining the cumulative damage assessment value of the sensor group and the original monitoring data of the corresponding sensor group; The cumulative damage assessment value is converted into a data correction coefficient through the material damage transfer function; Multiplying the data correction coefficient with the original monitoring data to generate the corrected monitoring data; The corrected monitoring data is input into the pre-trained fault diagnosis model to generate early warning signals.

8. An intelligent monitoring system for low-voltage switchgear based on the Internet of Things, used to implement the intelligent monitoring method for low-voltage switchgear based on the Internet of Things according to any one of claims 1 to 7, characterized in that: include: The arc detection module is used to collect current and voltage waveform data. When the current mutation rate exceeds the preset mutation threshold and is accompanied by high-frequency oscillation characteristics, an arc event is determined to have occurred. The coupling weight module is used to extract the skin effect characteristics of the cabinet and the parasitic capacitance of the sensor power supply circuit based on the arc location and the sensor installation coordinates, and calculate the coupling weight coefficient of the electromagnetic pulse to each sensor; An offset calculation module is used to capture the operating parameters of the corresponding sensor when an arc event occurs based on the coupling weight coefficient, and compare the operating parameters with the historical reference parameters of the corresponding sensor to calculate the offset of the key parameters; The entropy change monitoring module is used to calculate the information entropy value of the offset of key parameters of groups of sensors of the same type. When the information entropy value mutation amplitude exceeds the preset entropy change threshold, the nonlinear damage mark is triggered; A segmented mapping module is used to establish a segmented mapping relationship based on the key parameter offset and nonlinear damage marker, and output the cumulative damage assessment value of the sensor group; The early warning correction module is used to correct the sensor monitoring data according to the cumulative damage assessment value and input it into the fault diagnosis model to execute the early warning.

Images