Power distribution cabinet fault early warning system and method

Abstract

The invention relates to the technical field of power distribution cabinet safety monitoring, and discloses a power distribution cabinet fault early warning system and method, and the system comprises an acquisition module which comprises a sensor; the control module is electrically connected with the acquisition module and comprises a monitoring unit, a judgment unit, an early warning unit and a feature extraction unit; the monitoring unit is configured to collect sensor data and preprocess the sensor data; the feature extraction unit is configured to extract sensor data and perform windowing interpolation processing on the sensor data; the judgment unit is configured to input the sensor data into a pre-trained state model for comparison, and determine the current operation state of the power distribution cabinet based on a comparison result; the early warning unit is configured to determine abnormal areas based on the abnormal state and send corresponding early warning information based on the abnormal areas. According to the invention, harmonic abnormity caused by nonlinear load or equipment aging can be identified, and the early warning precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of safety monitoring of distribution cabinets, and more specifically, to a distribution cabinet fault warning system and method. Background Art

[0002] Distribution cabinets are the core equipment for power distribution and control in the power system, and are widely used in industrial, commercial, data center and civil power supply scenarios.

[0003] However, with the large-scale grid connection of new energy, the popularization of electric vehicle charging piles and the surge of industrial automation equipment, the limitations of its traditional architecture and monitoring are becoming increasingly prominent. First, the wide access of non-linear loads (such as frequency converters, LED power supplies, data center server power supplies) has led to a significant increase in the harmonic distortion rate (THD). The passive filters and fixed compensation capacitor banks relied on by traditional distribution cabinets are difficult to suppress high-frequency harmonics due to resonance risks, and may instead exacerbate the harmonic amplification phenomenon, causing problems such as overheating and bulging of capacitors and accelerated aging of cable insulation, and even deteriorating the regional power quality. These drawbacks will cause the power distribution system to be in a "passive response" state for a long time, not only reducing the power supply reliability, but also increasing the energy consumption.

[0004] Therefore, it is necessary to design a distribution cabinet fault warning system and method and system to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a distribution cabinet fault warning system and method, aiming to solve the problem of poor accuracy of current distribution cabinet safety monitoring.

[0006] On the one hand, the present invention proposes a distribution cabinet fault warning system, including:

[0007] An acquisition module, including sensors;

[0008] A control module, electrically connected to the acquisition module, the control module includes a monitoring unit, a judgment unit, a warning unit and a feature extraction unit;

[0009] The monitoring unit is configured to acquire sensor data and preprocess the sensor data;

[0010] The feature extraction unit is configured to extract the preprocessed sensor data and perform windowing interpolation processing on the sensor data;

[0011] The judgment unit is configured to input the sensor data into a pre-trained state model for comparison;

[0012] The judgment unit is further configured to determine the current operating state of the distribution cabinet based on the comparison result;

[0013] The warning unit is configured to determine an abnormal area based on an abnormal state and send corresponding warning information based on each abnormal area when the current operating state of the power distribution cabinet is abnormal.

[0014] Furthermore, when preprocessing the sensor data, it at least includes the following methods:

[0015] Pulse suppression, which is used to eliminate instantaneous pulse interference in the signal;

[0016] Wavelet denoising, which is used to eliminate continuous noise;

[0017] Sampling frequency unification, which is used to keep the data frequencies consistent;

[0018] Data normalization, which is used to unify the dimension.

[0019] Furthermore, when extracting the sensor data and performing windowing interpolation processing on the sensor data, it includes:

[0020] Multiplying the preprocessed sensor data point by point with the Nuttall window function based on the characteristics of the Nuttall window function to obtain a windowed signal;

[0021] Performing a Fourier transform on the windowed signal to determine its amplitude spectrum;

[0022] Based on the amplitude spectrum, extracting the main lobe spectral lines and determining the frequency offset coefficient based on the ratio method;

[0023] Determining a correction factor based on the characteristics of the Nuttall window function and performing frequency and phase correction on the sensor data.

[0024] Furthermore, when inputting the sensor data into a pre-trained state model for comparison, it includes:

[0025] Obtaining multi-channel time-series data of historical sensor data and extracting short-time high-frequency features based on large-kernel convolution;

[0026] Performing weighted fusion of the time-domain features and frequency-domain features on the short-time high-frequency features;

[0027] Extracting high-order features based on the fully connected layer from the weighted-fused time-domain features and frequency-domain features and mapping the high-order features to fault categories to form the state model;

[0028] Inputting the sensor data into the state model to obtain the fault classification probability and the health score.

[0029] Furthermore, when determining the current operating state of the power distribution cabinet based on the comparison result, it includes:

[0030] When the health score is lower than the threshold, a fault warning is triggered, and the fault category is determined based on the fault classification probability;

[0031] The fault categories include capacitor aging / damage, poor contact at busbar / cable connection points, non-linear load faults, arc discharge, resonant overvoltage, and insulation deterioration.

[0032] Further, when determining the abnormal area based on the abnormal state and sending corresponding warning information based on each abnormal area, it includes:

[0033] When the amplitude of the 5th or 7th harmonic > 8% of the nominal voltage, the phase offset > 5°, and the temperature of the capacitor housing continuously remains more than 15° higher than the ambient temperature, it is determined as capacitor aging / damage;

[0034] When the amplitude of the 3rd harmonic > 5%, the harmonic phase jitters periodically, and the temperature rise rate of the connection point > 1.5°C / min, it is determined as poor contact at the busbar / cable connection point;

[0035] When the amplitude of the high-frequency harmonic > 3%, the total harmonic distortion rate > 12%, and the proportion of non-integer harmonic energy > 10%, it is determined as a non-linear load fault;

[0036] When the energy of the impulse group in the 2kHz - 5MHz frequency band > 50%, the wide-spectrum noise floor rises, the even harmonics > 15%, and the discharge pulse interval shows a chaotic distribution, it is determined as arc discharge;

[0037] When the amplitude of a specific harmonic shows an exponential growth, the effective value fluctuation of the busbar voltage > 5%, and the distortion rate of the capacitor branch current > 20%, it is determined as resonant overvoltage;

[0038] When the discharge amplitude > 50pC, and the phase is concentrated near the voltage peak, and the discharge frequency increases by 3 times when the humidity > 85%, it is determined as insulation deterioration.

[0039] Further, when determining the abnormal area based on the abnormal state and sending corresponding warning information based on each abnormal area, it also includes:

[0040] When it is determined as capacitor aging / damage, the faulty capacitor bank is cut out, and the active power filter is started;

[0041] When it is determined as poor contact at the busbar / cable connection point, the fault point is marked, and forced cooling is carried out;

[0042] When it is determined as a non-linear load fault, load balancing control is started, and current limiting protection is carried out;

[0043] When it is determined as arc discharge, the faulty circuit is cut off;

[0044] When it is determined as resonant overvoltage, part of the capacitor bank is cut out, and the resonance point is adjusted;

[0045] When it is determined that the insulation has deteriorated, stop operation.

[0046] Furthermore, when the warning unit determines arc discharge or resonant overvoltage, generate a first-level warning message;

[0047] When the warning unit determines capacitor aging / damage or poor contact at the busbar / cable connection point or insulation deterioration, generate a second-level warning message;

[0048] When the warning unit determines a non-linear load fault, generate a third-level warning message.

[0049] Furthermore, it further includes: a remote monitoring module, the remote monitoring module is electrically connected to the warning unit, and the remote monitoring module is configured to monitor the operating state of the warning unit and send the operating state monitoring information to the Internet and the cloud data platform.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows: By performing windowing interpolation processing on the sensor data, the interference of spectral leakage and fence effect on harmonic measurement is suppressed, the detection accuracy of harmonic distortion rate and each harmonic component is improved, enabling the system to identify harmonic anomalies caused by non-linear loads or equipment aging, providing a reliable data basis for fault warning. Through the collaborative analysis of multi-dimensional data, composite fault characteristics (such as harmonic resonance and temperature rise coupling) that cannot be reflected by a single parameter can be captured, thereby improving the coverage and comprehensiveness of fault detection. The judgment unit uses a pre-trained 1D-CNN state model, which can automatically extract deep features from the original time-series data and identify hidden fault modes that are difficult to capture by traditional threshold methods. The model can quantify the equipment health score by comparing real-time data with the historical normal state, and make joint decisions by combining multi-source information such as harmonic parameters and temperature gradients, reducing the dependence on manual experience, reducing the risk of missed alarms or false alarms caused by subjective misjudgment, and shortening the fault identification time. The warning unit realizes hierarchical warning and positioning based on the type and severity of the abnormal state. The hierarchical mechanism further differentiates the urgency, ensuring that high-risk faults trigger protection actions first, and low-risk anomalies enter the preventive maintenance process. Through early warning and trend analysis, the system can detect potential risks in the initial stage of equipment deterioration.

[0051] On the other hand, the present application also provides a method for warning of distribution cabinet faults, which is applied to the distribution cabinet fault warning system according to any one of claims 1-9, and is characterized by including:

[0052] Obtain sensor data and perform preprocessing on the sensor data;

[0053] Extract the preprocessed sensor data and perform windowing interpolation processing;

[0054] Input the sensor data into a pre-trained state model for comparison;

[0055] Determine the current operating state of the power distribution cabinet based on the comparison result;

[0056] When the current operating state of the power distribution cabinet is abnormal, determine the abnormal area based on the abnormal state and send corresponding warning information based on each abnormal area.

[0057] It can be understood that the above-mentioned power distribution cabinet fault warning method has the same beneficial effects, which will not be elaborated here. Brief Description of the Drawings

[0058] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered as limiting the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0059] Figure 1 It is a functional block diagram of the power distribution cabinet fault warning system provided by an embodiment of the present invention.

[0060] Figure 2 It is a flowchart of the power distribution cabinet fault warning method provided by an embodiment of the present invention. Detailed Embodiments

[0061] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. Hereinafter, the present invention will be described in detail with reference to the drawings and in combination with the embodiments.

[0062] In some embodiments of the present application, referring to Figure 1 as shown, a power distribution cabinet fault warning system includes:

[0063] An acquisition module, including sensors.

[0064] A control module, electrically connected to the acquisition module, and the control module includes a monitoring unit, a judgment unit, a warning unit, and a feature extraction unit.

[0065] The monitoring unit is configured to acquire sensor data and preprocess the sensor data.

[0066] The feature extraction unit is configured to extract the preprocessed sensor data and perform windowing interpolation processing on the sensor data.

[0067] The judgment unit is configured to input the sensor data into a pre-trained state model for comparison.

[0068] The judgment unit is also configured to determine the current operating state of the power distribution cabinet based on the comparison result.

[0069] The warning unit is configured to, when the current operating state of the power distribution cabinet is abnormal, determine the abnormal area based on the abnormal state and send corresponding warning information based on each abnormal area.

[0070] Specifically, the sensor data includes, but is not limited to, current transformers for subsequent analysis of harmonic distortion rate and current unbalance degree, and detection of abnormal nonlinear loads such as overload and short circuit, voltage transformers for detecting voltage fluctuations, sag / swell time, identifying resonant overvoltage and interharmonic components, high-frequency current sensors (covering the frequency band of 2 kHz to 5 MHz) for capturing arc discharges, high-frequency harmonics and transient pulses, temperature sensors for detecting local overheating caused by poor contact, warning of abnormal temperature rise of capacitors and heat dissipation failure, ultra-high frequency sensors for detecting defects such as poor insulation and internal discharge, ultrasonic sensors for capturing mechanical vibrations caused by discharges and assisting in locating the position of the discharge point, humidity sensors for assisting in verifying arc discharges. By preprocessing the sensor data and then performing windowing interpolation processing, the windowed sensor data is obtained, and then it is compared through a pre-trained state model. The pre-trained state model is a 1D-CNN state model. Through the data output by the state model, key information such as fault probability distribution, health score, and harmonic parameters can be obtained. Furthermore, through probability + location warnings, the staff can identify abnormal areas and abnormal situations.

[0071] In some embodiments of the present application, when preprocessing the sensor data, at least the following methods are included:

[0072] Pulse suppression, used to eliminate instantaneous pulse interference in the signal.

[0073] Wavelet denoising, used to eliminate continuous noise.

[0074] Sampling frequency unification, used to keep the data frequencies consistent.

[0075] Data normalization, used to unify the dimensions.

[0076] It can be understood that in the power distribution cabinet fault warning system, the preprocessing of sensor data is the core link to ensure the reliability of subsequent analysis and decision-making. Through key technical means such as pulse suppression, wavelet denoising, sampling frequency unification, and data normalization, the system can optimize the data quality and provide high-quality input for fault diagnosis and condition assessment. The pulse suppression technology can identify and eliminate instantaneous spike interferences in the signal (such as lightning surges, switch arcs, etc.), avoiding the misguidance of subsequent analysis by abnormal pulses. It eliminates non-fault-related instantaneous noises, preventing the misjudgment of external interferences as equipment faults (such as misidentifying electromagnetic interference as arc discharge). It suppresses the impact of high-amplitude pulses on the acquisition circuit and extends the service life of sensors and data acquisition modules. While filtering out interferences, it ensures that real fault transient signals (such as weak partial discharge pulses) are not masked, enabling the identification of early faults. Wavelet denoising separates the continuous noises (such as thermal noise, power frequency harmonic interference) and effective components in the signal through multi-scale decomposition and threshold processing. While retaining the signal mutation points (such as harmonic distortion, phase jump), it suppresses the background noise, making it easier to extract fault features. Compared with traditional Fourier filtering, wavelet transform can process time-varying signals (such as non-steady-state harmonics caused by load mutations), avoiding signal distortion caused by frequency-domain filtering. The unification of the sampling frequencies of multi-sensor data solves the data alignment problem caused by asynchronous sampling through interpolation or resampling techniques, ensuring the strict synchronization of the timestamps of parameters such as current, voltage, and temperature, and avoiding feature misalignment caused by sampling delays (such as the phase deviation between temperature rise and harmonic mutation). Data normalization maps sensor data with different dimensions to a unified range (such as [-1, 1]) through linear or non-linear transformation, eliminating the feature scale differences (such as the current unit being amperes and the temperature unit being degrees Celsius), and avoiding gradient instability caused by parameter amplitude differences during model training.

[0077] In some embodiments of the present application, when extracting sensor data and performing windowing interpolation processing on the sensor data, it includes:

[0078] Multiplying the preprocessed sensor data point by point with the Nuttall window function based on the Nuttall window function characteristics to obtain a windowed signal.

[0079] Performing a Fourier transform on the windowed signal to determine its amplitude spectrum.

[0080] Based on the amplitude spectrum, extracting the main lobe spectral lines and determining the frequency deviation coefficient based on the ratio method.

[0081] Determining a correction factor based on the Nuttall window function characteristics and performing frequency and phase correction on the sensor data.

[0082] It can be understood that the window function uses a Blackman-Harris window (sidelobe attenuation -92dB) to suppress spectral leakage, and the Miura line interpolation correction: Among them, Y k-1 , Y k , Y k+1 are the amplitudes of three adjacent spectral lines near the main lobe of the harmonic after Fourier transform, representing the distribution of harmonic energy in the frequency domain. δ is the frequency deviation coefficient, reflecting the deviation between the actual signal frequency and the discrete spectrum frequency points (caused by asynchronous sampling). N is the number of sampling points, that is, the number of samples within a single signal period, which determines the frequency resolution. W(δ) is the window function spectrum correction factor, representing the amplitude attenuation characteristic of the selected window function (such as the Blackman-Harris window) at the frequency deviation δ, used to compensate for the amplitude error caused by windowing. A is the corrected harmonic amplitude, which is the true amplitude estimate after eliminating the fence effect through the interpolation algorithm. Specifically, apply the Blackman-Harris window to the current signal to suppress spectral leakage, perform Fourier transform and extract Y k-1 , Y k , Y k+1 , and then calculate the frequency deviation coefficient δ through Newton-Raphson iteration to correct the harmonic amplitude, phase and frequency.

[0083] In some embodiments of the present application, when inputting sensor data into a pre-trained state model for comparison, it includes:

[0084] Obtain the multi-channel time series data of historical sensor data, and extract short-time high-frequency features based on large kernel convolution.

[0085] Perform weighted fusion on the short-time high-frequency features in the time domain and frequency domain features.

[0086] Extract high-order features based on the fully connected layer for the weighted fusion of time domain features and frequency domain features, and map the high-order features to the fault category to form a state model.

[0087] Input the sensor data into the state model to obtain the fault classification probability and the health score.

[0088] Specifically, divide the sensor data into multiple channels and extract short-time high-frequency features through large kernel convolution: Among them, x is the input tensor, K is the convolution kernel size, representing the time series length covered by each convolution kernel, w k,j is the weight parameter of the j-th convolution kernel at position k, s is the stride, which determines the time series compression rate of the output feature map, b j is the bias term of the j-th convolution kernel. The large kernel convolution can capture short-time high-frequency transient features (such as harmonic spikes, arc pulses), and then extract time domain and frequency domain features, extract high-order features through feature fusion, and then the fault classification and health score can be obtained.

[0089] In some embodiments of the present application, when determining the current operating state of the power distribution cabinet based on the comparison result, it includes:

[0090] When the health score is lower than the threshold, a fault warning is triggered, and the fault category is determined based on the fault classification probability.

[0091] The fault categories include capacitor aging / damage, poor contact at the busbar / cable connection point, non-linear load fault, arc discharge, resonant overvoltage, and insulation deterioration.

[0092] Specifically, the threshold is determined by the health score, and the threshold is dynamically adjusted by the environmental humidity and temperature: η = β * υ + (1 - β) * Λ, where η is the threshold, β is the weight of historical data, representing the influence degree of the long-term health state of the device, υ is the historical baseline health, Λ is the environmental correction factor. When the environmental humidity > 80% or the temperature > 50°C, the environmental correction factor is 0.9, and in other cases it is 1.0. If the health score < the threshold, the fault classification determination is entered. When the health ≥ the threshold, it is marked as normal, and the historical baseline health is updated.

[0093] In some embodiments of the present application, when determining the abnormal area based on the abnormal state and sending corresponding warning information based on each abnormal area, it includes:

[0094] When the amplitude of the 5th or 7th harmonic > 8% of the rated voltage, the phase offset > 5°, and the temperature of the capacitor housing continuously remains more than 15° higher than the ambient temperature, it is determined as capacitor aging / damage.

[0095] When the amplitude of the 3rd harmonic > 5%, the harmonic phase has periodic jitter, and the temperature rise rate of the connection point > 1.5°C / min, it is determined as poor contact at the busbar / cable connection point.

[0096] When the amplitude of the high-frequency harmonic > 3%, the total harmonic distortion rate > 12%, and the energy ratio of non-integer harmonics > 10%, it is determined as a non-linear load fault.

[0097] When the pulse group energy in the 2kHz - 5MHz frequency band > 50%, the wide-spectrum noise floor rises, the even harmonics > 15%, and the discharge pulse interval shows a chaotic distribution, it is determined as arc discharge.

[0098] When the amplitude of a specific harmonic shows exponential growth, the effective value of the busbar voltage fluctuates > 5%, and the current distortion rate of the capacitor branch > 20%, it is determined as resonant overvoltage.

[0099] When the discharge amplitude > 50pC, and the phase is concentrated near the voltage peak, and the discharge frequency increases by 3 times when the humidity > 85%, it is determined as insulation deterioration.

[0100] Specifically, when the amplitudes of each harmonic (such as the 5th and 7th harmonics) increase significantly (>8% of the nominal voltage), the phase shift amount (>5°), and the temperature of the capacitor housing continuously exceeds the ambient temperature by more than 15°, it is determined that the capacitor is aged / damaged. For example, when the amplitude of the 5th harmonic continuously exceeds 8% for 3 minutes and the temperature gradient >2° / min, and the model output probability >85%, it is located to the corresponding group in the capacitor cabinet. When the amplitude of the 3rd harmonic suddenly increases (>5%), accompanied by periodic jitter of the odd harmonic phase (±3° to 10°), the temperature rise rate of the connection point >1.5°C / min, the absolute temperature >80°C, and the three-phase current unbalance >15% (normal operating condition <5%), it is determined that the busbar / cable connection point is poorly contacted. For example, when the amplitude of the 3rd harmonic >5% + the temperature rise rate >1°C / min + the current unbalance >10%, and the model probability >70%, positioning is performed. When the amplitudes of high-frequency harmonics (such as the 17th and 19th harmonics) are abnormal (>3%), the total harmonic distortion rate (THD) >12%, and the energy ratio of non-integer harmonics (such as 12.5 Hz and 150 Hz) suddenly increases, it is determined that there is a non-linear load fault (such as a frequency converter or an LED power supply). When the energy of the impulse group in the frequency band of 2 kHz to 5 MHz suddenly increases, the wide-spectrum noise base rises, the even harmonics (the 2nd and 4th harmonics) are abnormally active, and the discharge pulse interval shows a chaotic distribution (different from traditional power frequency interference), it is determined that there is an arc discharge (partial discharge or switch arc). When the amplitudes of specific harmonics (such as the 5th and 11th harmonics) increase exponentially (the increase amplitude >50% within 10 minutes), the effective value of the busbar voltage fluctuates >±10% (exceeding the GB / T 12325 standard), and the distortion rate of the capacitor branch current >25%, it is determined that there is a resonant overvoltage. When the discharge amplitude >50 pC, and the phase is concentrated near the voltage peak, and the discharge frequency increases by 3 times when the humidity >85%, it is determined that the insulation is deteriorated (the insulation of the busbar or cable).

[0101] In some embodiments of the present application, when determining the abnormal area based on the abnormal state and sending corresponding warning information based on each abnormal area, it further includes:

[0102] When it is determined that the capacitor is aged / damaged, the faulty capacitor bank is cut out, and the active power filter is started.

[0103] When it is determined that the busbar / cable connection point is poorly contacted, the fault point is marked, and the temperature is forced to drop.

[0104] When it is determined that there is a non-linear load fault, the load balancing control is started, and current limiting protection is performed.

[0105] When it is determined that there is an arc discharge, the faulty circuit is cut off.

[0106] When it is determined that there is a resonant overvoltage, part of the capacitor bank is cut out, and the resonance point is adjusted.

[0107] When it is determined that the insulation is deteriorated, the operation is stopped.

[0108] In some embodiments of the present application, when the warning unit determines arc discharge or resonant overvoltage, a first-level warning message is generated.

[0109] When the warning unit determines capacitor aging / damage, poor contact at the busbar / cable connection point, or insulation deterioration, a second-level warning message is generated.

[0110] When the warning unit determines a non-linear load fault, a third-level warning message is generated.

[0111] Specifically, the warning levels are divided into first-level (urgent), second-level (important), and third-level (general) to ensure differential treatment of faults of different severities. For example: Arc discharge and resonant overvoltage (first-level warning): directly trigger the highest-priority response (such as cutting off the circuit or adjusting the resonant point) to avoid catastrophic consequences such as equipment explosion or fire. Capacitor aging, poor busbar contact, insulation deterioration (second-level warning): prevent the expansion of faults through measures such as cutting out the faulty capacitor bank, forced cooling, or shutdown. Non-linear load fault (third-level warning): adopt load balancing and current limiting protection to reduce losses while ensuring power supply continuity.

[0112] Avoid problems of "over-treatment" or "insufficient response". Start the APF (active power filter) when a capacitor fails: when a traditional reactive power compensation device (such as a capacitor bank) fails, the APF can seamlessly take over the harmonic suppression and reactive power compensation functions to avoid voltage fluctuations or sudden drops in power factor. Adjust the resonant point during resonant overvoltage: quickly eliminate the resonant risk by cutting out some capacitor banks and reconfiguring the filtering parameters, which is more flexible than traditional fixed LC filters. Mark and force cooling for poor busbar / cable contact: use infrared temperature measurement or impedance analysis to locate the high-temperature point and suppress thermal runaway by forced air cooling or reducing the load current to avoid melting of the connection point. Immediately shut down in case of insulation deterioration: detect the deterioration trend in advance through partial discharge detection or insulation resistance monitoring to prevent short-circuit accidents caused by insulation breakdown. Load balancing and current limiting protection: for faults of non-linear loads such as frequency converters and rectifiers, dynamically adjust the phase current distribution to suppress harmonic pollution, and at the same time avoid equipment overload damage through current limiting protection.

[0113] In some embodiments of the present application, it further includes: a remote monitoring module, which is electrically connected to the warning unit and is configured to monitor the operating state of the warning unit and send the operating state monitoring information to the Internet and the cloud data platform.

[0114] Specifically, by introducing the remote monitoring module and deeply integrating it with the cloud data platform, a complete ecological system for intelligent management of distribution cabinets has been built, realizing all-weather, all-round monitoring and management of the equipment operation status. The direct connection between the remote monitoring module and the early warning unit ensures the real-time transmission of abnormal information, allowing managers to grasp the operation status of the distribution cabinet anytime and anywhere. The system uploads the operation status monitoring information of the early warning unit to the Internet and the cloud data platform in real time. The centralized monitoring function based on the cloud platform realizes the unified management of multiple distributed distribution cabinets. The operation and maintenance personnel can intuitively understand the real-time status of each site through the visual interface, which improves the management efficiency. Secondly, the historical data stored in the cloud provides a solid foundation for in-depth analysis. By mining the long-term accumulated early warning information, switching records and other data, it can identify the operation rules of equipment, predict the life cycle, and guide the formulation of preventive maintenance plans. In terms of emergency response, the remote monitoring module realizes the instant push of fault information and supports multi-channel alarms such as SMS, email, and mobile applications to ensure that relevant personnel are informed of abnormal situations in the first time, and gain valuable time for rapid disposal.

[0115] In summary, the beneficial effects of the present invention are: by windowing and interpolating sensor data, the interference of spectrum leakage and fence effect on harmonic measurement is suppressed, the detection accuracy of harmonic distortion rate and each harmonic component is improved, so that the system can identify harmonic anomalies caused by nonlinear loads or equipment aging, and provide a reliable data basis for fault warning. Through the collaborative analysis of multi-dimensional data, it can capture complex fault characteristics that cannot be reflected by a single parameter (such as harmonic resonance and temperature rise coupling), thereby improving the coverage and comprehensiveness of fault detection. The judgment unit adopts a pre-trained 1D-CNN state model, which can automatically extract deep features from the original time series data and identify hidden fault modes that are difficult to capture by traditional threshold methods. The model can quantify the equipment health score by comparing real-time data with historical normal states, and make joint decisions based on multi-source information such as harmonic parameters and temperature gradients, reducing the reliance on manual experience, reducing the risk of missed reports or false reports due to subjective misjudgment, and shortening the fault identification time. The early warning unit implements graded early warning and positioning based on the type and severity of abnormal conditions. The grading mechanism further distinguishes the degree of urgency, ensuring that high-risk faults trigger protection actions first, while low-risk abnormalities enter the preventive maintenance process. Through early warning and trend analysis, the system can detect potential risks in the early stages of equipment degradation.

[0116] In another preferred embodiment based on the above embodiment, refer to Figure 2 As shown, this embodiment provides a distribution cabinet fault warning method, which is used to apply the above distribution cabinet fault warning system, including:

[0117] S100: Acquire sensor data and pre-process the sensor data.

[0118] S200: Extract the preprocessed sensor data for windowing interpolation processing.

[0119] S300: Input the sensor data into a pre-trained state model for comparison.

[0120] S400: Determine the current operating state of the power distribution cabinet based on the comparison result.

[0121] S500: When the current operating state of the power distribution cabinet is abnormal, determine the abnormal area based on the abnormal state and send corresponding warning information based on each abnormal area.

[0122] Specifically, through the preprocessing of sensor data, it is possible to eliminate noise interference, abnormal fluctuations, and random errors caused by environmental factors in the original data, thereby improving the credibility and consistency of the data. The preprocessed data better conforms to the actual operating characteristics, laying a reliable foundation for subsequent analysis. At the same time, the windowing interpolation processing technology further optimizes the continuity and integrity of the data sequence, can make up for data missing problems caused by sensor sampling intervals or short-term failures, and ensures the complete capture of key feature information.

[0123] Inputting the preprocessed data into a pre-trained state model for comparison can break through the limitations of traditional threshold alarms. The state model establishes a dynamic benchmark for the normal operation of the device through learning historical data, can comprehensively consider the correlation relationships between multi-dimensional parameters, and identify hidden abnormalities that are difficult to detect by a single indicator. For example, when parameters such as temperature, current, and vibration are all within the normal range when analyzed separately, but their co-variation trend deviates from the model expectation, an alarm can still be triggered, thus achieving the sensitive capture of early fault signs, improving the sensitivity of fault detection, and reducing the probability of missed alarms and false alarms.

[0124] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, system, or computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of flows and / or blocks in the flowchart and / or block diagram. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate for implementing in the process Figure 1 one or more of the processes and / or blocks Figure 1 a device for the functions specified in one or more of the blocks.

[0126] These computer program instructions can also be stored in a computer-readable storage that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable storage generate a manufactured article including instruction means, and the instruction means implements in the process Figure 1 one or more of the processes and / or blocks Figure 1 the functions specified in one or more of the blocks.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide for implementing in the process Figure 1 one or more of the processes and / or blocks Figure 1 steps for the functions specified in one or more of the blocks.

[0128] 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 them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the specific implementation manners of the present invention can still be modified or equivalently replaced, and any modification or equivalent replacement without departing from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A power distribution cabinet fault warning system, characterized in that, Including: A collection module, including sensors; A control module, electrically connected to the collection module, the control module includes a monitoring unit, a judgment unit, an early warning unit and a feature extraction unit; The monitoring unit is configured to collect sensor data and preprocess the sensor data; The feature extraction unit is configured to extract the preprocessed sensor data and perform windowing interpolation processing on the sensor data; The judgment unit is configured to input the sensor data into a pre-trained state model for comparison; The judgment unit is further configured to determine the current operating state of the power distribution cabinet based on the comparison result; The early warning unit is configured to, when the current operating state of the power distribution cabinet is abnormal, determine the abnormal area based on the abnormal state and send corresponding warning information based on each abnormal area.

2. The power distribution cabinet fault warning system according to claim 1, wherein, When preprocessing the sensor data, the following methods are included: Pulse suppression, used to eliminate instantaneous pulse interference in the signal; Wavelet denoising, used to eliminate continuous noise; Unifying the sampling frequency, used to keep the data frequencies consistent; Data normalization, used to unify the dimension.

3. The power distribution cabinet fault warning system according to claim 2, characterized in that, When extracting the sensor data and performing windowing interpolation processing on the sensor data, it includes: Multiplying the preprocessed sensor data point by point by the Nuttall window function to obtain a windowed signal; Performing Fourier transform on the windowed signal to determine its amplitude spectrum; Based on the amplitude spectrum, extracting the main lobe spectral line and determining the frequency deviation coefficient based on the ratio method; Determining the correction factor based on the Nuttall window function and performing frequency and phase correction on the sensor data.

4. The power distribution cabinet fault warning system according to claim 3, wherein When inputting the sensor data into a pre-trained state model for comparison, it includes: Obtaining multi-channel time series data of historical sensor data and extracting short-time high-frequency features based on large kernel convolution; Performing weighted fusion of the short-time high-frequency features in the time domain and the frequency domain; Extracting high-order features based on the fully connected layer from the weighted fusion of the time domain feature and the frequency domain feature, and mapping the high-order features to the fault category to form the state model; Inputting the sensor data into the state model to obtain the fault classification probability and the health score.

5. The power distribution cabinet fault warning system according to claim 4, wherein, When determining the current operating state of the power distribution cabinet based on the comparison result, it includes: When the health score is lower than the threshold, triggering a fault warning and determining the fault category based on the fault classification probability; The fault categories include capacitor aging / damage, poor contact at the busbar / cable connection point, nonlinear load fault, arc discharge, resonant overvoltage, and insulation deterioration.

6. The power distribution cabinet fault warning system according to claim 5, characterized in that, When determining the abnormal area based on the abnormal state and sending corresponding warning information based on each abnormal area, it includes: When the amplitude of the 5th or 7th harmonic > 8% of the nominal voltage, the phase offset > 5°, and the temperature of the capacitor housing continuously remains more than 15° higher than the ambient temperature, it is determined as capacitor aging / damage; When the amplitude of the 3rd harmonic > 5%, the harmonic phase jitters periodically, and the temperature rise rate at the connection point > 1.5°C / min, it is determined as poor contact at the busbar / cable connection point; When the amplitude of the high-frequency harmonic > 3%, the total harmonic distortion rate > 12%, and the energy ratio of non-integer harmonics > 10%, it is determined as a nonlinear load fault; When the impulse group energy in the frequency band of 2 kHz to 5 MHz > 50%, the wide-spectrum noise base is lifted, the even harmonics > 15%, and the discharge pulse intervals are chaotically distributed, it is determined as arc discharge; When the amplitude of a specific harmonic grows exponentially, the effective value fluctuation of the bus voltage > 5%, and the distortion rate of the capacitor branch current > 20%, it is determined as resonance overvoltage; When the discharge amplitude > 50 pC, and the phase is concentrated near the voltage peak, and the discharge frequency increases by 3 times when the humidity > 85%, it is determined as insulation deterioration.

7. The power distribution cabinet fault warning system according to claim 6, wherein, When determining the abnormal area based on the abnormal state and sending corresponding warning information based on each abnormal area, it further includes: When it is determined that the capacitor is aged / damaged, the faulty capacitor bank is cut out, and the active power filter is started; When it is determined that the contact of the busbar / cable connection point is poor, the fault point is marked, and the temperature is forced to drop; When it is determined that the non-linear load fails, the load balancing control is started, and current limiting protection is carried out; When it is determined as arc discharge, the faulty circuit is cut off; When it is determined as resonance overvoltage, part of the capacitor bank is cut out, and the resonance point is adjusted; When it is determined as insulation deterioration, the operation is stopped.

8. The power distribution cabinet fault warning system according to claim 7, wherein When the warning unit determines it as arc discharge or resonance overvoltage, a first-level warning information is generated; When the warning unit determines it as capacitor aging / damage or busbar / cable connection point poor contact or insulation deterioration, a second-level warning information is generated; When the warning unit determines it as non-linear load failure, a third-level warning information is generated.

9. The power distribution cabinet fault warning system according to claim 8, wherein It further includes: A remote monitoring module, the remote monitoring module is electrically connected to the warning unit, and the remote monitoring module is configured to monitor the operation state of the warning unit and send the operation state monitoring information to the Internet and the cloud data platform.

10. A power distribution cabinet fault warning method, applied to the power distribution cabinet fault warning system according to any one of claims 1-9, characterized in that, It includes: Obtain sensor data and preprocess the sensor data; Extract the preprocessed sensor data for windowing interpolation processing; Input the sensor data into a pre-trained state model for comparison; Determine the current operation state of the power distribution cabinet based on the comparison result; When the current operation state of the power distribution cabinet is abnormal, determine the abnormal area based on the abnormal state and send corresponding warning information based on each abnormal area.

Images