Electrical fire hazard on-line monitoring and advanced early warning method and system

Abstract

The invention discloses an electrical fire hazard on-line monitoring and advanced early warning method and system, and relates to the technical field of electrical monitoring, and the method comprises the steps: injecting a group of excitation signals of a preset frequency spectrum into an electrical connection point when the electrical connection point bears a working current; synchronously acquiring an electric response signal and a thermal response signal generated by the excitation signal at the electrical connection point, and extracting energy and heat conversion efficiency at each frequency spectrum point to form a real-time efficiency characteristic spectrum; performing multi-scale decomposition on the real-time efficiency characteristic spectrum, and separating out a low-frequency trend component reflecting the overall oxidation state of the contact surface, an intermediate-frequency fluctuation component reflecting microcosmic cracks and chippings and a high-frequency noise component reflecting the characteristics of a surface film; according to the invention, directional tracing and comprehensive evaluation of the degradation type, the severity level and the remaining effective life of the electrical connection point are realized, and online monitoring and advanced early warning of electrical fire hazards are successfully achieved.

Description

Technical Field

[0001] This invention relates to the field of electrical monitoring technology, and in particular to a method and system for online monitoring and early warning of electrical fire hazards. Background Technology

[0002] Electrical connection points are critical nodes for energy transmission in electrical systems. They are constantly subjected to current carrying and environmental corrosion, making them prone to deterioration such as oxidation, cracking, and changes in surface characteristics. This deterioration directly leads to increased contact resistance and abnormal energy conversion, ultimately causing localized overheating or even electrical fires. Therefore, accurate diagnosis of the deterioration status of electrical connection points is the core foundation for online monitoring and early warning of electrical fire hazards. Only by understanding the type, severity, and development trend of deterioration in advance can fire risks be avoided at the source.

[0003] Existing methods for early warning of electrical fire hazards fail to focus on the core need of diagnosing the deterioration status of electrical connection points. They lack the ability to effectively capture multi-dimensional characteristic signals during the deterioration process and cannot establish stable health status reference benchmarks for individual connection points. This makes it difficult to identify early, hidden signs of deterioration, and even more difficult to accurately distinguish deterioration types, determine severity levels, and estimate remaining effective lifespan. This neglect of core diagnostic needs means that existing methods can only monitor obvious surface phenomena such as abnormal heating, failing to achieve early prediction of fire hazards and thus failing to meet the actual needs of electrical system safety control. Summary of the Invention

[0004] This invention provides a method and system for online monitoring and early warning of electrical fire hazards to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for online monitoring and early warning of electrical fire hazards, comprising: S1. When the electrical connection point carries the working current, inject a set of excitation signals with a preset spectrum into it; S2. Synchronously acquire the electrical response signal and thermal response signal generated by the excitation signal at the electrical connection point, and extract the energy and heat conversion efficiency at each spectrum point to form a real-time efficiency characteristic spectrum. S3. Perform multi-scale decomposition on the real-time efficiency characteristic spectrum to separate the low-frequency trend component reflecting the overall oxidation state of the contact surface, the mid-frequency fluctuation component reflecting micro-cracks and debris, and the high-frequency noise component reflecting the surface film characteristics. S4. Obtain the historical reference spectrum library corresponding to the electrical connection point. The historical reference spectrum library contains reference components of each scale established under healthy conditions. S5. Compare the low-frequency trend component, mid-frequency fluctuation component, and high-frequency noise component with the corresponding reference components at each scale in the historical reference spectrum library to obtain the deviation of each scale component. S6. Based on the combination of deviations of each scale component, and referring to the pre-determined correspondence between degradation type, severity level and remaining effective life, determine the specific degradation status of the electrical connection point and generate a diagnostic report.

[0006] Preferably, injecting a set of excitation signals with a preset spectrum into the electrical connection point when it carries operating current includes: A set of fundamental frequencies is selected based on the material and structural parameters of the electrical connection points, and the fundamental frequencies are sideband-extended to obtain the spectrum of the modulated excitation signal. The modulated excitation signal spectrum is injected into the working current path through a signal coupling device to form a composite current-carrying signal containing the excitation spectrum. The current component within a preset frequency range is separated from the composite current-carrying signal, and the power frequency carrier component is filtered out to obtain the excitation response signal characterizing the impedance characteristics of the electrical connection point.

[0007] Preferably, the synchronous acquisition of electrical response signals and thermal response signals generated by the excitation signal at the electrical connection point, and the extraction of their energy and heat conversion efficiency at each spectral point to form a real-time efficiency characteristic spectrum, includes: Within a preset synchronization time window, the electrical response signal is extracted from the composite current-carrying signal, and the surface thermal radiation change during the corresponding time period is captured by a non-contact sensor to obtain synchronized time-domain electrical response signal and thermal response signal. Synchronous spectral mapping of the time-domain electrical response signal and the thermal response signal yields the frequency domain distribution of electrical energy and the frequency domain distribution of thermal radiation, respectively. By correlating the amplitudes of corresponding spectral points in the frequency domain distribution of electrical energy with those in the frequency domain distribution of thermal radiation, a two-dimensional feature matrix characterizing the energy-to-heat conversion relationship is constructed, which serves as the real-time efficiency feature spectrum.

[0008] Preferably, the step of performing multi-scale decomposition on the real-time efficiency characteristic spectrum to separate the low-frequency trend component reflecting the overall oxidation state of the contact surface, the mid-frequency fluctuation component reflecting micro-cracks and debris, and the high-frequency noise component reflecting the surface film characteristics includes: Based on the frequency domain energy concentration characteristics of the real-time efficiency characteristic spectrum, multiple target analysis scales are determined, and the center frequency and bandwidth parameters corresponding to each target analysis scale are obtained. Based on the center frequency and bandwidth parameters, a group of bandpass filters with different passband ranges are configured to form a multi-scale filter bank. The real-time efficiency characteristic spectrum is input into a multi-scale filter bank, and the low-frequency trend component, mid-frequency fluctuation component and high-frequency noise component are separated in parallel through the passband filtering effect of each filter.

[0009] Preferably, the step of configuring a group of bandpass filters with different passband ranges based on center frequency and bandwidth parameters to form a multi-scale filter bank includes: The main characteristic frequencies of the energy distribution of each reference component are extracted by calling up the reference components at each scale in the historical reference spectrum library. Using the main characteristic frequency as a reference, the obtained center frequency and bandwidth parameters are calibrated and offset to obtain characteristic filter parameters that are adapted to the historical state of the current electrical connection point. Based on the characteristic filter parameters, the passband range of the corresponding filter in the multi-scale filter bank is reconstructed so that the separation target of the filter bank is aligned with the spectral features of the historical health state.

[0010] Preferably, the acquisition of the historical reference spectral library corresponding to the electrical connection point, the historical reference spectral library containing reference components at various scales established under healthy conditions, includes: Under the health verification state of the electrical connection point, the excitation signal of the preset spectrum is repeatedly injected, and the electrical response and thermal response are collected simultaneously each time to obtain multiple sets of benchmark datasets. Multiple sets of benchmark datasets are processed to obtain the real-time efficiency feature spectrum for each time, and all real-time efficiency feature spectra are averaged to obtain a stable benchmark efficiency feature spectrum. Multi-scale decomposition of the baseline efficiency characteristic spectrum separates the corresponding low-frequency baseline trend component, mid-frequency baseline fluctuation component and high-frequency baseline noise component. The low-frequency reference trend component, the mid-frequency reference fluctuation component, and the high-frequency reference noise component are normalized and then linked and stored in a dedicated database to form a historical reference spectrum library.

[0011] Preferably, the step of comparing the low-frequency trend component, mid-frequency fluctuation component, and high-frequency noise component with the corresponding reference components at each scale in the historical reference spectrum library to obtain the deviation of each scale component includes: Retrieve the low-frequency reference trend component, mid-frequency reference fluctuation component, and high-frequency reference noise component corresponding to the current electrical connection point from the historical reference spectrum library; Based on the energy-heat conversion relationship characteristics revealed by the two-dimensional feature matrix, each component is aligned separately; Based on the alignment results, the overall offset of the low-frequency trend component, the energy variability of the mid-frequency fluctuation component, and the distribution distortion of the high-frequency noise component are quantified respectively. Based on the overall offset, energy variability, and distribution distortion, and combined with the preset weighting relationship, a deviation degree is synthesized to characterize the overall deviation at each scale.

[0012] Preferably, the step of determining the specific degradation state of the electrical connection point and generating a diagnostic report based on the deviation combination of each scale component, referring to the predetermined correspondence between degradation type, severity level, and remaining effective life, includes: A three-dimensional state vector is constructed based on the deviation of the low-frequency trend component, the deviation of the mid-frequency fluctuation component, and the deviation of the high-frequency noise component. The three-dimensional state vector is matched with a preset degradation mode feature space, wherein the construction of the degradation mode feature space is based on the physical mechanism relationship between the components separated by multi-scale decomposition and the contact surface oxidation, micro-cracks, and surface film properties. The dominant degradation type and accompanying secondary degradation features are mapped based on the matching results. Based on the dominant degradation type and the accompanying secondary degradation characteristics, the corresponding severity level label and remaining effective lifetime range are indexed in the preset correspondence table. The severity level label, the remaining effective lifespan range, the current monitoring timestamp, and the location identifier of the electrical connection point are combined and packaged into a structured diagnostic report.

[0013] Preferably, the step of matching the three-dimensional state vector with a preset degradation mode feature space to map the dominant degradation type and accompanying secondary degradation features includes: Based on the historical failure case library, a degradation mode feature space is constructed with oxidation, cracking, and fouling as basis vectors, and an influence weight coefficient is associated with each basis vector; Determine the projection components of the three-dimensional state vector in the direction of each basis vector in the feature space of the degradation mode, and perform weighted correction on the projection components according to the influence weight coefficient; The degradation mode corresponding to the basis vector with the largest weighted projection component is selected as the dominant degradation type. At the same time, the degradation modes corresponding to the basis vectors whose projection components exceed the preset threshold are identified as the accompanying secondary degradation features.

[0014] To address the aforementioned problems, this invention also provides an online monitoring and early warning system for electrical fire hazards, the system comprising: The current excitation module is used to inject a set of excitation signals with a preset spectrum into the electrical connection point when it carries the working current. The feature extraction module is used to synchronously acquire the electrical response signal and thermal response signal generated by the excitation signal at the electrical connection point, and extract the energy and heat conversion efficiency at each spectral point to form a real-time efficiency feature spectrum. The feature decomposition module is used to perform multi-scale decomposition on the real-time efficiency feature spectrum, separating the low-frequency trend component reflecting the overall oxidation state of the contact surface, the mid-frequency fluctuation component reflecting micro-cracks and debris, and the high-frequency noise component reflecting the surface film characteristics. The reference acquisition module is used to acquire the historical reference spectrum library corresponding to the electrical connection point. The historical reference spectrum library contains reference components of each scale established under healthy conditions. The component comparison module is used to compare the low-frequency trend component, mid-frequency fluctuation component, and high-frequency noise component with the corresponding reference components at each scale in the historical reference spectrum library to obtain the deviation of each scale component. The report generation module is used to determine the specific degradation status of electrical connection points and generate a diagnostic report based on the combination of deviations of each scale component and with reference to the pre-determined correspondence between degradation type, severity level and remaining effective life.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention injects an excitation signal with a preset spectrum into electrical connection points, simultaneously collects electrical and thermal response signals and constructs a real-time efficiency feature spectrum. Through multi-scale decomposition, it separates the feature components corresponding to different degradations, and combines them with a dedicated historical benchmark spectrum library to compare and quantify the deviation. This accurately achieves targeted tracing and comprehensive evaluation of the degradation type, severity level and remaining effective life of electrical connection points, successfully realizing online monitoring and early warning of electrical fire hazards.

[0016] 2. The entire monitoring process does not require interruption of the normal operation of the electrical system. Data collection and analysis can be completed under the condition that the connection point carries the working current, adapting to the actual operating needs of various electrical equipment. The structured diagnostic report integrates key assessment results with location and time information, providing clear and implementable basis for operation and maintenance decisions, greatly improving the efficiency of early warning response and the targeting of operation and maintenance, while avoiding the waste of ineffective operation and maintenance costs, reducing the risk of electrical fires from the source, and ensuring the long-term safe and stable operation of the electrical system. Attached Figure Description

[0017] Figure 1 A flowchart illustrating an embodiment of the method for online monitoring and early warning of electrical fire hazards provided by the present invention; Figure 2 This is a functional module diagram of an online monitoring and early warning system for electrical fire hazards provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a method for online monitoring and early warning of electrical fire hazards. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for online monitoring and early warning of electrical fire hazards can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0020] Example 1, referring to Figure 1 The diagram shown is a flowchart illustrating an online monitoring and early warning method for electrical fire hazards provided in an embodiment of the present invention. In this embodiment, the online monitoring and early warning method for electrical fire hazards includes: S1. When the electrical connection point carries the operating current, inject a set of excitation signals with a preset spectrum into it, including: A set of fundamental frequencies is selected based on the material and structural parameters of the electrical connection points, and the fundamental frequencies are sideband-extended to obtain the spectrum of the modulated excitation signal. The modulated excitation signal spectrum is injected into the working current path through a signal coupling device to form a composite current-carrying signal containing the excitation spectrum. The current component within a preset frequency range is separated from the composite current-carrying signal, and the power frequency carrier component is filtered out to obtain the excitation response signal characterizing the impedance characteristics of the electrical connection point.

[0021] Specifically, the material is determined by eddy current flaw detector testing, the contact area in the structural parameters is calculated by scanning the surface profile of the connection point with a laser rangefinder, and the clamping force is obtained by measuring the pre-embedded strain gauges; the fundamental frequency is determined by using an impedance analyzer to perform an impedance spectrum scan of standard electrical connection points with the same material and structural parameters from 1kHz to 1MHz, and selecting the three frequency points with the largest changes in impedance response amplitude, for example, the fundamental frequency is determined to be 10kHz, 50kHz, and 200kHz after scanning.

[0022] Specifically, the sideband extension adopts a double-sideband modulation method. The sideband frequency offset is determined by calculating 10% of the fundamental frequency. This ratio is obtained by statistically analyzing the impedance change bandwidth of electrical connection points under different operating conditions. For example, the sideband frequencies corresponding to 10kHz are 9kHz and 11kHz, the sideband frequencies corresponding to 50kHz are 45kHz and 55kHz, and the sideband frequencies corresponding to 200kHz are 180kHz and 220kHz. Finally, a modulation excitation signal spectrum containing the fundamental frequency and the sideband frequencies is formed.

[0023] In this embodiment, sensitive frequencies are selectively screened, and sideband expansion broadens the spectral coverage, avoiding the problem that a single frequency signal cannot capture subtle changes in impedance, and significantly improving the sensitivity of subsequent response signals to changes in connection point states.

[0024] Furthermore, the modulated excitation signal is injected into the working current path through a signal coupling device to achieve common path transmission of the excitation signal and the working current, forming a composite current-carrying signal that simultaneously carries the working current and the excitation signal. The coupling device must meet the requirements of not affecting the normal transmission of the working current and being able to efficiently inject the excitation signal.

[0025] Specifically, the signal coupling device is a through-hole current coupler, and its rated current capacity is determined by measuring the maximum operating current of the electrical connection point and selecting 1.2 times the maximum operating current. For example, if the maximum operating current is 200A, then a coupler with a rated current of 240A is selected.

[0026] Specifically, the coupling coefficient of the coupler is determined by calibrating a standard resistive load, injecting an excitation signal of known amplitude into the coupler, measuring the amplitude of the induced signal at the load end, and calculating the ratio between the two. For example, the coupling coefficient after calibration is 0.98.

[0027] Specifically, the injection amplitude of the excitation signal is determined by calculating 5% of the peak value of the operating current. This proportion is obtained by statistically analyzing the influence of the excitation signal amplitude on the transmission of the operating current, ensuring that the injected signal neither interferes with the operating current nor fails to produce a detectable response. For example, if the peak value of the operating current is 300A, the injection amplitude of the excitation signal is 15A. The modulated excitation signal is generated by the signal generator and connected in series with the operating current path via a coupler to form a composite current-carrying signal.

[0028] Furthermore, a bandpass filter is used to screen the current components in the composite current-carrying signal within a preset frequency range, and a notch filter is used to filter out the power frequency carrier component. Based on the amplitude and phase relationship between the excitation signal and the response signal, an excitation-response signal that can characterize the impedance characteristics of the electrical connection point is extracted.

[0029] Specifically, the preset frequency range is determined by the fundamental and sideband frequency range of the excitation signal. The passband of the bandpass filter is set to cover all fundamental and sideband frequencies. The passband range is determined by calculating 90% to 110% of the lowest frequency of the excitation signal. For example, if the frequency range of the excitation signal is 9kHz-220kHz, then the passband of the bandpass filter is set to 8.1kHz-242kHz.

[0030] Specifically, the power frequency carrier component is filtered out using a digital notch filter. The notch frequency is determined by measuring the power frequency of the operating current in real time using a power system frequency measuring device. For example, if the measured power frequency is 50Hz, the notch frequency is set to 50Hz. The notch bandwidth is determined by calculating 2% of the power frequency, for example, 1Hz.

[0031] The composite current-carrying signal is acquired by a data acquisition card, and the sampled data is sent to a processing system, such as a LabVIEW environment, to separate the current component in a preset frequency range and filter out the power frequency component to obtain the excitation response signal.

[0032] In summary, the passband setting of a bandpass filter can fully preserve the components of the excitation response signal, while a digital notch filter can accurately filter out power frequency interference, preventing power frequency components from masking the response signal related to impedance characteristics, and ensuring that the extracted excitation response signal can accurately characterize the impedance state of the electrical connection point.

[0033] S2. Synchronously acquire the electrical response and thermal response signals generated by the excitation signal at the electrical connection points, and extract their energy and heat conversion efficiency at each spectral point to form a real-time efficiency characteristic spectrum, including: Within a preset synchronization time window, the electrical response signal is extracted from the composite current-carrying signal, and the surface thermal radiation change during the corresponding time period is captured by a non-contact sensor to obtain synchronized time-domain electrical response signal and thermal response signal. Synchronous spectral mapping of the time-domain electrical response signal and the thermal response signal yields the frequency domain distribution of electrical energy and the frequency domain distribution of thermal radiation, respectively. By correlating the amplitudes of corresponding spectral points in the frequency domain distribution of electrical energy with those in the frequency domain distribution of thermal radiation, a two-dimensional feature matrix characterizing the energy-to-heat conversion relationship is constructed, which serves as the real-time efficiency feature spectrum. The formula for constructing the real-time efficiency feature spectrum is as follows:

[0034] In the formula, Represents the two-dimensional characteristic matrix at the th The frequency band, the first The values ​​in each feature dimension, the matrix as a whole, constitute the real-time efficiency feature spectrum.

[0035] Represents the frequency domain distribution of electrical energy; Indicates at a specific characteristic frequency The amplitude of the frequency domain distribution of thermal radiation at that location.

[0036] Indicates the first The defined frequency band ranges are pre-defined based on the frequency points and subsequent multi-scale analysis requirements.

[0037] Indicates the first The integral of electrical energy within a frequency band reflects the mapping relationship of energy to heat conversion efficiency.

[0038] Specifically, the synchronization time window is determined by taking twice the longest period of the excitation signal. For example, if the longest period of the excitation signal is 10ms, then the window is set to 20ms.

[0039] The electrical response signal is extracted from the composite current-carrying signal by a data acquisition card. The sampling rate of the acquisition card is determined by 5 times the highest frequency of the excitation signal. For example, if the highest frequency is 220kHz, the sampling rate is set to 1.1MS / s.

[0040] The thermal response signal is acquired in a non-contact manner by an infrared thermal imager. The non-contact sensor is an infrared thermal imager, and its temperature measurement range is determined by measuring the normal operating temperature range of the electrical connection point, ranging from 0.8 times the lowest temperature to 1.2 times the highest temperature. For example, if the normal operating temperature is 50-80℃, then the temperature measurement range of the thermal imager is set to 40-96℃.

[0041] The frame rate of the thermal imager is determined by multiplying the reciprocal of the synchronization time window by 10. For example, for a 20ms window, the frame rate is set to 500fps.

[0042] Acquisition synchronization is achieved through the external trigger synchronization interface between the acquisition card and the thermal imager, capturing the time-domain electrical response signal and thermal response signal for the corresponding time period. The synchronization time window ensures that the complete excitation signal cycle is included, and the high sampling rate and frame rate preserve signal details. Synchronous acquisition guarantees the time correspondence between the two types of signals, providing a foundation for subsequent correlation analysis.

[0043] Furthermore, the implementation method of synchronous spectrum mapping of time-domain signals is to perform the same spectrum mapping process on the synchronously acquired time-domain electrical response signal and thermal response signal to obtain the frequency domain distribution of electrical energy and the frequency domain distribution of thermal radiation, respectively.

[0044] Specifically, the spectrum mapping uses Fast Fourier Transform (FFT). The number of transform points is determined by multiplying the synchronization time window and the sampling rate and taking the nearest power of 2. For example, if the synchronization time window is 20ms and the sampling rate is 1.1MS / s, the product is 22000, and 32768 points are taken. The FFT is performed simultaneously on the time-domain electrical response signal and the thermal response signal using LabVIEW software to determine the square of the amplitude at each frequency point, thereby obtaining the frequency domain distribution of electrical energy and the frequency domain distribution of thermal radiation.

[0045] Furthermore, by correlating the amplitude of the corresponding spectral points of the frequency domain distribution of electrical energy and thermal radiation, a two-dimensional feature matrix characterizing the relationship between energy and heat conversion is constructed using a given formula.

[0046] Specifically, the number of frequency bands is determined by counting the number of spectral points of the excitation signal and taking 10% of that number. For example, if the number of spectral points is 32768, then it is divided into 3277 frequency bands; Frequency band range The frequency range of the excitation signal is determined by dividing it into a predetermined number of frequency bands. For example, if the frequency range of 8.1kHz-242kHz is divided into 3277 frequency bands, then each frequency band has a width of approximately 71Hz. (Feature dimension) The number is determined by selecting the five frequency points with the largest amplitude in the frequency domain distribution of thermal radiation. For example, if the frequency points with the largest amplitude are 10kHz, 50kHz, 200kHz, 11kHz, and 55kHz, then... Set to 5; calculate matrix elements based on the formula constructed from the real-time efficiency feature spectrum, where By calculating the first All frequency points within a frequency band The sum of the values ​​multiplied by the bandwidth determines the frequency. For example, the first bandwidth is 71Hz, and the frequency within the bandwidth... The sum of the values ​​is 0.01 W, so the integral result is 0.71 W·Hz. Take the amplitude of the thermal radiation frequency domain distribution at the j-th characteristic frequency, for example... A value of 1 corresponds to an amplitude of 0.005W at 10kHz. Substituting this into the formula yields... The values, when combined with all elements, form a two-dimensional feature matrix, i.e., the real-time efficiency feature spectrum.

[0047] In summary, the formula quantifies the conversion efficiency by the ratio of the frequency band electrical energy integral to the characteristic frequency thermal radiation amplitude. The two-dimensional matrix comprehensively presents the conversion relationship under different frequency bands and characteristic dimensions, enabling the real-time efficiency characteristic spectrum to accurately characterize the energy conversion state of the electrical connection point.

[0048] In summary, by pre-setting a synchronization time window and externally triggering synchronization, the precise time correspondence between the electrical response and thermal response signals is ensured, avoiding analytical errors caused by time deviations and laying a reliable data foundation for subsequent correlation analysis. The time-domain signal is converted into a frequency-domain distribution through Fast Fourier Transform, accurately extracting the energy and thermal radiation characteristics of each spectral point. Compared with time-domain analysis, this method is more likely to capture subtle changes and enhances the signal's sensitivity to anomalies. By constructing a formula using the feature matrix, the frequency-domain amplitudes of electrical energy and thermal radiation are correlated, quantifying the energy-to-heat conversion efficiency and comprehensively presenting the conversion patterns under different frequency bands and feature dimensions, accurately characterizing the energy conversion state of the connection point.

[0049] Overall, the real-time efficiency feature spectrum generated in this embodiment provides a high-quality data source for multi-scale decomposition, ensuring effective separation of components at each scale and providing crucial information for assessing degradation type, severity level, and remaining lifetime. S3. Perform multi-scale decomposition on the real-time efficiency characteristic spectrum to separate the low-frequency trend component reflecting the overall oxidation state of the contact surface, the mid-frequency fluctuation component reflecting micro-cracks and debris, and the high-frequency noise component reflecting the surface film characteristics, including: Based on the frequency domain energy concentration characteristics of the real-time efficiency characteristic spectrum, multiple target analysis scales are determined, and the center frequency and bandwidth parameters corresponding to each target analysis scale are obtained. Based on the center frequency and bandwidth parameters, a group of bandpass filters with different passband ranges are configured to form a multi-scale filter bank. The real-time efficiency characteristic spectrum is input into a multi-scale filter bank. Through the passband filtering effect of each filter, the low-frequency trend component, mid-frequency fluctuation component and high-frequency noise component are separated in parallel. The multi-scale decomposition formula is as follows:

[0050] In the formula, Indicates the separated first Each scale component The frequency domain form of the real-time efficiency characteristic spectrum.

[0051] Indicates the corresponding number The frequency response function of a bandpass filter at each scale is determined by the configuration of the center frequency and bandwidth parameters.

[0052] This indicates a scale index, corresponding to multiple target analysis scales.

[0053] The integral operation embodies the core filtering and separation action of inputting the real-time efficiency characteristic spectrum into a multi-scale filter bank and using the passband filtering function of each filter.

[0054] In this embodiment of the invention, based on the center frequency and bandwidth parameters, a group of bandpass filters with different passband ranges are configured to form a multi-scale filter bank, including: The main characteristic frequencies of the energy distribution of each reference component are extracted by calling up the reference components at each scale in the historical reference spectrum library. Using the main characteristic frequency as a reference, the obtained center frequency and bandwidth parameters are calibrated and offset to obtain characteristic filter parameters that are adapted to the historical state of the current electrical connection point. Based on the characteristic filter parameters, the passband range of the corresponding filter in the multi-scale filter bank is reconstructed so that the separation target of the filter bank is aligned with the spectral features of the historical health state.

[0055] The adaptive calibration formula for the filter parameters is as follows:

[0056] In the formula, Represents the original center frequency. This represents the original bandwidth parameter. This indicates the first spectral unit extracted from the historical benchmark spectral library. The principal characteristic frequencies of each scale reference component This represents the calibration coefficient, a preset adjustment factor that controls the calibration range.

[0057] This represents the characteristic filter parameters obtained after calibration offset.

[0058] Specifically, the number of target analysis scales is determined by statistically analyzing the number of frequency point clusters in the real-time efficiency characteristic spectrum where the frequency domain energy accounts for more than 1% of the total energy. For example, if three clusters are obtained, then three target analysis scales are set, corresponding to the low-frequency trend component, the mid-frequency fluctuation component, and the high-frequency noise component, respectively.

[0059] Specifically, the center frequency of each scale is determined by the energy centroid frequency of the corresponding frequency point cluster. The energy centroid frequency is the sum of the products of the frequency value and the corresponding energy value of each frequency point in the cluster divided by the total energy of the cluster. For example, after calculating the frequency points and energy in the low-frequency cluster, the center frequency is 10kHz, the mid-frequency cluster is 50kHz, and the high-frequency cluster is 200kHz.

[0060] Specifically, the bandwidth parameter for each scale is determined by the energy half-width of the cluster at the corresponding frequency point. The energy half-width is the frequency range difference when the energy in the cluster reaches 50% of the peak value. For example, the half-width of the low-frequency cluster is 5kHz, the mid-frequency cluster is 10kHz, and the high-frequency cluster is 20kHz.

[0061] Furthermore, a bandpass filter is configured based on the center frequency and bandwidth parameters, and the filter bank is reconstructed after calling the historical reference spectrum library to calibrate the parameters.

[0062] Specifically, the historical reference spectrum library stores multiple sets of real-time efficiency characteristic spectra of the same type of electrical connection point under healthy conditions and corresponding scale reference components. The main characteristic frequency of each scale reference component is determined by extracting the frequency point with the highest energy in the frequency domain distribution of the reference component. For example, the main characteristic frequency of the low-frequency reference component is 9.8kHz, the mid-frequency is 49.5kHz, and the high-frequency is 198kHz.

[0063] Calibration coefficient The deviation distribution between the principal characteristic frequencies at each scale and the original center frequency in the historical reference spectral library is statistically analyzed, and the reciprocal of the standard deviation of this distribution is used to determine the center frequency. For example, if the standard deviation is 0.2, then... Set it to 5.

[0064] Specifically, the parameter calculation is performed according to the adaptive calibration formula of the filter parameters. First, the deviation between the historical main characteristic frequency and the original center frequency is calculated. Then, the deviation value is multiplied by the calibration coefficient to obtain the parameter adjustment amount. Finally, the original center frequency and bandwidth parameters are added to the adjustment amount to obtain the characteristic filter parameters.

[0065] For example, the original low-frequency center frequency is 10kHz, the bandwidth is 5kHz, the deviation is 9.8kHz-10kHz=-0.2kHz, the adjustment is 5×(-0.2kHz)=-0.1kHz, the calibrated center frequency is 10kHz+(-0.1kHz)=9.9kHz, and the bandwidth is 5kHz+(-0.1kHz)=4.9kHz; the calibrated intermediate frequency center frequency is 49.75kHz, the bandwidth is 4.75kHz; the calibrated high-frequency center frequency is 199kHz, the bandwidth is 19kHz.

[0066] Specifically, three bandpass filters are configured according to the characteristic parameters, with passband ranges of ± bandwidth / 2 after calibration, forming a reconstructed multi-scale filter bank.

[0067] Furthermore, the real-time efficiency feature spectrum is input into the reconstructed multi-scale filter bank, and the three target components are separated in parallel through the operation logic of the multi-scale decomposition formula.

[0068] Specifically, the reconstructed multi-scale filter bank is loaded using a digital signal processing platform, and the real-time efficiency characteristic spectrum is first converted into frequency domain form. Then, compare the frequency response functions of filters at each scale. Perform the product operation to obtain the frequency domain filtering results for each scale.

[0069] Specifically, an integral operation is performed on the frequency domain filtering results at each scale. The integration range covers the entire frequency interval. The integration is achieved using the trapezoidal numerical integration method. The integration step size is set as the frequency domain resolution. The frequency domain resolution is determined by the ratio of the sampling rate to the number of Fourier transform points. For example, if the sampling rate is 1.1 MS / s and the number of transform points is 32768, then the integration step size is 33.55 Hz.

[0070] Specifically, during the integration process, the frequency domain filtering result is compared with the complex exponential function. After multiplication, the components are accumulated point by point to complete the conversion from the frequency domain to the time domain, and the time domain components at each scale are obtained.

[0071] Finally, through parallel processing by three filters, low-frequency trend components, mid-frequency fluctuation components and high-frequency noise components are output respectively, which reflect the overall oxidation state of the contact surface, micro-cracks and debris, and surface film characteristics.

[0072] In summary, this solution precisely separates three components—contact surface oxidation, micro-cracks and debris, and surface film characteristics—through multi-scale decomposition, enabling accurate differentiation of different types of degradation features and targeted tracing of potential causes. Based on the frequency domain energy concentration characteristics, the analysis scale is determined and a bandpass filter bank is configured. Combined with calibration parameters from a historical benchmark spectral library, the separation target is aligned with the healthy state spectral characteristics, significantly improving the accuracy of component separation. Parallel separation is employed to improve processing efficiency and avoid the problem of missing key degradation information in single-scale analysis. The separated components at each scale provide high-quality data support for subsequent comparisons with historical benchmark components and quantification of deviations, ensuring the accuracy of subsequent degradation type and severity level assessments and providing core technical support for early warning.

[0073] S4. Obtain the historical reference spectrum library corresponding to the electrical connection point. The historical reference spectrum library contains reference components at various scales established under healthy conditions, including: Under the health verification state of the electrical connection point, the excitation signal of the preset spectrum is repeatedly injected, and the electrical response and thermal response are collected simultaneously each time to obtain multiple sets of benchmark datasets. Multiple sets of benchmark datasets are processed to obtain the real-time efficiency feature spectrum for each time, and all real-time efficiency feature spectra are averaged to obtain a stable benchmark efficiency feature spectrum. Multi-scale decomposition of the baseline efficiency characteristic spectrum separates the corresponding low-frequency baseline trend component, mid-frequency baseline fluctuation component and high-frequency baseline noise component. The low-frequency reference trend component, the mid-frequency reference fluctuation component, and the high-frequency reference noise component are normalized and then linked and stored in a dedicated database to form a historical reference spectrum library.

[0074] Specifically, when the electrical connection point is in a health verification state, the operation of injecting excitation signals and synchronously acquiring response signals is repeatedly performed to obtain multiple sets of benchmark datasets.

[0075] Specifically, the health verification status is determined by measuring the contact resistance of electrical connection points. The contact resistance is measured using a micro-ohmmeter. When the measured value is less than or equal to the factory standard contact resistance value for that type of connection point, it is determined to be in the health verification status. For example, if the factory standard contact resistance of a certain type of connection point is 50 micro-ohms, then when the measured value is ≤50 micro-ohms, it is in the health verification status.

[0076] Specifically, the injection parameters of the excitation signal are the same as in step S1, including the fundamental frequency of the preset spectrum, the sideband extension parameters, and the injection amplitude. The response signal acquisition uses the same equipment and parameter configuration as in step S2, with a data acquisition card sampling rate of 1.1 MS / s, an infrared thermal imager temperature measurement range of 40-96℃, and a frame rate of 500 fps.

[0077] The number of repeated samplings is determined by statistically analyzing the coefficient of variation (COP) of historical data from similar connection points. Sampling is stopped when the COP is less than 0.5%. For example, if statistics show that 20 samplings are needed to reduce the COP to less than 0.5%, then 20 repeated samplings are performed to obtain 20 baseline datasets. Multiple repeated samplings reduce the impact of random interference on the data and ensure the reliability of the baseline dataset.

[0078] Furthermore, following the real-time efficiency feature spectrum processing procedure in step S2, each set of benchmark datasets is processed one by one to obtain 20 real-time efficiency feature spectra.

[0079] Specifically, an averaging process is performed on all real-time efficiency feature spectra. The averaging method involves summing the matrix elements at corresponding positions in each real-time efficiency feature spectrum, and then dividing the sum by the number of real-time efficiency feature spectra to obtain a stable baseline efficiency feature spectrum. For example, if the sum of the matrix elements at the same position in 20 real-time efficiency feature spectra is 10, then the baseline matrix element value at that position is 10 ÷ 20 = 0.5.

[0080] The averaging process here can offset the random errors of a single acquisition, enabling the baseline efficiency characteristic spectrum to stably characterize the energy conversion properties under healthy conditions.

[0081] Furthermore, the same multi-scale decomposition method as step S3 is used to decompose the baseline efficiency characteristic spectrum and separate the corresponding three types of baseline components.

[0082] Specifically, based on the frequency domain energy concentration characteristics of the baseline efficiency characteristic spectrum, the target analysis scale and corresponding center frequency and bandwidth parameters are first determined, using the same method as in S3. Then, an initial multi-scale filter bank that has not undergone historical baseline calibration is invoked, and the baseline efficiency characteristic spectrum is input into this filter bank. The multi-scale decomposition logic is then used to complete the decomposition, yielding the low-frequency baseline trend component, the mid-frequency baseline fluctuation component, and the high-frequency baseline noise component. The integration step size remains 33.55 Hz, and the parallel processing method is the same as in S3. This decomposition method ensures that the decomposition dimensions of the baseline components and the components to be evaluated are consistent, guaranteeing their comparability.

[0083] Furthermore, the implementation method for constructing the historical reference spectrum library is as follows: normalization processing is performed on the three types of reference components that are separated, and then the processed reference components are associated and stored in a dedicated database to form the historical reference spectrum library.

[0084] Specifically, the normalization parameter is determined by statistically analyzing the maximum and minimum amplitudes of each reference component. For example, if the maximum amplitude of the low-frequency reference trend component is 2.5 and the minimum amplitude is 0.5, then normalization is achieved by dividing the value obtained by subtracting the minimum amplitude from the current amplitude by the value obtained by subtracting the minimum amplitude from the maximum amplitude.

[0085] Specifically, the dedicated database adopts a relational database, storing content including normalized low-frequency reference trend components, mid-frequency reference fluctuation components, high-frequency reference noise components, and corresponding connection point models, materials, structural parameters, and other related information.

[0086] In summary, this solution collects data multiple times and averages it under the healthy condition of electrical connection points, effectively reducing random interference and obtaining a stable baseline efficiency characteristic spectrum, ensuring the reliability of the baseline data. It separates three types of baseline components using the same multi-scale decomposition method, ensuring consistency between the baseline and the component to be evaluated in dimensions, guaranteeing comparability for subsequent comparisons. After normalization and associated storage, a dedicated historical baseline spectrum library is constructed, which can accurately retrieve baseline components matching the current connection point's model, material, and structural parameters. This provides accurate reference for quantifying deviations at each scale, avoiding evaluation errors caused by baseline mismatch, laying a solid foundation for identifying degradation states, determining severity levels, and estimating remaining effective lifespan, thereby improving the accuracy and reliability of overall monitoring and early warning.

[0087] S5. Compare the low-frequency trend component, mid-frequency fluctuation component, and high-frequency noise component with the corresponding reference components at each scale in the historical reference spectrum library to obtain the deviation of each scale component, including: Retrieve the low-frequency reference trend component, mid-frequency reference fluctuation component, and high-frequency reference noise component corresponding to the current electrical connection point from the historical reference spectrum library; Based on the energy-heat conversion relationship characteristics revealed by the two-dimensional feature matrix, each component is aligned separately; Based on the alignment results, the overall offset of the low-frequency trend component, the energy variability of the mid-frequency fluctuation component, and the distribution distortion of the high-frequency noise component are quantified respectively. Based on the overall offset, energy variability, and distribution distortion, and combined with the preset weighting relationship, a deviation degree is synthesized to characterize the overall deviation at each scale.

[0088] Specifically, three types of reference components matching the current electrical connection point are extracted from a historical reference spectrum library. The matching is based on the model, material, and structural parameters of the current electrical connection point. The model is determined by reading the equipment nameplate, the material by eddy current testing, and the structural parameters by scanning and calculating with a laser rangefinder. Using the query function of a dedicated database, the corresponding low-frequency reference trend component, mid-frequency reference fluctuation component, and high-frequency reference noise component are precisely retrieved by inputting the above matching parameters. This method ensures that the retrieved reference components completely match the basic attributes of the current connection point, providing a prerequisite for the accuracy of subsequent comparisons.

[0089] Furthermore, based on the energy-heat conversion relationship characteristics revealed by the two-dimensional feature matrix, alignment processing is performed on the current scale components and their corresponding reference components.

[0090] Specifically, the alignment process using cross-correlation is as follows: First, the peak positions of the energy-heat conversion characteristics of the current component and the reference component are extracted. Then, the offset between the two peak positions is calculated. Finally, the current component is shifted along the frequency axis by this offset so that the characteristic peak positions of the two components coincide. For example, if the characteristic peak position of the current low-frequency trend component is 9.9kHz, and the corresponding characteristic peak position of the reference component is 9.8kHz, the calculated offset is 0.1kHz. The current low-frequency trend component is then shifted along the frequency axis by 0.1kHz to complete the alignment. The alignment process for the other two types of components is the same. This alignment process eliminates frequency offset errors during signal acquisition, ensuring that the current component and the reference component are compared on the same characteristic dimension.

[0091] Furthermore, based on the alignment results, the quantization of the low-frequency overall offset, mid-frequency energy variability, and high-frequency distribution distortion is completed respectively.

[0092] Specifically, the process of determining the overall offset is as follows: first, calculate the difference between the corresponding matrix elements of the current low-frequency trend component and the low-frequency reference trend component after alignment; then, obtain the absolute value of all differences; finally, calculate the average value of these absolute values, which is the overall offset.

[0093] Specifically, the calculation process of energy variability is as follows: first, calculate the total energy of the current intermediate frequency fluctuation component and the intermediate frequency reference fluctuation component after alignment, and then divide the total energy of the current intermediate frequency component by the total energy of the reference component. The resulting ratio is the energy variability.

[0094] Specifically, the calculation process of the distribution distortion degree is as follows: First, the amplitude range of the current high-frequency noise component and the high-frequency reference noise component after alignment is statistically analyzed. The amplitude range is then divided into 20 intervals on average. The number of intervals is determined by statistically analyzing the amplitude distribution density of historical high-frequency components. Next, the probability density of the amplitude of the two components in each interval is calculated. Finally, the integral is completed by calculating the divergence between the two probability density distributions. The integral result is the distribution distortion degree.

[0095] Furthermore, by combining the preset weighting relationships, the three deviation indicators are synthesized to obtain the comprehensive deviation degree.

[0096] Specifically, the weighting relationship is determined by statistically analyzing the contribution of each component deviation index to the fault identification result under different fault modes. The contribution is obtained by calculating the correlation coefficient between each index and the fault degree. The weight value is equal to the proportion of the correlation coefficient of each index to the sum of the correlation coefficients of the three indices.

[0097] For example, the correlation coefficients of the low-frequency overall offset, mid-frequency energy variability, and high-frequency distribution distortion are statistically found to be 0.5, 0.3, and 0.2, respectively, and the corresponding weights are 0.5, 0.3, and 0.2, respectively.

[0098] Specifically, the synthesis process is as follows: multiply each deviation index by its corresponding weight, and then add the three product results together. The sum obtained is the deviation degree that characterizes the overall deviation of each scale.

[0099] For example, if the overall offset is 0.05, the energy variability is 1.2, and the distribution distortion is 0.2, the calculated value can be 0.425. The weight allocation reflects the different degrees of influence of each deviation index on the assessment of the connection point status, and the synthesized comprehensive deviation can fully reflect the overall abnormality of the connection point.

[0100] In summary, this embodiment accurately calls upon the reference component that matches the current electrical connection point's model, material, and structural parameters, ensuring the relevance and accuracy of the comparison benchmark. Alignment processing based on energy-heat conversion characteristics eliminates frequency offset errors, ensuring that the current component and the reference component are compared in the same feature dimension. Overall offset, energy variability, and distribution distortion are quantified dimensionally to accurately capture anomalies related to degradation at different scales. Combined with preset weights, a comprehensive deviation is synthesized to reflect the different contributions of each indicator to the state assessment, comprehensively reflecting the overall abnormal condition of the connection point. This provides a quantitative basis for subsequent three-dimensional state vector construction and degradation mode matching, effectively avoiding the one-sidedness of single-dimensional assessment and significantly improving the accuracy of subsequent degradation type and severity level determination.

[0101] S6. Based on the deviation combination of each scale component, and referring to the predetermined correspondence between degradation type, severity level, and remaining effective life, determine the specific degradation state of the electrical connection point and generate a diagnostic report, including: A three-dimensional state vector is constructed based on the deviation of the low-frequency trend component, the deviation of the mid-frequency fluctuation component, and the deviation of the high-frequency noise component. The three-dimensional state vector is matched with a preset degradation mode feature space, wherein the construction of the degradation mode feature space is based on the physical mechanism relationship between the components separated by multi-scale decomposition and the contact surface oxidation, micro-cracks, and surface film properties. The dominant degradation type and accompanying secondary degradation features are mapped based on the matching results. Based on the dominant degradation type and the accompanying secondary degradation characteristics, the corresponding severity level label and remaining effective lifetime range are indexed in the preset correspondence table. The severity level label, the remaining effective lifespan range, the current monitoring timestamp, and the location identifier of the electrical connection point are combined and packaged into a structured diagnostic report.

[0102] In embodiments of the present invention, a three-dimensional state vector is matched with a preset degradation mode feature space to map the dominant degradation type and accompanying secondary degradation features, including: Based on the historical failure case library, a degradation mode feature space is constructed with oxidation, cracking, and fouling as basis vectors, and an influence weight coefficient is associated with each basis vector; Determine the projection components of the three-dimensional state vector in the direction of each basis vector in the feature space of the degradation mode, and perform weighted correction on the projection components according to the influence weight coefficient; The degradation mode corresponding to the basis vector with the largest weighted projection component is selected as the dominant degradation type. At the same time, the degradation modes corresponding to the basis vectors whose projection components exceed the preset threshold are identified as the accompanying secondary degradation features.

[0103] The formula for identifying degradation type projection is as follows.

[0104] In the formula, Three-dimensional state vector; The first characteristic in the feature space of the degradation mode basis vectors This represents the vector dot product operation, which is to "determine the projection components of the three-dimensional state vector in the direction of each basis vector in the feature space of the degradation mode". Represents basis vectors The modulus length is used for normalization. This represents the "influence weight coefficient" that is associated with each basis vector. This represents the final projected component value after weighted correction.

[0105] Specifically, a three-dimensional state vector characterizing the state of electrical connection points is constructed based on the deviation of each scale component.

[0106] Specifically, the three dimensions of the three-dimensional state vector correspond to the overall offset of the low-frequency trend component, the energy variability of the mid-frequency fluctuation component, and the distribution distortion of the high-frequency noise component, respectively. The values ​​of the three dimensions are directly obtained by quantizing the corresponding deviation results in step S5.

[0107] Specifically, the three deviations are arranged in the order of low frequency, medium frequency, and high frequency to form a three-dimensional ordered array, which is the three-dimensional state vector. For example, if the overall offset obtained in step S5 is 0.05, the energy variability is 1.2, and the distribution distortion is 0.2, then the constructed three-dimensional state vector is [0.05, 1.2, 0.2].

[0108] Furthermore, a degradation mode feature space is constructed based on a historical failure case library, and the weighted projection components of the three-dimensional state vector in each basis vector direction in this space are calculated.

[0109] Specifically, a historical fault case library of the same type of electrical connection point is collected. The case library contains the degradation type and corresponding three-dimensional state vector at the time of the fault. Three typical degradation types, namely oxidation, cracking and contamination, are extracted from the case library as basis vectors of the feature space. The value of each basis vector is determined by statistically averaging the historical three-dimensional state vectors under the corresponding degradation type.

[0110] Specifically, the influence weighting coefficient is determined by statistically analyzing the frequency proportion of each type of degradation leading to connection point failure in historical failure cases. The frequency proportion is the number of failure cases caused by a certain type of degradation divided by the total number of failure cases. For example, if the number of failure cases caused by oxidation, cracking, and fouling in 1000 failure cases are 500, 300, and 200 respectively, then the corresponding influence weighting coefficients are... , , The values ​​are 0.5, 0.3, and 0.2 respectively.

[0111] Specifically, the calculation process for the projection components is as follows: First, calculate the dot product between the 3D state vector and each basis vector; then, normalize the dot product by dividing the result by the magnitude of the corresponding basis vector; finally, multiply the normalized result by the corresponding influence weight coefficient to obtain the weighted corrected projection components. .

[0112] For example, if the basis vectors The value is [0.04, 0.3, 0.1], its magnitude is calculated to be 0.32, and the three-dimensional state vector is... The dot product is 0.05 × 0.04 + 1.2 × 0.3 + 0.2 × 0.1 = 0.382. After normalization, it becomes 0.382 ÷ 0.32 ≈ 1.194. Multiplying by the weighting factor of 0.5, we get... ≈0.597; the calculation process for the projection components corresponding to the remaining basis vectors is the same as above.

[0113] Furthermore, based on the weighted and corrected projection component results, the dominant degradation type and accompanying secondary degradation characteristics are determined. The preset threshold is determined by statistically analyzing the projection component distribution of each basis vector direction under historical health conditions, and the 95th quantile of this distribution is taken as the preset threshold. For example, if the 95th quantile of the projection component distribution obtained from statistically analyzing 100 sets of health condition data is 0.3, then the preset threshold is set to 0.3.

[0114] Specifically, by comparing all weighted and corrected projection components, the degradation type represented by the basis vector corresponding to the projection component with the largest value is selected as the dominant degradation type. Simultaneously, the remaining projection components are compared one by one with a preset threshold, and the degradation types corresponding to projection components with values ​​exceeding the preset threshold are identified as secondary degradation features. For example, if... ≈0.597 ≈0.4 If the value is approximately 0.2 and the preset threshold is 0.3, then oxidation is selected as the dominant degradation type, and cracks are selected as the accompanying secondary degradation feature.

[0115] Furthermore, based on the identified degradation characteristics, the severity level label and remaining effective lifespan range are obtained by indexing in the preset correspondence table.

[0116] Specifically, the correlation data of different degradation type combinations, deviation ranges and severity levels, and remaining effective lifespan in historical failure cases are statistically analyzed. The degradation type combinations and deviation ranges are used as index items, and the severity level labels and remaining effective lifespan ranges are used as corresponding values ​​to form a structured correspondence table.

[0117] Specifically, the identified dominant degradation type, secondary degradation features, and deviation range of each component are used as joint index conditions to retrieve matching items in the preset correspondence table, and the severity level label and remaining effective lifetime interval corresponding to the matching item are extracted.

[0118] For example, if the dominant degradation type is oxidation and the secondary degradation feature is cracking, with an overall offset of 0.05, energy variability of 1.2, and distribution distortion of 0.2, the severity level label obtained through the joint index is level two, and the remaining effective lifetime range is 12 to 18 months.

[0119] Specifically, relevant information is integrated and packaged into a structured diagnostic report. The current monitoring timestamp is obtained through the clock module with millisecond-level accuracy. For example, if the monitoring time is 15:30:25.300 on December 22, 2025, the timestamp record will be 2025-12-22 15:30:25.300.

[0120] Specifically, the location of electrical connection points is determined by reading equipment ledger information, which includes the equipment number to which the connection point belongs and its specific installation location.

[0121] Furthermore, the severity level label, remaining effective lifespan, current monitoring timestamp, and electrical connection point location identifier are organized according to a preset structured format and encapsulated in XML format to form a complete diagnostic report. For example, the encapsulated diagnostic report includes core information such as the severity level label (Level 2), lifespan range (12 to 18 months), timestamp (2025-12-22 15:30:25.300), and location identifier (#1 transformer high-voltage side terminal).

[0122] In summary, this embodiment constructs a three-dimensional state vector based on deviations at various scales, comprehensively integrating multi-dimensional anomaly information to avoid the one-sidedness of single-indicator evaluation. It builds a degradation mode feature space based on physical mechanism correlations and a historical case library, accurately identifying dominant degradation types and secondary degradation characteristics through weighted projection matching, thus improving the accuracy of hazard cause determination. By using a preset correspondence table indexing severity levels and remaining effective lifespan intervals, it achieves quantitative assessment and early warning of degradation status. The core assessment results, monitoring timestamps, and location identifiers are encapsulated into a structured diagnostic report, providing complete information and standardized format, offering a clear basis for operation and maintenance decisions, while significantly improving the efficiency of early warning report generation, facilitating timely and targeted prevention and control measures, and effectively reducing the risk of electrical fires.

[0123] Example 2, as Figure 2 The diagram shown is a functional module diagram of an online monitoring and early warning system for electrical fire hazards provided in an embodiment of the present invention.

[0124] This invention discloses an online monitoring and early warning system for electrical fire hazards that can be installed in electronic devices. Depending on the functions implemented, such a system may include a current excitation module 101, a feature decomposition module 102, a feature decomposition module 103, a reference acquisition module 104, a component comparison module 105, and a report generation module 106. These modules, also referred to as units, are a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0125] In this embodiment, the functions of each module / unit are as follows: The current excitation module 101 is used to inject a set of excitation signals with a preset spectrum into the electrical connection point when the working current is carried there. The feature extraction module 102 is used to synchronously acquire the electrical response signal and thermal response signal generated by the excitation signal at the electrical connection point, and extract the energy and heat conversion efficiency at each spectral point to form a real-time efficiency feature spectrum. The feature decomposition module 103 is used to perform multi-scale decomposition on the real-time efficiency feature spectrum to separate the low-frequency trend component reflecting the overall oxidation state of the contact surface, the mid-frequency fluctuation component reflecting micro-cracks and debris, and the high-frequency noise component reflecting the surface film characteristics. The reference acquisition module 104 is used to acquire the historical reference spectrum library corresponding to the electrical connection point. The historical reference spectrum library contains reference components of each scale established under healthy conditions. The component comparison module 105 is used to compare the low-frequency trend component, the mid-frequency fluctuation component and the high-frequency noise component with the corresponding reference components at each scale in the historical reference spectrum library to obtain the deviation of each scale component. The report generation module 106 is used to determine the specific deterioration status of electrical connection points and generate a diagnostic report based on the combination of deviations of each scale component and with reference to the pre-determined correspondence between deterioration type, severity level and remaining effective life.

[0126] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0127] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0128] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0129] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0130] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for online monitoring and early warning of electrical fire hazards, characterized in that, The method includes: S1. When the electrical connection point carries the working current, inject a set of excitation signals with a preset spectrum into it; S2. Synchronously acquire the electrical response signal and thermal response signal generated by the excitation signal at the electrical connection point, and extract the energy and heat conversion efficiency at each spectrum point to form a real-time efficiency characteristic spectrum. S3. Perform multi-scale decomposition on the real-time efficiency characteristic spectrum to separate the low-frequency trend component reflecting the overall oxidation state of the contact surface, the mid-frequency fluctuation component reflecting micro-cracks and debris, and the high-frequency noise component reflecting the surface film characteristics. S4. Obtain the historical reference spectrum library corresponding to the electrical connection point. The historical reference spectrum library contains reference components of each scale established under healthy conditions. S5. Compare the low-frequency trend component, mid-frequency fluctuation component, and high-frequency noise component with the corresponding reference components at each scale in the historical reference spectrum library to obtain the deviation of each scale component. S6. Based on the combination of deviations of each scale component, and referring to the pre-determined correspondence between degradation type, severity level and remaining effective life, determine the specific degradation status of the electrical connection point and generate a diagnostic report.

2. The method for online monitoring and early warning of electrical fire hazards as described in claim 1, characterized in that, When the electrical connection point carries the operating current, injecting a set of excitation signals with a preset spectrum into it includes: A set of fundamental frequencies is selected based on the material and structural parameters of the electrical connection points, and the fundamental frequencies are sideband-extended to obtain the spectrum of the modulated excitation signal. The modulated excitation signal spectrum is injected into the working current path through a signal coupling device to form a composite current-carrying signal containing the excitation spectrum. The current component within a preset frequency range is separated from the composite current-carrying signal, and the power frequency carrier component is filtered out to obtain the excitation response signal characterizing the impedance characteristics of the electrical connection point.

3. The method for online monitoring and early warning of electrical fire hazards as described in claim 1, characterized in that, The synchronous acquisition of electrical response and thermal response signals generated by the excitation signal at the electrical connection point, and the extraction of their energy and heat conversion efficiency at each spectral point to form a real-time efficiency characteristic spectrum, includes: Within a preset synchronization time window, the electrical response signal is extracted from the composite current-carrying signal, and the surface thermal radiation change during the corresponding time period is captured by a non-contact sensor to obtain synchronized time-domain electrical response signal and thermal response signal. Synchronous spectral mapping of the time-domain electrical response signal and the thermal response signal yields the frequency domain distribution of electrical energy and the frequency domain distribution of thermal radiation, respectively. By correlating the amplitudes of corresponding spectral points in the frequency domain distribution of electrical energy with those in the frequency domain distribution of thermal radiation, a two-dimensional feature matrix characterizing the energy-to-heat conversion relationship is constructed, which serves as the real-time efficiency feature spectrum.

4. The method for online monitoring and early warning of electrical fire hazards as described in claim 1, characterized in that, The multi-scale decomposition of the real-time efficiency characteristic spectrum separates the low-frequency trend component reflecting the overall oxidation state of the contact surface, the mid-frequency fluctuation component reflecting micro-cracks and debris, and the high-frequency noise component reflecting the surface film characteristics, including: Based on the frequency domain energy concentration characteristics of the real-time efficiency characteristic spectrum, multiple target analysis scales are determined, and the center frequency and bandwidth parameters corresponding to each target analysis scale are obtained. Based on the center frequency and bandwidth parameters, a group of bandpass filters with different passband ranges are configured to form a multi-scale filter bank. The real-time efficiency characteristic spectrum is input into a multi-scale filter bank, and the low-frequency trend component, mid-frequency fluctuation component and high-frequency noise component are separated in parallel through the passband filtering effect of each filter.

5. The method for online monitoring and early warning of electrical fire hazards as described in claim 4, characterized in that, The method involves configuring a set of bandpass filters with different passband ranges based on center frequency and bandwidth parameters to form a multi-scale filter bank, including: The main characteristic frequencies of the energy distribution of each reference component are extracted by calling up the reference components at each scale in the historical reference spectrum library. Using the main characteristic frequency as a reference, the obtained center frequency and bandwidth parameters are calibrated and offset to obtain characteristic filter parameters that are adapted to the historical state of the current electrical connection point. Based on the characteristic filter parameters, the passband range of the corresponding filter in the multi-scale filter bank is reconstructed so that the separation target of the filter bank is aligned with the spectral features of the historical health state.

6. The method for online monitoring and early warning of electrical fire hazards as described in claim 1, characterized in that, The acquisition of the historical reference spectral library corresponding to the electrical connection point, the historical reference spectral library containing reference components at various scales established under healthy conditions, including: Under the health verification state of the electrical connection point, the excitation signal of the preset spectrum is repeatedly injected, and the electrical response and thermal response are collected simultaneously each time to obtain multiple sets of benchmark datasets. Multiple sets of benchmark datasets are processed to obtain the real-time efficiency feature spectrum for each time, and all real-time efficiency feature spectra are averaged to obtain a stable benchmark efficiency feature spectrum. The baseline efficiency characteristic spectrum is decomposed into multiple scales to separate the corresponding low-frequency baseline trend component, mid-frequency baseline fluctuation component and high-frequency baseline noise component. The low-frequency reference trend component, the mid-frequency reference fluctuation component, and the high-frequency reference noise component are normalized and then linked and stored in a dedicated database to form a historical reference spectrum library.

7. The method for online monitoring and early warning of electrical fire hazards as described in claim 6, characterized in that, The step of comparing the low-frequency trend component, mid-frequency fluctuation component, and high-frequency noise component with the corresponding benchmark components at each scale in the historical benchmark spectral library to obtain the deviation of each scale component includes: Retrieve the low-frequency reference trend component, mid-frequency reference fluctuation component, and high-frequency reference noise component corresponding to the current electrical connection point from the historical reference spectrum library; Based on the energy-heat conversion relationship characteristics revealed by the two-dimensional feature matrix, each component is aligned separately; Based on the alignment results, the overall offset of the low-frequency trend component, the energy variability of the mid-frequency fluctuation component, and the distribution distortion of the high-frequency noise component are quantified respectively. Based on the overall offset, energy variability, and distribution distortion, and combined with the preset weighting relationship, a deviation degree is synthesized to characterize the overall deviation at each scale.

8. The method for online monitoring and early warning of electrical fire hazards as described in claim 7, characterized in that, The method, based on the deviation combination of each scale component and referring to the predetermined correspondence between degradation type, severity level, and remaining effective life, determines the specific degradation state of the electrical connection point and generates a diagnostic report, including: A three-dimensional state vector is constructed based on the deviation of the low-frequency trend component, the deviation of the mid-frequency fluctuation component, and the deviation of the high-frequency noise component. The three-dimensional state vector is matched with a preset degradation mode feature space, wherein the degradation mode feature space is constructed based on the physical mechanism relationship between the components separated by multi-scale decomposition and the contact surface oxidation, micro-cracks, and surface film properties. The dominant degradation type and its accompanying secondary degradation features are mapped based on the matching results. Based on the dominant degradation type and the accompanying secondary degradation characteristics, the corresponding severity level label and remaining effective lifetime range are indexed in the preset correspondence table. The severity level label, the remaining effective lifespan range, the current monitoring timestamp, and the location identifier of the electrical connection point are combined and packaged into a structured diagnostic report.

9. The method for online monitoring and early warning of electrical fire hazards as described in claim 8, characterized in that, The process of matching the three-dimensional state vector with a preset degradation mode feature space to map the dominant degradation type and accompanying secondary degradation features includes: Based on the historical failure case library, a degradation mode feature space is constructed with oxidation, cracking, and fouling as basis vectors, and an influence weight coefficient is associated with each basis vector; Determine the projection components of the three-dimensional state vector in the direction of each basis vector in the feature space of the degradation mode, and perform weighted correction on the projection components according to the influence weight coefficient; The degradation mode corresponding to the basis vector with the largest weighted projection component is selected as the dominant degradation type. At the same time, the degradation modes corresponding to the basis vectors whose projection components exceed the preset threshold are identified as the accompanying secondary degradation features.

10. An online monitoring and early warning system for electrical fire hazards, characterized in that, The system is used to implement the online monitoring and early warning method for electrical fire hazards according to any one of claims 1-9, the system comprising: The current excitation module is used to inject a set of excitation signals with a preset spectrum into the electrical connection point when it carries the working current. The feature extraction module is used to synchronously acquire the electrical response signal and thermal response signal generated by the excitation signal at the electrical connection point, and extract the energy and heat conversion efficiency at each spectral point to form a real-time efficiency feature spectrum. The feature decomposition module is used to perform multi-scale decomposition on the real-time efficiency feature spectrum, separating the low-frequency trend component reflecting the overall oxidation state of the contact surface, the mid-frequency fluctuation component reflecting micro-cracks and debris, and the high-frequency noise component reflecting the surface film characteristics. The reference acquisition module is used to acquire the historical reference spectrum library corresponding to the electrical connection point. The historical reference spectrum library contains reference components of each scale established under healthy conditions. The component comparison module is used to compare the low-frequency trend component, mid-frequency fluctuation component, and high-frequency noise component with the corresponding reference components at each scale in the historical reference spectrum library to obtain the deviation of each scale component. The report generation module is used to determine the specific degradation status of electrical connection points and generate a diagnostic report based on the combination of deviations of each scale component and with reference to the pre-determined correspondence between degradation type, severity level and remaining effective life.

Images