Distribution box operation state monitoring method and system

Abstract

The invention discloses a distribution box operation state monitoring method and system, and relates to the field of distribution boxes, and the system comprises an acquisition module which is used for collecting current and voltage signals of each loop in a distribution box, surface temperature distribution of a box body and opening and closing state signals of a cabinet door, and converting the collected analog signals into a digital signal sequence; the extraction module is used for receiving the digital signal sequence and extracting a signal fluctuation feature, a temperature field distribution feature and a state signal change feature to generate a three-dimensional feature set; the system can obtain loop electrical parameters, box body temperature and cabinet door state in real time, accurately analyze operation characteristics to deduce state trend, quickly position defective loop components and influence range in case of abnormity, scientifically evaluate health level, timely give an early warning, synchronously disconnect power supply of associated equipment in case of severity, effectively reduce fault shutdown, and improve work efficiency. The power utilization safety and the maintenance efficiency are improved, and stable operation of the distribution box is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of distribution box, in particular to a distribution box operation state monitoring method and system. BACKGROUND

[0002] The distribution box is the core equipment of the power system, which is composed of switch devices and protection appliances, and is responsible for power distribution and circuit on-off control. Overload, short circuit and leakage protection are realized through circuit breakers and leakage protectors to ensure power safety and reliability. The current distribution box is transforming towards intelligence and modularization, integrating Internet of Things and AI technologies, and can remotely monitor the operation state, fault diagnosis and predictive maintenance to adapt to smart grids and new energy demands.

[0003] The application number 202411555332.6 discloses a distribution box operation state monitoring method, which aims to solve the problems of "manual detection of running load relying on personnel professional accomplishment, insufficient reliability, affecting operation safety; running environment analysis not comprehensively considering distribution safety and operation efficiency, not comprehensive enough to accurately judge the influence of environment on operation state; and abnormality warning after the occurrence of abnormality, difficult to maintain in time, abnormality warning using fixed threshold, not considering device self-loss, easy to misjudge, leading to unsatisfactory monitoring effect".

[0004] For daily operation and maintenance management and monitoring of the distribution box, the existing technology has few comprehensive safety monitoring and deduction functions. Therefore, a new distribution box operation state monitoring method and system are proposed to realize daily operation and maintenance management and monitoring of the distribution box. SUMMARY

[0005] In view of the above shortcomings of the prior art, the present application provides a distribution box operation state monitoring method and system, which can effectively solve the problems of the prior art.

[0006] To achieve the above purpose, the present application realizes the following technical solutions: The present application discloses a distribution box operation state monitoring system, which comprises: The acquisition module is used for acquiring current and voltage signals of each loop in the distribution box, temperature distribution of the box surface and cabinet door opening and closing state signals, and converting the acquired analog signals into digital signal sequences; the extraction module is used for receiving the digital signal sequences, extracting signal fluctuation characteristics, temperature field distribution characteristics and state signal change characteristics to generate a three-dimensional feature set; the deduction module is used for receiving the three-dimensional feature set, analyzing and deducing an evolution trend of the distribution box operation state based on the correlation between the characteristics, and outputting a trend deduction result; the positioning module is used for obtaining the trend deduction result, positioning the loop, component and associated influence range corresponding to the abnormality when the result display has an abnormal tendency, and generating a positioning message; the evaluation module is used for obtaining the trend deduction result and the positioning message, combining the three-dimensional feature set, and evaluating a health level of the current operation state of the distribution box; the monitoring module is used for receiving the current health level evaluation result of the distribution box, continuously monitoring the dynamic change of the health level, and outputting a warning signal containing a warning level and associated abnormal information when the level is lower than a preset threshold or the change rate is abnormal. The acquisition module is connected with the extraction module through a wireless network, the extraction module is connected with the deduction module and the positioning module through a wireless network, the deduction module and the positioning module are connected with the evaluation module through a wireless network, and the evaluation module is connected with the monitoring module through a wireless network.

[0007] Further, the acquisition module triggers each loop current signal and voltage signal for the same frequency sampling through a synchronous clock in the running stage, so that the sampling time deviation is always controlled within a preset time range. The acquisition module is integrated with a temperature acquisition array, the acquisition points of the temperature acquisition array are arranged on the heat accumulation area of the box surface and the outer wall area corresponding to each loop core element, the coordinate information of each acquisition point and the model and installation position information of the corresponding loop element are pre-stored in the acquisition module, the acquisition module is integrated with a non-contact Hall sensor, and the non-contact Hall sensor generates a cabinet door opening and closing state signal by detecting the magnetic flux change between the cabinet door and the box.

[0008] Further, the extraction module is subjected to the following when performing feature extraction: The current signal and the voltage signal in the digital signal sequence are respectively subjected to short-time Fourier transform, the frequency domain peak value, the characteristic frequency band energy proportion and the phase offset of adjacent sampling points in the transform result are extracted to constitute a current and voltage fluctuation characteristic subset; The digital signal of the box surface temperature distribution is subjected to spatial gradient calculation to obtain the gradient amplitude and gradient direction characteristics of the temperature field, and a density clustering algorithm is used to cluster the temperature values of all temperature acquisition points to extract the temperature value of the cluster center and the intra-class temperature variance to constitute a temperature field distribution characteristic subset; The time sequence of the cabinet door opening and closing state signal is labeled, and the opening and closing frequency in a unit of time, the duration of a single opening and closing, and the interval period between adjacent two openings and closings are extracted to constitute a state signal change feature subset; Finally, the three feature subsets are input into an attention mechanism network, the network learns the matching degree of each feature vector and the abnormal features in the historical fault samples of the distribution box, generates a dynamic attention weight for each feature, and obeys the generation logic that the higher the matching degree, the greater the weight, and performs weighted summation on the feature vectors according to the weight to output the three-dimensional feature set.

[0009] Further, the process of deducing the operation state evolution trend based on the correlation analysis between features is: Calculate the mutual information between any two different types of features in the three-dimensional feature set: ; In the formula: H (X) is the information entropy of the first feature sequence X; H (Y) is the information entropy of the second feature sequence Y; H (X, Y) is the joint information entropy of X and Y; Calculate the gray correlation degree between each feature sequence and the standard operation state feature sequence of the distribution box, wherein the standard operation state feature sequence of the distribution box is generated by factory debugging and long-term normal operation data statistics of the distribution box: ; In the formula: is the gray correlation degree; is the sample number of the feature sequence; is the i-th sample value of the k-th feature sequence to be analyzed; is the i-th sample value of the standard feature sequence; is the resolution coefficient, ∈ (0, 1); represents the minimum value of the sample difference between all feature sequences and the standard sequence; represents the maximum value of the sample difference between all feature sequences and the standard sequence; Again, the , are normalized to the interval [0, 1], respectively, and are denoted as , and the weight is configured to calculate the comprehensive correlation degree between features: wherein are positive numbers and the sum is 1; Based on , a feature correlation network is constructed, wherein the nodes of the network are single features in the three-dimensional feature set, and the weight of the edge between the nodes is the , then the first-order difference of the feature time sequence of each node is calculated to obtain the feature change rate, the change synchronization between the associated features is analyzed combined with the edge weight, and then the trend slope reflecting the state evolution speed is quantified by comparing the change trend of the associated features with the feature change trend under the standard running state. The trend slope is positive, indicating evolution towards the normal direction, negative, indicating evolution towards the abnormal direction, and the larger the absolute value, the faster the evolution. At the same time, the proportion of the number of edges with an edge weight lower than a preset correlation threshold in the network is calculated, and the proportion is taken as the abnormal risk probability. Finally, the trend deduction result including the trend slope and the abnormal risk probability is output.

[0010] Further, when the positioning module positions the loop, component and associated influence range corresponding to the anomaly, it is subject to: Logic-step 1: Determine whether there is an abnormal tendency according to the abnormal risk probability and the trend slope in the trend deduction result; if the determination result is that there is an abnormal tendency, determine the feature type corresponding to the abnormal tendency based on the feature associated with the abnormal risk probability and the trend slope most strongly in the trend deduction process; Logic-step 2: Construct a distribution box fault propagation matrix M, the matrix dimension is loop component number x loop component number, and the element represents the probability of the i-th loop component fault propagating to the j-th loop component, The value of is mapped by the feature comprehensive correlation degree of the i-th and j-th components, subject to the higher the correlation degree, the larger the value; Logic-step 3: Set a comprehensive correlation threshold, and mark the loop components corresponding to the features related to the feature type corresponding to the abnormal tendency and not less than the comprehensive correlation threshold as candidate abnormal sources; Logic-step 4: For each candidate abnormal source, calculate its fault influence degree within a preset time through the fault propagation matrix. The fault influence degree is the product of the sum of the propagation probabilities within the preset time and the number of affected components. The candidate abnormal source with the highest influence degree and a unique starting point of the fault propagation path is determined as the abnormal core component; Logic-step 5: Determine the associated influence range of the anomaly according to the fault propagation path, and generate a positioning message containing the core component model, installation location identifier, associated loop number and influence range boundary.

[0011] Further, the logic for the evaluation module to evaluate the health level of the current running state of the distribution box is: For each feature in the three-dimensional feature set , , wherein n represents the total number of features, and the abnormal coefficient is calculated. In the abnormal coefficient calculation formula taking its actual sampling value, representing the characteristic of normal operation threshold, =0 represents no exception, the greater the abnormality is more serious; According to determine the correlation weight of each , representing the average correlation degree with all other characteristics; Finally, the health index of the current distribution box is calculated ; Wherein, set health threshold H1, H2, H3, and subject to 1>H1>H2>H3>0, then: H∈[H1,1] when, health level is excellent, that is, no abnormal risk; H∈[H2,H1) when, health level is good, that is, slight abnormality, no need to stop; H∈[H3,H2) when, health level is medium, that is, moderate abnormality, need to be inspected; H∈[0,H3) when, health level is poor, that is, serious abnormality, need to stop maintenance.

[0012] Further, when the health level of the current running state of the distribution box output by the evaluation module is poor, all power equipment connected with the distribution box is synchronously disconnected with the power connection of the distribution box.

[0013] Further, when the monitoring module continuously monitors the dynamic change of the health level, it is subject to: The health index H corresponding to the health level is continuously sampled by using a sliding time window, the window length is a preset time T, the sampling number in the window is N, and the change rate of the health index in each sliding window is calculated , H represents the health index at the end of the window, H represents the health index at the beginning of the window, the positive and negative of v are used to represent the rise and fall of the health index, and the absolute value is used to represent the change speed, and the exponential smoothing method is used to predict the health index in the future preset time period to obtain the predicted health index; The health warning threshold and the rate threshold are set synchronously, and when any of the following conditions is met, an early warning signal is output: The current H is less than the health warning threshold; the absolute value of v is greater than the rate threshold, which is only applicable to the case that v is less than zero; the predicted health index is less than the health warning threshold; The early warning signal includes early warning level and associated abnormal information.

[0014] ​In another aspect, a power distribution box operation state monitoring method comprises: Collecting current and voltage signals of each loop in the power distribution box, temperature distribution on the surface of the box body and cabinet door opening and closing state signals, converting the analog signals into digital signal sequences; receiving the digital signal sequences, extracting fluctuation characteristics such as current and voltage frequency domain peak values, distribution characteristics such as temperature field gradients and state change characteristics such as cabinet door opening and closing frequencies, generating a three-dimensional feature set through attention mechanism weighting; calculating mutual information between the three-dimensional features and the gray correlation degree between the features and standard operation characteristics, obtaining a comprehensive correlation degree after normalization, then constructing a feature correlation network to deduce the operation state evolution trend, and outputting a result containing a trend slope and an abnormal risk probability; if the result shows an abnormal tendency, determining the corresponding feature type and constructing a fault propagation matrix, screening candidate abnormal sources and calculating fault influence degrees, locating abnormal core components and associated impact ranges, and synchronously generating a positioning message; calculating feature abnormality coefficients and correlation weights, obtaining a health index and dividing a health level in combination with the three-dimensional feature set and other information, and disconnecting power equipment connected to the power distribution box when the health level is poor; applying a sliding time window to monitor dynamic changes in the health index, calculating a change rate and predicting a future health index, and outputting an early warning signal containing an early warning level and associated abnormal information when a threshold is triggered.

[0015] Compared with the known prior art, the technical solution provided by the present application has the following beneficial effects: The power distribution box operation state monitoring method and system can accurately collect current and voltage signals of each loop in the power distribution box, temperature distribution on the surface of the box body and cabinet door opening and closing state signals and convert them into digital signal sequences, extract signal fluctuations, temperature field distribution and state change characteristics, determine exclusive feature bands for different loop elements, optimize the accuracy of the feature set in combination with the attention mechanism, more reliably deduce the operation state evolution trend based on the correlation between the features, accurately locate the abnormal corresponding loop, core components and associated impact range when there is an abnormal tendency, reasonably evaluate the current operation health level to determine the maintenance requirements, continuously monitor the dynamic changes in the health level and predict the future trend, output a signal containing the early warning level and abnormal information when the early warning conditions are met, and simultaneously disconnect the associated power equipment connection when the health level is poor, effectively improving the monitoring comprehensiveness and accuracy, timely avoiding fault risks and ensuring the long-term safe and stable operation of the power distribution box. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creating any inventive labor.

[0017] Figure 1 It is a structural schematic view of a power distribution box operation state monitoring system. Figure 2 It is a flow schematic view of a power distribution box operation state monitoring method. DETAILED DESCRIPTION

[0018] To make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0019] The present application will be further described below with reference to the embodiments.

[0020] Embodiment 1 A power distribution box operation state monitoring system of the present embodiment, as shown in Figure 1 includes: A collection module, configured to collect current and voltage signals of each circuit in the power distribution box, temperature distribution on the surface of the box body and cabinet door opening and closing state signals, and convert the collected analog signals into digital signal sequences; The collection module triggers through a synchronous clock in the running stage, and performs same-frequency sampling on the current and voltage signals of each circuit, so that the sampling time deviation is always controlled within a preset time range; The collection module is integrated with a temperature collection array, and the collection points of the temperature collection array are arranged at heat accumulation areas on the surface of the box body and the outer wall areas corresponding to each circuit core element. The coordinate information of each collection point and the model and installation position information of the corresponding circuit element are pre-stored in the collection module. The collection module is integrated with a non-contact Hall sensor, which generates a cabinet door opening and closing state signal by detecting the magnetic flux change between the cabinet door and the box body; The heat accumulation areas on the surface of the box body at least include the outer wall corresponding to the circuit terminal and the area where the heat dissipation hole is opened, and the core elements at least include the circuit breaker and the contactor; An extraction module, configured to receive the digital signal sequences, extract signal fluctuation features, temperature field distribution features and state signal change features, and generate a three-dimensional feature set; When performing feature extraction, the extraction module is subject to: The current and voltage signals in the digital signal sequences are respectively subjected to short-time Fourier transform, the frequency domain peak value, characteristic frequency band energy proportion and phase offset of adjacent sampling points in the transform result are extracted, so as to constitute a current and voltage fluctuation feature subset; The characteristic frequency band is determined by the following method: collecting the electrical parameter signals of the distribution box during the factory debugging stage and within a preset time length of continuous normal operation, performing Fourier transform on the signals to obtain the electrical parameter frequency domain distribution data during normal operation; counting the frequency domain intervals with an energy ratio exceeding a preset confidence level in the frequency domain distribution data, and taking the intervals as the "basic characteristic frequency band"; excluding typical interference frequency bands (such as electromagnetic interference frequency bands of the environment where the distribution box is located, and harmonic interference frequency bands of the power grid power frequency signal) contained in the basic characteristic frequency band, and the remaining frequency domain intervals are the "characteristic frequency band" for feature extraction; and for different types of circuit elements (such as circuit breakers, contactors, and relays), the above steps are repeated to generate exclusive characteristic frequency bands matching the element types. The spatial gradient of the digital signal of the surface temperature distribution of the box is calculated to obtain the gradient amplitude and gradient direction characteristics of the temperature field, and the density clustering algorithm is used to cluster the temperature values of all temperature collection points to extract the temperature values of the cluster centers and the temperature variance within the cluster to constitute a temperature field distribution feature subset, wherein the clustering parameters are preset through box thermal field uniformity analysis; The timing sequence of the cabinet door opening and closing state signal is labeled, and the opening and closing frequency per unit time, the duration of a single opening and closing, and the interval period between adjacent two openings and closings are extracted to constitute a state signal change feature subset; Finally, the three feature subsets are input into the attention mechanism network, and the network learns the matching degree of each feature vector and the abnormal features in the historical fault samples of the distribution box to generate dynamic attention weights for each feature, which obeys the generation logic that the higher the matching degree, the greater the weight, and the feature vectors are weighted and summed according to the weights to output a three-dimensional feature set; The inference module is configured to receive the three-dimensional feature set, analyze and infer the evolution trend of the running state of the distribution box based on the correlation between the features, and output the trend inference result. The process of the inference module for analyzing and inferring the evolution trend of the running state based on the correlation between the features is as follows: Calculate the mutual information between any two different types of features in the three-dimensional feature set: ; In the formula: H(X) is the information entropy of the first feature sequence X; H(Y) is the information entropy of the second feature sequence Y; H(X,Y) is the joint information entropy of X and Y; The formula above constructs a quantitative correlation relationship among the information entropy of the first feature sequence, the information entropy of the second feature sequence and the joint information entropy of the two, which distinguishes the limitations of traditional single feature analysis through such operation logic, provides basic data support for subsequent feature correlation comprehensive evaluation, applies the mutual information concept in information theory to the distribution box operation feature correlation analysis, realizes the quantitative expression of the correlation strength between features, and avoids the subjectivity and fuzziness of qualitative analysis; wherein, , respectively represent the total number of samples in the first feature sequence X, the total number of samples in the second feature sequence Y, represents the occurrence probability of the i th sample in X, represents the joint probability of the i th sample in X and the j th sample in Y; The grey correlation degrees between each feature sequence and the distribution box standard operation state feature sequence are calculated, wherein the distribution box standard operation state feature sequence is generated by distribution box factory debugging and long-term normal operation data statistics: ; In the formula: is the grey correlation degree; is the sample number of the feature sequence; is the i th sample value of the k th feature sequence to be analyzed; is the i th sample value of the standard feature sequence; is the resolution coefficient, ∈(0,1), The value of is determined according to the distribution box operation stability requirement, and is smaller when higher sensitivity is required to distinguish the subtle differences between the feature sequence and the standard operation state feature sequence, and is larger when lower sensitivity is allowed; represents the minimum value of the sample difference between all feature sequences and the standard sequence; represents the maximum value of the sample difference between all feature sequences and the standard sequence; The formula above takes the sample number of the feature sequence as the basis, adjusts the analysis sensitivity by introducing the resolution coefficient, combines the minimum and maximum values of the sample difference between all feature sequences and the standard sequence, constructs a similarity quantitative model of the feature sequence to be analyzed and the standard operation state feature sequence, and the resolution coefficient can be flexibly adjusted according to the distribution box operation stability requirement, and is smaller when high sensitivity is required to distinguish subtle differences, and is larger when low sensitivity is allowed, which effectively solves the problem that traditional correlation analysis cannot adapt to different stability requirement scenarios, and can more accurately reflect the deviation degree of the feature to be analyzed and the standard feature, providing support for judging whether the feature is abnormal or not; Again, normalize , respectively, so that both are normalized to the interval [0,1], denoted as and configure weights to calculate the comprehensive correlation degree between features: wherein are positive numbers and the sum is 1, when the correlation between different types of features is more critical to the influence on the judgment of the running state evolution trend of the distribution box, The greater the value is; on the contrary, The smaller the value is, when the similarity between the feature to be analyzed and the standard running state feature of the distribution box is more critical to the influence on the judgment of the running state evolution trend, The greater the value is; on the contrary, The smaller the value is; The above formula comprehensively considers the internal correlation between features and the degree of fit between features and standard states by complementing the advantages of the two correlation indicators, laying a foundation for constructing a feature correlation network and accurately deducing the running state evolution trend; Based on construct a feature correlation network, wherein the nodes of the network are single features in the three-dimensional feature set, and the weight of the edge between the nodes is the , then calculate the first-order difference of the feature time sequence of each node to obtain the feature change rate, analyze the change synchronization between the correlated features combined with the edge weight, and then compare the change trend of the current correlated features with the feature change trend under the standard running state to quantitatively obtain the trend slope reflecting the speed of state evolution. The slope is positive, indicating evolution in the normal direction, negative indicating evolution in the abnormal direction, and the greater the absolute value is, the faster the evolution is. At the same time, the proportion of the number of edges in the network whose weight is lower than the preset correlation degree threshold is calculated, and the proportion is taken as the abnormal risk probability. Finally, the trend deduction result containing the trend slope and the abnormal risk probability is output; A positioning module is configured to obtain the trend deduction result, and when the result shows an abnormal tendency, the module locates the circuit, component and associated impact range corresponding to the abnormality and generates a positioning message; When the positioning module locates the circuit, component and associated impact range corresponding to the abnormality, it is subject to: Logic-step1: Determine whether there is an abnormal tendency according to the abnormal risk probability and the trend slope in the trend deduction result. If the determination result is that there is an abnormal tendency, determine the feature type corresponding to the abnormal tendency based on the feature with the strongest correlation with the abnormal risk probability and the trend slope in the trend deduction process; Logic-step2: Construct a distribution box fault propagation matrix M, which has a dimension of circuit component number x circuit component number, and the element represents the probability of the i-th circuit component fault propagating to the j-th circuit component, The value of is generated by mapping the feature comprehensive correlation degree between the i-th and j-th components, subject to the higher the correlation degree, Logic-step 3: Set a comprehensive correlation threshold, which will be related to the feature type corresponding to the abnormal tendency and Loop components corresponding to features that are not less than the comprehensive correlation threshold are marked as candidate anomaly sources; Logic-step4: For each candidate anomaly source, calculate its fault impact within a preset time using the fault propagation matrix. The fault impact is the product of the sum of propagation probabilities within the preset time and the number of affected components. The candidate anomaly source with the highest impact and the unique starting point of the fault propagation path is identified as the core component of the anomaly. Logic-step 5: Determine the scope of the anomaly's associated impact based on the fault propagation path (e.g., the circuit containing the core component and its propagation probability). For loops with a value of ≥0.5, a positioning message is generated that includes the core component model, installation location identifier, corresponding loop number, and boundary of the affected area. The fault propagation path starts from the faulty core component and the propagation probability is selected based on the fault propagation matrix. The inter-component relationships are determined to be no less than a preset propagation probability threshold; The assessment module is used to obtain trend projection results and location messages, and combined with the three-dimensional feature set, assess the health level of the current operating status of the distribution box; The evaluation module assesses the health level of the distribution box's current operating status using the following logic: For each feature in the three-dimensional feature set , Represents the total number of features, and calculates their anomaly coefficient. In the formula for calculating the anomaly coefficient Take its actual sampled value, Representation of features Normal operating threshold, =0 indicates no abnormality. The larger the value, the more severe the abnormality. according to Determine each Association weight , express The mean of the overall correlation with all other features; Finally, calculate the current health index of the distribution box. ; Where health thresholds H1, H2, and H3 are set, and they follow the order 1 > H1 > H2 > H3 > 0, then: When H∈[H1,1], the health level is excellent, which means there is no abnormal risk. When H∈[H2,H1), the health level is good, which means there is a slight abnormality and no need to stop the machine. When H∈[H3, H2), the health level is medium, which means moderate abnormality, and needs to be inspected; When H∈[0, H3), the health level is poor, which means serious abnormality, and needs to be shut down for maintenance; When the health level of the current running state of the distribution box output by the evaluation module is poor, all power equipment connected with the distribution box is synchronously disconnected with the power connection of the distribution box; The monitoring module receives the evaluation result of the current health level of the distribution box, continuously monitors the dynamic change of the health level, and outputs a warning signal containing a warning level and associated abnormal information when the level is lower than a preset threshold or the change rate is abnormal; When the monitoring module continuously monitors the dynamic change of the health level, it is subject to: The health index H corresponding to the health level is continuously sampled by using a sliding time window, the window length is a preset time T, the sampling number in the window is N, and the change rate of the health index in each sliding window is calculated , represents the health index at the end of the window, represents the health index at the beginning of the window, the positive and negative of v represent the increase and decrease of the health index, and the absolute value represents the change speed, and the exponential smoothing method is used to predict the health index in the future preset time period to obtain the predicted health index; Wherein, the smoothing coefficient is preset by the historical change stability of the health index, and the smoothing coefficient is preset according to the standard deviation of the historical health index sequence, that is, the smaller the ratio of the standard deviation to the coefficient of variation, the larger the smoothing coefficient, and vice versa. The smoothing coefficient is smaller; The health warning threshold and the rate threshold are set synchronously, and when any of the following conditions is met, a warning signal is output: The current H is less than the health warning threshold; the absolute value of v is greater than the rate threshold, which is only applicable to the case where v is less than zero; the predicted health index is less than the health warning threshold; The warning signal contains a warning level and associated abnormal information; Wherein, the warning level is set according to the deviation degree of H from the health warning threshold and the deviation degree of the absolute value of v from the rate threshold to divide into first-level warning, second-level warning and third-level warning, and the associated abnormal information includes the specific feature category in the three-dimensional feature set that causes the abnormality of the health level or the warning and the abnormal core component positioning information; The acquisition module is connected with the extraction module through wireless network interaction, the extraction module is connected with the deduction module and the positioning module through wireless network interaction, the deduction module and the positioning module are connected with the evaluation module through wireless network interaction, and the evaluation module is connected with the monitoring module through wireless network interaction.

[0021] In the embodiment, the acquisition module collects the current and voltage signals of each circuit in the distribution box, the temperature distribution of the box surface, and the opening and closing state signal of the cabinet door, converts the collected analog signals into digital signal sequences, the extraction module receives the digital signal sequences, extracts the signal fluctuation characteristics, temperature field distribution characteristics, and state signal change characteristics, generates a three-dimensional feature set, the deduction module receives the three-dimensional feature set, analyzes and deduces the evolution trend of the running state of the distribution box based on the correlation between the characteristics, outputs the trend deduction result, the positioning module further obtains the trend deduction result, and when the result display shows an abnormal tendency, the positioning module locates the abnormal circuit, component, and associated impact range, generates a positioning message, and the evaluation module obtains the trend deduction result and the positioning message, combines the three-dimensional feature set, evaluates the health level of the current running state of the distribution box, and finally the monitoring module receives the current health level evaluation result of the distribution box, continuously monitors the dynamic change of the health level, and when the level is lower than the preset threshold or the change rate is abnormal, outputs a warning signal containing the warning level and associated abnormal information.

[0022] In the above embodiment, the system can obtain the circuit electrical parameters, box temperature, and cabinet door state in real time in the distribution box running scene, accurately analyze the running characteristics to deduce the state trend, quickly locate the problem circuit components and impact range when abnormal, scientifically evaluate the health level, timely warn, synchronously disconnect the power supply of associated equipment when serious, effectively reduce the fault downtime, improve the power safety and maintenance efficiency, and ensure the stable operation of the distribution box.

[0023] Embodiment 2: In the specific implementation level, based on embodiment 1, the embodiment refers to Figure 2 The embodiment 1 is further specifically described as a distribution box running state monitoring system: A distribution box running state monitoring method, comprising: Collecting the current and voltage signals of each circuit in the distribution box, the temperature distribution of the box surface, and the opening and closing state signal of the cabinet door, converting the analog signals into digital signal sequences; Receiving the digital signal sequences, extracting the fluctuation characteristics such as current and voltage frequency domain peak values, the distribution characteristics such as temperature field gradient, and the state change characteristics such as cabinet door opening and closing frequency, generating a three-dimensional feature set through attention mechanism weighting; Calculating the mutual information between the three-dimensional characteristics and the gray correlation degree between the characteristics and the standard running characteristics, obtaining the comprehensive correlation degree after normalization, and constructing a feature correlation network to deduce the evolution trend of the running state, outputting a result containing the trend slope and abnormal risk probability; If the result shows an abnormal tendency, determine the corresponding feature type and construct a fault propagation matrix, screen the candidate abnormal sources and calculate the fault impact degree, locate the abnormal core components and associated impact range, and synchronously generate a positioning message; The health index is obtained by combining the three-dimensional feature set and other information, and the health level is divided, and when the health level is poor, the power equipment connected to the distribution box is disconnected; The sliding time window is applied to monitor the dynamic change of the health index, the change rate is calculated, and the future health index is predicted, and when the threshold is triggered, the early warning signal containing the early warning level and the associated abnormal information is output.

[0024] In summary, in the above-mentioned embodiments, the system and method can accurately collect the current and voltage signals of each circuit in the distribution box, the temperature distribution on the surface of the box, and the cabinet door opening and closing state signal and convert them into digital signal sequences, extract signal fluctuation, temperature field distribution and state change characteristics, determine the exclusive feature band for different circuit elements, optimize the accuracy of the feature set combined with the attention mechanism, analyze and deduce the evolution trend of the running state based on the correlation between features, accurately locate the abnormal circuit, core components and associated impact range when there is an abnormal tendency, reasonably evaluate the current running health level to determine the maintenance requirements, continuously monitor the dynamic change of the health level and predict the future trend, output the signal containing the early warning level and abnormal information when the early warning condition is met, disconnect the associated power equipment connection when the health level is poor, effectively improve the monitoring comprehensiveness and accuracy, avoid fault risks in time, and ensure the long-term safe and stable operation of the distribution box.

[0025] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A power distribution box operation status monitoring system, characterized in that, include: The acquisition module is used to acquire current and voltage signals of each circuit in the distribution box, temperature distribution on the surface of the box, and opening and closing status signals of the cabinet door, and convert the acquired analog signals into digital signal sequences. The extraction module is used to receive digital signal sequences and extract signal fluctuation features, temperature field distribution features, and state signal change features to generate a three-dimensional feature set. The deduction module is used to receive a three-dimensional feature set, analyze the correlation between features to deduce the evolution trend of the operating status of the distribution box, and output the trend deduction results. The positioning module is used to obtain trend projection results. When the results show an abnormal trend, it locates the circuit, component and related impact range corresponding to the abnormality and generates a positioning message. The assessment module is used to obtain trend projection results and location messages, and combined with the three-dimensional feature set, assess the health level of the current operating status of the distribution box; The monitoring module is used to receive the current health level assessment results of the distribution box, continuously monitor the dynamic changes in the health level, and output an early warning signal containing the warning level and related abnormal information when the level is lower than the preset threshold or the rate of change is abnormal.

2. The distribution box operation status monitoring system according to claim 1, characterized in that, During the operation phase, the acquisition module is triggered by a synchronous clock to sample the current and voltage signals of each circuit at the same frequency, so that the sampling time deviation is always controlled within a preset time range. The acquisition module integrates a temperature acquisition array. The acquisition points of the temperature acquisition array are arranged in the heat accumulation area on the surface of the cabinet and the corresponding outer wall area of ​​each core component of the circuit. The coordinate information of each acquisition point and the model and installation position information of the corresponding circuit component are pre-stored inside the acquisition module. The acquisition module integrates a non-contact Hall sensor, which generates a cabinet door opening and closing status signal by detecting the change in magnetic flux between the cabinet door and the cabinet.

3. The distribution box operation status monitoring system according to claim 1, characterized in that, When performing feature extraction, the extraction module follows the following rules: Short-time Fourier transforms are performed on the current and voltage signals in the digital signal sequence, and the frequency domain peak, characteristic frequency band energy ratio, and phase offset of adjacent sampling points are extracted from the transform results to form a feature subset of current and voltage fluctuations. Spatial gradient calculation is performed on the digital signal of temperature distribution on the surface of the box to obtain the gradient amplitude and gradient direction characteristics of the temperature field. At the same time, density clustering algorithm is used to cluster the temperature values ​​of all temperature acquisition points, and the temperature values ​​of the cluster centers and the temperature variance within the cluster are extracted to form a subset of temperature field distribution characteristics. The time sequence of cabinet door opening and closing status signals is labeled, and the opening and closing frequency, duration of a single opening and closing, and interval period between two adjacent opening and closing are extracted to form a subset of status signal change features. Finally, the three feature subsets are input into the attention mechanism network. The network learns the matching degree between each feature vector and the abnormal features in the historical fault samples of the distribution box, and generates dynamic attention weights for each feature. The generation logic is that the higher the matching degree, the greater the weight. The feature vectors are then weighted and summed according to the weights to output the three-dimensional feature set.

4. The distribution box operation status monitoring system according to claim 1, characterized in that, The process of the deduction module, which analyzes the correlation between features to deduce the evolution trend of the operating state, is as follows: Calculate the mutual information between any two features of different types in a 3D feature set: ； In the formula: The information entropy of the first feature sequence X; The information entropy of the second feature sequence Y; Let X be the joint information entropy of Y. Calculate the grey relational degree between each feature sequence and the feature sequence of the standard operating state of the distribution box, where the feature sequence of the standard operating state of the distribution box is generated by statistical analysis of the distribution box's factory commissioning and long-term normal operation data: ； In the formula: Grey relational degree; The number of samples in the feature sequence; This represents the i-th sample value of the k-th feature sequence to be analyzed. This represents the i-th sample value of the standard feature sequence; The resolution coefficient, ∈ (0,1); This represents the minimum difference between all characteristic sequences and standard sequence samples; This represents the maximum value of the differences between all feature sequences and the standard sequence samples; Again , Normalize them separately to bring them to the interval [0,1], denoted as . And configure weights to calculate the overall correlation between features: ,in All are positive numbers and their sum is 1; based on Construct a feature association network, where each network node is a single feature in a 3D feature set, and the weights of edges between nodes are taken from the weights of the corresponding two features. Then, the first-order difference is calculated for the feature time sequence of each node to obtain the feature change rate. Combined with the edge weight analysis, the synchronicity of changes between related features is analyzed. Then, by comparing the change trend of the current related features with the feature change trend under the standard operating state, the trend slope reflecting the speed of state evolution is quantified. A positive slope indicates evolution in the normal direction, and a negative slope indicates evolution in the abnormal direction. The larger the absolute value, the faster the evolution. At the same time, the proportion of the number of edges with edge weights lower than the preset correlation degree threshold in the network is counted. This proportion is used as the probability of abnormal risk. Finally, the trend inference result containing the trend slope and the probability of abnormal risk is output.

5. The distribution box operation status monitoring system according to claim 1, characterized in that, When the positioning module locates the circuit, component, and associated impact range corresponding to the anomaly, it follows the following rules: Logic-step1: Determine whether there is an abnormal tendency based on the abnormal risk probability and trend slope in the trend extrapolation results; if the judgment result is that there is an abnormal tendency, then determine the feature type corresponding to the abnormal tendency based on the feature that is most strongly correlated with the abnormal risk probability and trend slope in the trend extrapolation process. Logic-step 2: Construct the fault propagation matrix M of the distribution box. The matrix dimension is the number of circuit components × the number of circuit components, and the elements are... This represents the probability that a failure in the i-th loop component will propagate to the j-th loop component. The value is determined by the comprehensive correlation degree of the features corresponding to the i-th and j-th components. Mapping generation follows a higher degree of correlation. The larger; Logic-step 3: Set a comprehensive correlation threshold, which will be related to the feature type corresponding to the abnormal tendency and Loop components corresponding to features that are not less than the comprehensive correlation threshold are marked as candidate anomaly sources; Logic-step4: For each candidate anomaly source, calculate its fault impact degree within a preset time using the fault propagation matrix. The fault impact degree is the product of the sum of propagation probabilities within the preset time and the number of affected components. The candidate anomaly source with the highest impact degree and the unique starting point of the fault propagation path is determined as the core component of the anomaly. Logic-step5: Determine the associated impact range of the anomaly based on the fault propagation path, and generate a location message containing the core component model, installation location identifier, corresponding circuit number, and impact range boundary.

6. The distribution box operation status monitoring system according to claim 1, characterized in that, The evaluation module assesses the health level of the current operating status of the distribution box using the following logic: For each feature in the three-dimensional feature set , Represents the total number of features, and calculates their anomaly coefficient. In the formula for calculating the anomaly coefficient Take its actual sampled value, Representation of features Normal operating threshold, =0 indicates no abnormality. The larger the value, the more severe the abnormality. according to Determine each Association weight , express The mean of the overall correlation with all other features; Finally, calculate the current health index of the distribution box. ; Where health thresholds H1, H2, and H3 are set, and they follow the order 1 > H1 > H2 > H3 > 0, then: When H∈[H1,1], the health level is excellent, which means there is no abnormal risk. When H∈[H2,H1), the health level is good, which means there is a slight abnormality and no need to stop the machine. When H∈[H3,H2), the health level is medium, which means moderate abnormality and requires routine inspection; When H∈[0,H3), the health level is poor, which means that there is a serious abnormality and the machine needs to be shut down for maintenance.

7. The distribution box operation status monitoring system according to claim 6, characterized in that, When the evaluation module outputs a poor health level for the current operating status of the distribution box, all electrical equipment connected to the distribution box will simultaneously disconnect from the power supply.

8. The distribution box operation status monitoring system according to claim 1, characterized in that, When the monitoring module continuously monitors the dynamic changes in health level, it follows the following rules: A sliding time window is used to continuously sample the health index H corresponding to the health level. The window length is a preset time T, and the number of samplings within the window is denoted as N. The rate of change of the health index within each sliding window is calculated. , The health index at the end of the window. The health index at the start of the window is represented by the positive or negative value of v, which indicates whether the health index is rising or falling. The absolute value indicates the rate of change. The exponential smoothing method is then used to predict the health index within a preset time period in the future to obtain the predicted health index. Synchronously set health warning thresholds and rate thresholds. When any of the following conditions are met, an warning signal will be output: The current H is less than the health warning threshold; the absolute value of v is greater than the rate threshold, which only applies when v is less than zero; the predicted health index is less than the health warning threshold. The warning signal includes the warning level and associated anomaly information.

9. The distribution box operation status monitoring system according to claim 1, characterized in that, The acquisition module is interactively connected to the extraction module via a wireless network. The extraction module is interactively connected to the inference module and the positioning module via a wireless network. The inference module and the positioning module are interactively connected to the evaluation module via a wireless network. The evaluation module is interactively connected to the monitoring module via a wireless network.

10. A method for monitoring the operating status of a distribution box, wherein the method is an implementation method of a distribution box operating status monitoring system as described in any one of claims 1-9, characterized in that, include: Collect current and voltage signals of each circuit in the distribution box, temperature distribution on the surface of the box, and opening and closing status signals of the cabinet door, and convert the analog signals into digital signal sequences; The system receives digital signal sequences, extracts fluctuation features such as peak values ​​in the frequency domain of current and voltage, distribution features such as temperature field gradient, and state change features such as cabinet door opening and closing frequency, and generates a three-dimensional feature set by weighting through an attention mechanism. Calculate the mutual information between three-dimensional features and the grey correlation degree between features and standard operating features. After normalization, obtain the comprehensive correlation degree. Then, construct a feature correlation network to infer the evolution trend of the operating status and output the results including trend slope and abnormal risk probability. If the results show an abnormal tendency, determine the corresponding feature type and construct a fault propagation matrix, screen candidate anomaly sources and calculate the fault impact, locate the core components of the anomaly and the associated impact range, and generate a location message simultaneously. Calculate the anomaly coefficient and correlation weight of each feature, combine information such as the three-dimensional feature set to obtain the health index and classify the health level. When the health level is poor, disconnect the power equipment connected to the distribution box. The system uses a sliding time window to monitor the dynamic changes in health indices, calculates the rate of change, predicts future health indices, and outputs an early warning signal containing the warning level and related abnormal information when a threshold is triggered.

Images