A method and system for analyzing and providing early warning of temperature changes in the busbar of a power distribution cabinet.

Abstract

This invention discloses a method and system for analyzing and warning of temperature changes on the busbar of a power distribution cabinet, relating to the field of power distribution cabinet technology. The method includes: collecting temperature data on the three-phase busbar of the power distribution cabinet; dividing the temperature sequence into spatiotemporal units; calculating the spatial and temporal characteristics of each spatiotemporal unit; performing frequency domain analysis on the temperature data of the spatiotemporal units; decomposing the temperature data to obtain low-frequency, mid-frequency, and high-frequency components; extracting sub-components at different time scales; and calculating temporal coupling characteristics, spatial coupling characteristics, and physical coupling characteristics; collecting fault-free data to establish a normal mode library; collecting historical fault data to establish a fault mode library; calculating the deviation of features from the baseline in real time; marking and updating potential new faults; locating temperature anomalies based on distributed monitoring; calculating reconstruction errors at each scale; determining the scale of anomaly origin; and triggering graded warnings. This method solves the problem of insufficient fault identification capability in traditional monitoring methods.

Description

Technical Field

[0001] This invention relates to the field of power distribution cabinet technology, specifically to a method and system for analyzing and providing early warning of changes in busbar temperature in power distribution cabinets. Background Technology

[0002] As an important component of the power system, the safe and stable operation of the distribution network is crucial. The busbars in the distribution cabinet undertake the core task of power transmission, and their temperature changes directly reflect the operating status of the equipment. In the existing technology, the temperature monitoring of the distribution cabinet busbars is mainly achieved by installing a limited number of temperature sensors, such as thermocouples or infrared temperature measuring devices. These sensors are usually deployed at key locations on the busbars, and by periodically collecting temperature data and combining it with threshold judgments, an early warning function is achieved.

[0003] However, since only a small number of sensors are deployed at key locations on the busbar, it is impossible to capture the temperature gradient changes along the entire length of the busbar in real time, which may lead to missed detection of local anomalies. Secondly, the early warning mechanism relies on static thresholds, which cannot adapt to the effects of dynamic load changes and ambient temperature fluctuations. For example, under high load or high temperature environments, fixed thresholds are prone to false alarms or missed alarms, reducing the accuracy of early warnings. In addition, existing technologies mainly rely on time-domain threshold judgments and lack frequency-domain feature analysis of temperature changes, making it impossible to effectively distinguish between normal fluctuations and potential faults, such as poor contact or insulation aging. At the same time, the temperature data is not decomposed into multiple scales, making it difficult to identify the abnormal transmission patterns at different time scales. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for analyzing and warning of temperature changes in distribution cabinet busbars. It utilizes distributed fiber optic sensors to achieve real-time monitoring of the temperature gradient along the entire length of the busbar, and combines dynamic noise threshold adaptive adjustment technology to solve the false alarm problem caused by static threshold changes under load and environmental variations. By employing frequency domain analysis and multi-scale wavelet decomposition techniques, it accurately extracts the frequency domain and time-scale features of temperature data, effectively distinguishing between normal fluctuations and potential faults. Furthermore, by establishing dynamically updated normal mode and fault mode libraries, combined with cluster analysis and a hierarchical early warning mechanism, it significantly improves the accuracy and timeliness of anomaly detection, solving the problems of localized missed detections, high false alarm rates, and insufficient fault identification capabilities in traditional monitoring methods.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for analyzing and providing early warning of temperature changes in a distribution cabinet busbar, comprising:

[0008] Temperature data is collected from the three-phase busbar of the distribution cabinet. The temperature data is preprocessed and a smoothed temperature sequence is output. The temperature sequence is divided into spatiotemporal units, and the spatial and temporal characteristics of each spatiotemporal unit are calculated.

[0009] Frequency domain analysis is performed on the temperature data of the spatiotemporal unit. The frequency band boundary is determined by fast Fourier transform. Low-frequency, mid-frequency and high-frequency empirical wavelet basis functions are constructed. The temperature data is decomposed to obtain low-frequency, mid-frequency and high-frequency components. Multi-level wavelet decomposition is performed on each component to extract sub-components at different time scales. Temporal coupling characteristics, spatial coupling characteristics and physical coupling characteristics are calculated.

[0010] Collect fault-free data to establish a normal mode library, calculate the initial baseline range of each feature, update the baseline range according to the load interval, collect historical fault data to establish a fault mode library, extract the typical range of fault features through cluster analysis, calculate the deviation of features from the baseline in real time, trigger fault mode library retrieval, and mark and update potential new faults.

[0011] Based on distributed monitoring, the system locates temperature anomalies, expands the anomaly area and records the duration, calculates the reconstruction error at each scale, determines the scale of the anomaly origin, and triggers graded early warnings based on the anomaly propagation range, energy growth factor, and duration.

[0012] Furthermore, the temperature data, current data, and cabinet door status data of the three-phase busbars of the distribution cabinet are collected, including the following steps:

[0013] Distributed fiber optic sensing points are laid every 0.5 meters along the length of the three-phase busbars A / B / C of the distribution cabinet, with both ends of the fiber optic cable connected to the DTS host; temperature and humidity sensors are installed at the top, middle, and bottom of the distribution cabinet; current transformers are installed at the three-phase input terminals to collect current signals; and magnetic proximity switches are installed at the cabinet door hinges to output cabinet door status signals.

[0014] Furthermore, the preprocessing includes the following steps:

[0015] The noise threshold is dynamically adjusted based on load current and ambient temperature: when the load current does not exceed 80% of the rated current and the ambient temperature does not exceed 35℃, the noise threshold is 3℃; when the load current exceeds 80% of the rated current, the noise threshold increases by 0.2℃ for every 10 amperes exceeding the rated current; when the ambient temperature exceeds 35℃, the noise threshold increases by 0.1℃ for every 1℃ exceeding the rated current.

[0016] For each temperature data point, data within 10 seconds before and after it are used to form an analysis window. The weight of each data point in the window is calculated using a Gaussian kernel function. The temperature change curve in the window is fitted using the weighted least squares method. If the deviation of any data point from the temperature change curve exceeds the noise threshold, it is identified as a noise point and replaced with the value of the temperature change curve at that time. The processed data is then smoothed twice using the moving average method, and the smoothed temperature sequence is output.

[0017] Furthermore, a fast Fourier transform is performed on the temperature data of each spatiotemporal unit to obtain the power spectrum, and the boundary between low and mid frequencies is initially established. And the boundary between mid-frequency and high-frequency frequencies. Calculate the Pearson correlation coefficient between power and load current in the mid-frequency band;

[0018] If the absolute value of the correlation coefficient is not less than 0.7, then the current boundary is retained. and ;

[0019] If the absolute value of the correlation coefficient is less than 0.7, the boundary is adjusted with a step size of 0.0005 Hz until the absolute value of the correlation coefficient is not less than 0.7.

[0020] Furthermore, based on and Construct empirical wavelet basis functions, including low-frequency basis functions, mid-frequency basis functions, and high-frequency basis functions;

[0021] The temperature data of the three phases A, B, and C are decomposed using empirical wavelet basis functions to obtain the low-frequency, mid-frequency, and high-frequency components of each phase. The instantaneous phase difference of the high-frequency components of the three phases is calculated. If the phase difference between any two phases exceeds 30°, the center frequency of the empirical wavelet basis function is adjusted and the data is decomposed again until all phase differences do not exceed 30°.

[0022] Furthermore, the db4 wavelet basis was selected to perform three-level wavelet decomposition on the low-frequency, mid-frequency and high-frequency components respectively. The low-frequency components were decomposed into daily, hourly and minute scales, the mid-frequency components were decomposed into 30-minute, 10-minute and 1-minute scales, and the high-frequency components were decomposed into 30-second, 10-second and 1-second scales.

[0023] Furthermore, fault-free data for several months after equipment commissioning were collected, and the 95% confidence intervals of each feature were calculated as the initial baseline range. For load-related features, the baselines were calculated separately for each load interval. The baseline range was updated using the incremental mean method with a 7-day window.

[0024] Furthermore, historical fault data is collected, features are extracted and standardized; Euclidean distance, local density and relative distance between samples are calculated; cluster centers are selected according to the product of local density and relative distance, and the samples are divided into M classes; typical range of features is extracted for each class of faults.

[0025] Furthermore, based on distributed monitoring, the physical coordinates of temperature anomaly points are located; if the spatial gradient of an anomaly point exceeds 5℃ / meter, the anomaly area is expanded to an adjacent 0.5-meter range; the duration of the anomaly is recorded; the reconstruction error at each scale is calculated to determine the scale of the anomaly origin; and graded early warnings are triggered based on the anomaly propagation range, energy growth factor, and duration.

[0026] If the anomaly occurs in a single phase and propagates to a 1-minute timescale, and the energy increases by 2 to 5 times within 15 minutes and continues for more than 15 minutes, a Level 1 warning will be triggered.

[0027] If the anomaly occurs in both phases and propagates to the 10-minute scale, and the energy increases by 5 to 10 times within 15 minutes and continues for more than 15 minutes, a level 2 warning will be triggered.

[0028] If the anomaly occurs in the three phases, it will propagate to the hourly scale, with energy increasing more than 10 times within 15 minutes, triggering a level three warning.

[0029] A system for analyzing and warning of temperature changes on the busbar of a distribution cabinet, comprising:

[0030] The data acquisition module collects temperature data on the three-phase busbar of the distribution cabinet, preprocesses the temperature data, outputs a smoothed temperature sequence, divides the temperature sequence into spatiotemporal units, and calculates the spatial and temporal characteristics of each spatiotemporal unit.

[0031] The frequency domain analysis module performs frequency domain analysis on the temperature data of the spatiotemporal unit, determines the frequency band boundary through fast Fourier transform, constructs low-frequency, mid-frequency and high-frequency empirical wavelet basis functions, decomposes the temperature data to obtain low-frequency, mid-frequency and high-frequency components, performs multi-level wavelet decomposition on each component, extracts sub-components at different time scales, and calculates temporal coupling characteristics, spatial coupling characteristics and physical coupling characteristics.

[0032] The pattern library management module collects fault-free data to build a normal pattern library, calculates the initial baseline range of each feature, updates the baseline range according to the load interval, collects historical fault data to build a fault pattern library, extracts typical ranges of fault features through cluster analysis, calculates the deviation between features and baseline in real time, triggers fault pattern library retrieval, and marks and updates potential new faults.

[0033] The anomaly early warning module locates temperature anomalies based on distributed monitoring, expands the anomaly area and records the duration, calculates the reconstruction error at each scale, determines the scale of the anomaly origin, and triggers graded early warnings based on the anomaly propagation range, energy growth factor and duration.

[0034] (III) Beneficial Effects

[0035] This invention provides a method and system for analyzing and providing early warning of temperature changes in the busbar of a power distribution cabinet, which has the following beneficial effects:

[0036] (1) By dynamically adjusting the noise threshold, false alarms and missed alarms caused by load fluctuations and changes in ambient temperature are effectively reduced. The temperature curve is fitted by weighted least squares method and smoothed by moving average method, which significantly improves the accuracy and stability of temperature data. The division of spatiotemporal units and the introduction of cabinet door status correction coefficient enhance the comprehensiveness and reliability of monitoring, providing a high-quality data foundation for subsequent analysis.

[0037] (2) By frequency domain analysis and adaptive empirical wavelet decomposition, the low-frequency trend, mid-frequency load fluctuation and high-frequency transient anomaly features of temperature data are accurately extracted. The sub-components of different time scales are refined by multi-layer wavelet decomposition to fully capture the anomaly transmission law, calculate the spatiotemporal coupling and physical coupling features, and deeply integrate multi-dimensional information such as time correlation, spatial distribution and current change, which significantly improves the accuracy of fault identification and early warning capability, and effectively distinguishes between normal fluctuations and potential faults.

[0038] (3) By dynamically establishing and updating the normal mode library, the baseline adaptability to load and environmental changes is significantly improved, misjudgment is reduced, cluster analysis is used to accurately extract the typical feature range of historical faults, a fault mode library is constructed, feature deviation is calculated in real time and fault library retrieval is triggered, known fault modes are quickly matched, potential new faults are marked and feedback is provided for closed-loop updates, and the coverage and accuracy of the fault library are continuously optimized, thereby greatly improving the accuracy of fault identification and the system's self-learning ability.

[0039] (4) By dynamically expanding the abnormal area, local missed detection is avoided, the duration of the abnormality is recorded and the characteristic change curve is drawn, providing a quantitative basis for the evolution of the fault, calculating the reconstruction error at each scale, accurately identifying the scale of the abnormal origin, quickly locating the root cause of the fault, and triggering graded early warning based on the abnormal propagation range, energy growth multiple and duration, so as to achieve accurate assessment of the severity of the fault and targeted response, significantly improving the timeliness of early warning and operation and maintenance efficiency. Attached Figure Description

[0040] Figure 1 This is a schematic diagram illustrating the steps of the method for analyzing and providing early warning of temperature changes on the busbar of the distribution cabinet according to the present invention.

[0041] Figure 2This is a schematic diagram of the structure of the analysis and early warning system for the temperature change of the distribution cabinet busbar of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Please see Figure 1 This invention provides a method for analyzing and providing early warning of temperature changes in the busbar of a power distribution cabinet, comprising the following steps:

[0044] Step 1: Collect temperature data, current data and cabinet door status data on the three-phase busbar of the distribution cabinet, preprocess the temperature data, output a smoothed temperature sequence, divide the temperature sequence into spatiotemporal units, and calculate the spatial and temporal characteristics of each spatiotemporal unit.

[0045] Step one includes the following:

[0046] Step 101: Lay one distributed optical fiber sensing point every 0.5m along the length of the three-phase busbars A / B / C of the distribution cabinet. Connect both ends of the optical fiber to the DTS host. Install one temperature and humidity sensor at the top, middle and bottom of the distribution cabinet. The top of the distribution cabinet is located 0.3m from the top of the cabinet, the middle is located at the vertical center of the cabinet, and the bottom is located 0.3m from the bottom of the cabinet. Install a current transformer at the three-phase input terminal to collect the current signal through the secondary side. Install a magnetic proximity switch at the cabinet door hinge to output a "1" (closed) or "0" (open) status signal.

[0047] Step 102: Set the rated current as the baseline (e.g., 630 amps), and the real-time load current and ambient temperature as variables. The calculation rules for the dynamic noise threshold are as follows:

[0048] When the load current does not exceed 80% of the rated current and the ambient temperature does not exceed 35℃, the noise threshold is 3℃ (the noise threshold is the noise threshold of the temperature data, in ℃).

[0049] When the load current exceeds 80% of the rated current, the noise threshold increases by 0.2°C for every 10 amperes exceeding the rated current.

[0050] When the ambient temperature exceeds 35℃, the noise threshold increases by 0.1℃ for every 1℃ increase.

[0051] Step 103: For each temperature data point, take all data within 10 seconds before and after it (a total of 21 data points) to form an analysis window. Calculate the weight of each data point within the window using a Gaussian kernel function (the closer to the current point, the higher the weight). Calculate, where, w Indicates weight, x The distance to the current point is represented by 0.8, which is the standard deviation of the Gaussian kernel, and exp() is the exponential function. The temperature change curve within the window is fitted using the weighted least squares method. If the deviation of any data point from the temperature change curve exceeds the dynamic noise threshold, it is identified as a noise point and replaced with the value of the temperature change curve at that data point. The processed data is then smoothed twice using the moving average method to eliminate high-frequency jitter caused by the inherent noise of the sensor, and the smoothed temperature sequence is output.

[0052] Step 104: Divide the temperature sequence into spatiotemporal units according to the standard of "10 minutes × 0.5 meters". Each unit contains temperature data of 120 time points (collected once every 5 seconds within 10 minutes) × 1 spatial point (monitoring point every 0.5 meters). Calculate the spatial and temporal characteristics for each unit. When calculating the spatial characteristics, take the temperature difference between two adjacent 0.5-meter data points and divide it by 0.5 meters to obtain the temperature gradient. When calculating the temporal characteristics, divide the difference between the maximum and minimum temperature values ​​in the unit by 600 seconds to obtain the rate of change. If the cabinet door status corresponding to the unit is "open" (status signal is 0), multiply the ambient temperature data of the unit by a weighting coefficient of 1.2 to compensate for the accelerated heat dissipation caused by the cabinet door being open, and mark "cabinet door open" in the unit label.

[0053] When using this method, refer to steps 101 to 104:

[0054] By dynamically adjusting the noise threshold, false alarms and missed alarms caused by load fluctuations and changes in ambient temperature are effectively reduced. The temperature curve is fitted using the weighted least squares method and then smoothed using the moving average method, which significantly improves the accuracy and stability of the temperature data. The division of the time and space into units and the introduction of a cabinet door status correction coefficient enhance the comprehensiveness and reliability of the monitoring, providing a high-quality data foundation for subsequent analysis.

[0055] Step 2: Perform frequency domain analysis on the temperature data of the spatiotemporal unit, determine the frequency band boundary through fast Fourier transform, construct low-frequency, mid-frequency and high-frequency empirical wavelet basis functions, decompose the temperature data to obtain low-frequency components, mid-frequency components and high-frequency components, perform multi-level wavelet decomposition on each component, extract sub-components at different time scales, and calculate temporal coupling characteristics, spatial coupling characteristics and physical coupling characteristics.

[0056] Step two includes the following:

[0057] Step 201: Perform a Fast Fourier Transform on the temperature data of each spatiotemporal unit to obtain the power spectrum in the frequency range of 0~0.1 Hz, and initially define the boundary between low frequency and mid frequency from the power spectrum. And the boundary between mid-frequency and high-frequency frequencies. For example, 0.001 Hz and 0.01 Hz, calculate the mid-frequency band (boundary). and The Pearson correlation coefficient between power and load current (between two values):

[0058] If the absolute value of the correlation coefficient is not less than 0.7, then the current boundary is retained. and ;

[0059] If the absolute value of the correlation coefficient is less than 0.7, the boundary is adjusted in steps of 0.0005 Hz (the boundary between low and mid frequencies is limited to 0.0005~0.002 Hz, and the boundary between mid and high frequencies is limited to 0.008~0.02 Hz) until the absolute value of the correlation coefficient is not less than 0.7.

[0060] Step 202: Based on and Three empirical wavelet basis functions were constructed: a low-frequency basis function, a mid-frequency basis function, and a high-frequency basis function. The low-frequency basis function covers the frequency range from 0 to the low-frequency / mid-frequency boundary and is used to extract long-term temperature trends. The mid-frequency basis function covers the frequency range from the mid-frequency / low-frequency boundary to the mid-frequency / high-frequency boundary and is used to extract temperature changes related to load fluctuations. The high-frequency basis function covers the frequency range from the mid-frequency / high-frequency boundary to 0.1 Hz and is used to extract rapid fluctuations related to contact conditions, such as whether the contact is tight or whether there are local sparks.

[0061] Step 203: Decompose the temperature data of the three phases A / B / C using the empirical wavelet basis function to obtain the low-frequency, mid-frequency, and high-frequency components of each phase. Calculate the instantaneous phase difference of the high-frequency components of the three phases. For example, the phase difference between phase A and phase B is the absolute value of their phase angles. If the phase difference between any two phases exceeds 30°, adjust the center frequency of the empirical wavelet basis function (adjust by 0.0001 Hz each time) and decompose again until all phase differences do not exceed 30°.

[0062] Step 204: Select the db4 wavelet basis to perform three-level wavelet decomposition on the low-frequency, mid-frequency and high-frequency components respectively. The low-frequency components are decomposed into daily, hourly and minute scales, the mid-frequency components are decomposed into 30-minute, 10-minute and 1-minute scales, and the high-frequency components are decomposed into 30-second, 10-second and 1-second scales.

[0063] Step 205: Within the spatiotemporal unit, record the peak occurrence time of the high-frequency component on a 10-second scale and the peak occurrence time of the mid-frequency component on a 1-minute scale. Calculate the time difference (the absolute value of the difference between the two time points) as a time coupling feature. For any monitoring point of phase A and the symmetrical point of phase B (at the same distance from the top of the cabinet), calculate the temperature difference of the low-frequency components of the two. Then divide the temperature difference by 0.5 meters (the phase-to-phase distance is approximately 0.5 meters) to obtain the spatial gradient as a spatial coupling feature. Within the spatiotemporal unit, calculate the maximum fluctuation amplitude of the mid-frequency component and the current change during the corresponding time period. The ratio of the two is the coupling coefficient, which serves as a physical coupling feature.

[0064] When using this method, refer to steps 201 to 205:

[0065] By using frequency domain analysis and adaptive empirical wavelet decomposition, the low-frequency trend, mid-frequency load fluctuation and high-frequency transient anomaly features of temperature data are accurately extracted. Combined with multi-layer wavelet decomposition to refine sub-components at different time scales, the abnormal transmission law is fully captured, the spatiotemporal coupling and physical coupling features are calculated, and multi-dimensional information such as time correlation, spatial distribution and current change are deeply integrated, which significantly improves the accuracy of fault identification and early warning capability, and effectively distinguishes between normal fluctuations and potential faults.

[0066] Step 3: Collect fault-free data to establish a normal mode library, calculate the initial baseline range of each feature, update the baseline range according to the load interval, collect historical fault data to establish a fault mode library, extract the typical range of fault features through cluster analysis, calculate the deviation between features and baseline in real time, trigger fault mode library retrieval, and mark and update potential new faults.

[0067] Step three includes the following:

[0068] Step 301: Collect fault-free data for several months after the equipment is put into operation. During this period, there are no warning records and the operation and maintenance have confirmed that the equipment is normal. Process the data according to the methods in Step 1 to Step 2 to obtain a three-dimensional dataset of "component-scale-feature". For each feature, calculate its 95% confidence interval as the initial baseline range. For example, the normal range of time-coupled features is 0~5 seconds.

[0069] Step 302: For load-related features (such as time difference, spatial gradient, coupling coefficient), calculate the baseline for each load range (0~30%, 30%~60%, 60%~100% rated current). For example, when the load is 30%~60%, the normal range of the coupling coefficient is 0.2~0.5℃ / Ampere. With a 7-day window, add 24 hours of normal data each day and update the baseline range using the incremental average (attenuation coefficient of 0.9). For example, when updating the baseline with the newly added 24 hours of normal data, the historical baseline and the average of the new data are weighted by the "attenuation coefficient of 0.9". The new baseline = historical baseline × 0.9 + average of the new data × 0.1. If the average of a certain feature increases from 2℃ to 2.1℃, the upper limit of the 95% range is adjusted upward accordingly.

[0070] Step 303: Collect historical fault records, such as poor contact, insulation aging, and short circuit. Each fault must contain complete data from "1 hour before occurrence → when it occurs → after handling". Extract features using the methods in Steps 1 and 2. Take the cross-scale coupling features as input and standardize them to the [0, 1] interval. Calculate the Euclidean distance between samples. Pre-set the cutoff distance (e.g., 0.3). Calculate the local density and relative distance. The relative distance is the minimum distance to samples with higher density. For each sample, calculate the Euclidean distance between the sample and all other samples. Divide each distance by the cutoff distance, take the square, and then take the negative of the squared value as the exponent. Calculate the natural exponential function value. Finally, add the natural exponential function values ​​corresponding to all samples. The sum is the local density of the sample.

[0071] Step 304: Sort by the product of local density and relative distance, take the top M peaks as cluster centers, divide the samples into M classes, and extract the typical range of its characteristics for each type of fault. For example, poor contact: time difference exceeds 10 seconds, spatial gradient exceeds 5℃ / m, and high frequency component fluctuation frequency on the 10-second scale is 5~10 times / minute; insulation aging: daily offset of low frequency component exceeds 2℃ (compared to last week), and coupling coefficient increases by 0.1℃ / Ampere per month; short circuit precursor: high frequency component jump on the 1-second scale exceeds 5℃, spatial gradient exceeds 10℃ / m, and duration exceeds 30 seconds.

[0072] Step 305: The normal mode library calculates the deviation of the current feature from the baseline every 10 minutes. The deviation is the Euclidean distance divided by the baseline standard deviation. For example, for the feature "time difference", assuming its normal data fluctuates within the range of 0-5 seconds, by statistically analyzing the dispersion of normal data, the square root of the average of the squares of the deviations of each data point from the mean is obtained as the baseline standard deviation of the time difference. When the deviation exceeds the deviation threshold, a fault mode library search is triggered.

[0073] The fault database uses the K-nearest neighbor algorithm, which takes the K nearest neighbor samples and calculates the similarity between the current feature and the samples in the database. The similarity is obtained by subtracting the ratio of the distance between samples to the maximum distance from 1. If the similarity is less than 50%, it is marked as a "potential new fault" and pushed to the operation and maintenance APP through the server. It includes feature curves and location information. The operation and maintenance personnel need to provide feedback of "confirmed fault" or "false alarm" within 48 hours. After confirmation, the new fault feature is automatically added to the fault database and the cluster center is updated. False alarms will expand the baseline range of the normal pattern database by 10%.

[0074] When using this method, refer to steps 301 to 305:

[0075] By dynamically establishing and updating the normal mode library, the baseline's adaptability to load and environmental changes is significantly improved, reducing misjudgments. Cluster analysis is used to accurately extract the typical feature range of historical faults, construct a fault mode library, calculate feature deviation in real time and trigger fault library retrieval, quickly match known fault modes, mark potential new faults and provide feedback for closed-loop updates, continuously optimize the coverage and accuracy of the fault library, thereby greatly improving the accuracy of fault identification and the system's self-learning ability.

[0076] Step 4: Based on distributed monitoring, locate temperature anomaly points, expand the anomaly area and record the duration, calculate the reconstruction error at each scale, determine the scale of the anomaly origin, and trigger graded early warnings based on the anomaly propagation range, energy growth factor and duration.

[0077] Step four includes the following:

[0078] Step 401: Based on distributed monitoring using fiber optic sensors, locate the physical coordinates of temperature anomaly points, such as phase A being 2.3 meters from the top of the cabinet. If the spatial gradient of the anomaly point exceeds 5℃ / meter, expand the anomaly area to an adjacent 0.5-meter range, record the moment when the anomaly first exceeds the baseline, calculate the duration, and plot the curve of the characteristic value changing over time, such as the curve of the coupling coefficient increasing from 0.4℃ / Ampere to 0.8℃ / Ampere.

[0079] Step 402: Calculate the reconstruction error (difference between the original signal and the reconstructed signal) of each decomposition scale. The scale with the highest error percentage is the scale from which the anomaly originates. For example, if the error percentage of the 10-second scale is 65%, the anomaly originates from high-frequency rapid fluctuations. If the anomaly originates from the 1-second scale, it is necessary to check whether the 10-second scale and the 1-minute scale have anomalies that occur synchronously, i.e., exceed the baseline, and the time difference between the occurrence of anomalies at each scale does not exceed 3 minutes.

[0080] The original signal refers to the "smoothed temperature sequence" output after secondary smoothing processing, which is the original monitoring data of the bus temperature in the space-time unit.

[0081] The reconstructed signal refers to the temperature signal synthesized in reverse from the decomposition results of each scale after multi-scale decomposition: for the three-level decomposition results of low-frequency, mid-frequency and high-frequency components, according to the reconstruction rules of wavelet transform, the detail components and approximate components of each level are superimposed to synthesize the complete signals of the corresponding components respectively. Then, the reconstructed signals of low-frequency, mid-frequency and high-frequency are added together to obtain a reconstructed signal with the same length as the original signal.

[0082] Energy is defined as the sum of the amplitude of each feature multiplied by its duration. For example, if the amplitude of a high-frequency component on a 10-second scale is 0.5℃ and it lasts for 2 minutes, then the energy is 0.5℃ × 120 seconds = 60℃·second.

[0083] Step 403: If an anomaly occurs in a single phase, the anomaly propagates from the fine scale to the 1-minute scale, the energy increases by 2 to 5 times within 15 minutes, and the duration exceeds 15 minutes, triggering a Level 1 warning, which notifies the maintenance team.

[0084] If an anomaly occurs in two phases, such as phase A and phase B, and the anomaly propagates to a 10-minute timescale, with energy increasing 5 to 10 times within 15 minutes and lasting for more than 15 minutes, a level 2 warning is triggered, and the level 2 warning notifies the operations and maintenance manager.

[0085] If all three phases are abnormal, the abnormality will propagate to the hourly scale. If the energy increases more than 10 times within 15 minutes, a Level 3 warning will be triggered. The Level 3 warning will notify the operation and maintenance manager and the emergency repair team.

[0086] When using this method, please refer to the content of steps 401 to 403:

[0087] By dynamically expanding the abnormal area, avoiding local missed detections, recording the duration of the abnormality and plotting characteristic change curves, a quantitative basis for fault evolution is provided. The reconstruction error at each scale is calculated, the scale of the abnormality origin is accurately identified, the root cause of the fault is quickly located, and graded early warnings are triggered based on the abnormality propagation range, energy growth factor and duration, so as to achieve accurate assessment of the severity of the fault and targeted response, significantly improving the timeliness of early warning and operation and maintenance efficiency.

[0088] Please see Figure 2 This invention provides an analysis and early warning system for temperature changes in the busbar of a power distribution cabinet, comprising: a data acquisition module, a frequency domain analysis module, a pattern library management module, and an anomaly early warning module, wherein:

[0089] The data acquisition module collects temperature data on the three-phase busbar of the distribution cabinet, preprocesses the temperature data, outputs a smoothed temperature sequence, divides the temperature sequence into spatiotemporal units, and calculates the spatial and temporal characteristics of each spatiotemporal unit.

[0090] The frequency domain analysis module performs frequency domain analysis on the temperature data of the spatiotemporal unit, determines the frequency band boundary through fast Fourier transform, constructs low-frequency, mid-frequency and high-frequency empirical wavelet basis functions, decomposes the temperature data to obtain low-frequency, mid-frequency and high-frequency components, performs multi-level wavelet decomposition on each component, extracts sub-components at different time scales, and calculates temporal coupling characteristics, spatial coupling characteristics and physical coupling characteristics.

[0091] The pattern library management module collects fault-free data to build a normal pattern library, calculates the initial baseline range of each feature, updates the baseline range according to the load interval, collects historical fault data to build a fault pattern library, extracts typical ranges of fault features through cluster analysis, calculates the deviation between features and baseline in real time, triggers fault pattern library retrieval, and marks and updates potential new faults.

[0092] The anomaly early warning module locates temperature anomalies based on distributed monitoring, expands the anomaly area and records the duration, calculates the reconstruction error at each scale, determines the scale of the anomaly origin, and triggers graded early warnings based on the anomaly propagation range, energy growth factor and duration.

[0093] In the application, the various formulas mentioned are all calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The coefficients in the formulas are set by those skilled in the art according to the actual situation.

[0094] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, and combinations thereof. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0095] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0096] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for analyzing and providing early warning of temperature changes on the busbar of a distribution cabinet, characterized in that: include: Temperature data is collected from the three-phase busbar of the distribution cabinet. The temperature data is preprocessed and a smoothed temperature sequence is output. The temperature sequence is divided into spatiotemporal units, and the spatial and temporal characteristics of each spatiotemporal unit are calculated. Frequency domain analysis is performed on the temperature data of the spatiotemporal unit. The frequency band boundary is determined by fast Fourier transform. Low-frequency, mid-frequency and high-frequency empirical wavelet basis functions are constructed. The temperature data is decomposed to obtain low-frequency, mid-frequency and high-frequency components. Multi-level wavelet decomposition is performed on each component to extract sub-components at different time scales. Temporal coupling characteristics, spatial coupling characteristics and physical coupling characteristics are calculated. Collect fault-free data to establish a normal mode library, calculate the initial baseline range of each feature, update the baseline range according to the load interval, collect historical fault data to establish a fault mode library, extract the typical range of fault features through cluster analysis, calculate the deviation of features from the baseline in real time, trigger fault mode library retrieval, and mark and update potential new faults. Based on distributed monitoring, the system locates temperature anomalies, expands the anomaly area and records the duration, calculates the reconstruction error at each scale, determines the scale of the anomaly origin, and triggers graded early warnings based on the anomaly propagation range, energy growth factor, and duration.

2. The method for analyzing and early warning of temperature changes in a distribution cabinet busbar according to claim 1, characterized in that: Distributed fiber optic sensing points are laid every 0.5 meters along the length of the three-phase busbars A / B / C of the distribution cabinet, with both ends of the fiber optic cable connected to the DTS host; temperature and humidity sensors are installed at the top, middle, and bottom of the distribution cabinet; current transformers are installed at the three-phase input terminals to collect current signals; and magnetic proximity switches are installed at the cabinet door hinges to output cabinet door status signals.

3. The method for analyzing and early warning of temperature changes in a distribution cabinet busbar according to claim 2, characterized in that: Preprocessing includes the following steps: The noise threshold is dynamically adjusted based on load current and ambient temperature: when the load current does not exceed 80% of the rated current and the ambient temperature does not exceed 35℃, the noise threshold is 3℃; when the load current exceeds 80% of the rated current, the noise threshold increases by 0.2℃ for every 10 amperes exceeding the rated current; when the ambient temperature exceeds 35℃, the noise threshold increases by 0.1℃ for every 1℃ exceeding the rated current. For each temperature data point, data within 10 seconds before and after it are used to form an analysis window. The weight of each data point in the window is calculated using a Gaussian kernel function. The temperature change curve in the window is fitted using the weighted least squares method. If the deviation of any data point from the temperature change curve exceeds the noise threshold, it is identified as a noise point and replaced with the value of the temperature change curve at that time. The processed data is then smoothed twice using the moving average method, and the smoothed temperature sequence is output.

4. The method for analyzing and early warning of temperature changes in the busbar of a distribution cabinet according to claim 1, characterized in that: Perform a Fast Fourier Transform on the temperature data of each spatiotemporal unit to obtain the power spectrum, and initially define the boundary between low and mid frequencies. And the boundary between mid-frequency and high-frequency frequencies. Calculate the Pearson correlation coefficient between power and load current in the mid-frequency band; If the absolute value of the correlation coefficient is not less than 0.7, then the current boundary is retained. and ; If the absolute value of the correlation coefficient is less than 0.7, the boundary is adjusted with a step size of 0.0005 Hz until the absolute value of the correlation coefficient is not less than 0.

7.

5. The method for analyzing and early warning of temperature changes in the busbar of a distribution cabinet according to claim 4, characterized in that: based on and Construct empirical wavelet basis functions, including low-frequency basis functions, mid-frequency basis functions, and high-frequency basis functions; The temperature data of the three phases A, B, and C are decomposed using empirical wavelet basis functions to obtain the low-frequency, mid-frequency, and high-frequency components of each phase. The instantaneous phase difference of the high-frequency components of the three phases is calculated. If the phase difference between any two phases exceeds 30°, the center frequency of the empirical wavelet basis function is adjusted and the data is decomposed again until all phase differences do not exceed 30°.

6. The method for analyzing and early warning of temperature changes in a distribution cabinet busbar according to claim 5, characterized in that: The db4 wavelet basis was selected to perform three-level wavelet decomposition on the low-frequency, mid-frequency and high-frequency components respectively. The low-frequency components were decomposed into daily, hourly and minute scales, the mid-frequency components were decomposed into 30-minute, 10-minute and 1-minute scales, and the high-frequency components were decomposed into 30-second, 10-second and 1-second scales.

7. The method for analyzing and early warning of temperature changes in a distribution cabinet busbar according to claim 1, characterized in that: Collect fault-free data for several months after equipment commissioning, and calculate the 95% confidence interval for each feature as the initial baseline range; for load-related features, calculate the baseline separately according to the load interval; update the baseline range using the incremental mean method with a 7-day window.

8. The method for analyzing and early warning of temperature changes in a distribution cabinet busbar according to claim 7, characterized in that: Collect historical fault data, extract features and standardize them; calculate Euclidean distance, local density and relative distance between samples; sort the samples by the product of local density and relative distance and select cluster centers to divide the samples into M classes; extract the typical range of features for each class of faults.

9. The method for analyzing and early warning of temperature changes in a distribution cabinet busbar according to claim 8, characterized in that: The physical coordinates of temperature anomaly points are located based on distributed monitoring; if the spatial gradient of the anomaly point exceeds 5℃ / meter, the anomaly area is expanded to an adjacent 0.5-meter range; the duration of the anomaly is recorded. Calculate the reconstruction error at each scale to determine the scale at which the anomaly originated; A tiered early warning system is triggered based on the abnormal conduction range, energy increase factor, and duration: If the anomaly occurs in a single phase and propagates to a 1-minute timescale, and the energy increases by 2 to 5 times within 15 minutes and continues for more than 15 minutes, a Level 1 warning will be triggered. If the anomaly occurs in both phases and propagates to the 10-minute scale, and the energy increases by 5 to 10 times within 15 minutes and continues for more than 15 minutes, a level 2 warning will be triggered. If the anomaly occurs in the three phases, it will propagate to the hourly scale, with the energy increasing more than 10 times within 15 minutes, triggering a level three warning.

10. A system for analyzing and warning of temperature changes in a distribution cabinet busbar, used to implement the method described in any one of claims 1 to 9, characterized in that: include: The data acquisition module collects temperature data on the three-phase busbar of the distribution cabinet, preprocesses the temperature data, outputs a smoothed temperature sequence, divides the temperature sequence into spatiotemporal units, and calculates the spatial and temporal characteristics of each spatiotemporal unit. The frequency domain analysis module performs frequency domain analysis on the temperature data of the spatiotemporal unit, determines the frequency band boundary through fast Fourier transform, constructs low-frequency, mid-frequency and high-frequency empirical wavelet basis functions, decomposes the temperature data to obtain low-frequency, mid-frequency and high-frequency components, performs multi-level wavelet decomposition on each component, extracts sub-components at different time scales, and calculates temporal coupling characteristics, spatial coupling characteristics and physical coupling characteristics. The pattern library management module collects fault-free data to build a normal pattern library, calculates the initial baseline range of each feature, updates the baseline range according to the load interval, collects historical fault data to build a fault pattern library, extracts typical ranges of fault features through cluster analysis, calculates the deviation between features and baseline in real time, triggers fault pattern library retrieval, and marks and updates potential new faults. The anomaly early warning module locates temperature anomalies based on distributed monitoring, expands the anomaly area and records the duration, calculates the reconstruction error at each scale, determines the scale of the anomaly origin, and triggers graded early warnings based on the anomaly propagation range, energy growth factor and duration.

Images