Line aging monitoring system applied to safe electric device

Abstract

The invention discloses a line aging monitoring system applied to a safe electric device, and the system comprises the steps: collecting time series data of line state parameters (such as multi-point temperature, key node current and voltage, and leakage current), and training a dynamic baseline model through employing historical data collected under a normal operation condition; the model can learn and predict an expected parameter value after considering the influence of factors such as load and environment in a normal operation state of the line; then, comparing a line state parameter time sequence acquired in real time with a parameter baseline value time sequence predicted by the dynamic baseline model, and calculating a residual vector sequence between the two time sequences; furthermore, the abnormal state of the line is accurately judged by analyzing the mode characteristics of the residual error sequence in a period of time, so that the problem that normal fluctuation and real aging abnormity are difficult to distinguish by a traditional method is effectively solved, and the accuracy and reliability of line aging monitoring are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring, and more specifically, to a line aging monitoring system for safety electrical devices. Background Art

[0002] As a vital component of the power system, the operational status of power lines is directly related to the stability and safety of the power supply. Safety electrical devices used in critical locations, in particular, place extremely high demands on power supply reliability and safety. However, over the long term, power lines are subject to multiple factors, including the thermal effects of current flow, changes in ambient temperature and humidity, chemical corrosion, external damage, and changes in the physical and chemical properties of the insulation materials themselves, inevitably leading to aging. Line aging typically manifests as decreased insulation performance, increased conductor resistance, and increased contact resistance of joints. These factors not only lead to increased power loss but, more seriously, can cause serious safety incidents such as overheating, short circuits, leakage, and even fire, posing a significant threat to human life and property. Therefore, developing a line aging monitoring solution for safety electrical devices that can timely and accurately monitor line aging and provide early warnings when aging is in its early stages or abnormal conditions occur is of vital importance for ensuring the safe and stable operation of power systems and preventing electrical fires.

[0003] Currently, there is a certain technical foundation for monitoring line status, such as by installing sensors to monitor parameters such as line temperature, current, and voltage, and setting fixed thresholds for alarms. However, these traditional methods often have significant flaws. Line operating parameters, such as temperature, current, and voltage, inherently change dynamically with normal operating conditions, such as load changes and ambient temperature fluctuations. Simple threshold alarm methods struggle to effectively distinguish these normal parameter fluctuations from the slow, continuous abnormal changes caused by line aging. Furthermore, a core challenge in line aging monitoring lies in accurately distinguishing normal operational fluctuations from true abnormal aging signals that represent potential risks. Relying solely on whether instantaneous values exceed certain limits is prone to numerous false alarms (misinterpreting normal fluctuations as abnormalities) or missed alarms (failing to identify aging anomalies in their infancy, where the numerical changes are not yet significant). This fails to meet the high-reliability monitoring requirements of safety electrical devices.

[0004] Therefore, a more optimized line aging monitoring solution for safety electrical devices is expected. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a line aging monitoring system for safety electrical devices, which collects time series data of line state parameters (such as multi-point temperature, key node current and voltage, leakage current) and uses historical data collected under normal operating conditions to train a dynamic baseline model. The model can learn and predict the expected parameter values under normal operating conditions of the line after considering the influence of factors such as load and environment; then, the line state parameter time series collected in real time is compared with the parameter baseline value time series predicted by the dynamic baseline model, and the residual vector sequence between the two is calculated; furthermore, by analyzing the pattern characteristics of the residual sequence over a period of time, an accurate judgment of the abnormal state of the line is achieved, thereby effectively solving the problem that traditional methods are difficult to distinguish between normal fluctuations and real aging anomalies, and significantly improving the accuracy and reliability of line aging monitoring.

[0006] According to one aspect of the present application, a line aging monitoring system for a safety electrical device is provided, comprising:

[0007] A data acquisition module, configured to acquire line state parameters through a sensor component to obtain a time series of the line state parameters;

[0008] A dynamic baseline construction module is used to input each line state parameter in the time series of the line state parameter into the trained dynamic baseline model to obtain a time series of predicted line state parameter baseline values;

[0009] A residual calculation module, configured to input the time series of the predicted line state parameter baseline value and the time series of the line state parameter into the residual calculation module to obtain a time series of the line state residual vector;

[0010] A residual sequence pattern feature extraction module is used to extract residual sequence pattern features from the time series of the line state residual vector to obtain a line state residual sequence pattern feature encoding vector;

[0011] The abnormality judgment module is used to obtain an abnormality judgment result based on the line state residual sequence pattern feature coding vector.

[0012] Compared with the prior art, the present application provides a line aging monitoring system for safety electrical devices. The system collects time series data of line status parameters (such as multi-point temperature, current and voltage at key nodes, and leakage current) and uses historical data collected under normal operating conditions to train a dynamic baseline model. The model can learn and predict the expected parameter values under normal operating conditions of the line, taking into account the influence of factors such as load and environment. Subsequently, the time series of line status parameters collected in real time is compared with the time series of parameter baseline values predicted by the dynamic baseline model, and the residual vector sequence between the two is calculated. Furthermore, by analyzing the pattern characteristics of the residual sequence over a period of time, an accurate judgment of the abnormal state of the line is achieved, thereby effectively solving the problem that traditional methods have difficulty in distinguishing normal fluctuations from real aging anomalies, and significantly improving the accuracy and reliability of line aging monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0014] Figure 1 is a block diagram of a line aging monitoring system applied to a safety electrical device according to an embodiment of the present application;

[0015] Figure 2 Schematic diagram of data flow in a line aging monitoring system for safety electrical devices according to an embodiment of the present application;

[0016] Figure 3 4 is a block diagram of a residual sequence pattern feature extraction module in a line aging monitoring system for a safety electrical device according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0018] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0019] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0020] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0021] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0022] In the technical solution of the present application, a line aging monitoring system for safety electrical devices is proposed. Figure 1 4 is a block diagram of a line aging monitoring system applied to a safety electrical device according to an embodiment of the present application. Figure 2 Schematic diagram of data flow in a line aging monitoring system for a safety electrical device according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to an embodiment of the present application, a line aging monitoring system 300 for a safety electrical device includes: a data acquisition module 310 for acquiring line state parameters through a sensor assembly to obtain a time series of line state parameters; a dynamic baseline construction module 320 for inputting each line state parameter in the time series of the line state parameters into a trained dynamic baseline model to obtain a time series of predicted line state parameter baseline values; a residual calculation module 330 for inputting the time series of predicted line state parameter baseline values and the time series of line state parameters into the residual calculation module to obtain a time series of line state residual vectors; a residual sequence pattern feature extraction module 340 for performing residual sequence pattern feature extraction on the time series of the line state residual vectors to obtain a line state residual sequence pattern feature coding vector; and an abnormality judgment module 350 for obtaining an abnormality judgment result based on the line state residual sequence pattern feature coding vector.

[0023] In particular, the data acquisition module 310 is configured to acquire line state parameters via a sensor assembly to obtain a time series of line state parameters. It should be understood that as electrical lines are affected by multiple factors during long-term operation, such as ambient temperature changes, load fluctuations, and gradual degradation of insulation materials, manual inspections or static threshold alarms alone are no longer able to promptly and accurately detect potential aging hazards. Therefore, in the technical solution of the present application, a sensor assembly is used to acquire line state parameters to obtain a time series of line state parameters. These line state parameters include temperature values at multiple temperature measurement points, current and voltage values at key nodes, and leakage current values. Line state parameters not only reflect the thermal conditions of the line (e.g., abnormal temperatures often indicate overload or insulation degradation), but also reveal changes in electrical characteristics (e.g., current and voltage fluctuations indicate load anomalies or poor contact), while leakage current is an important indicator for determining insulation aging and damage. Converting these key parameters into a continuous time series not only provides comprehensive and reliable data support for subsequent dynamic baseline modeling and residual analysis, but also greatly improves sensitivity and response speed to complex dynamic processes.

[0024] Specifically, the dynamic baseline construction module 320 is configured to input each line state parameter in the time series of line state parameters into a trained dynamic baseline model to obtain a time series of predicted line state parameter baseline values. The dynamic baseline model is an intelligent model trained based on historical data. This historical data is labeled as line parameters under normal operating conditions and reflects the typical fluctuation patterns and dynamic characteristics of various line indicators under normal conditions. It should be understood that in actual operation, electrical lines are affected by various factors such as ambient temperature fluctuations and load fluctuations. Line state parameters exhibit normal dynamic variations. Using static or fixed thresholds as a benchmark makes it difficult to accurately distinguish between normal fluctuations and true abnormalities. By using the dynamic baseline model to predict line state parameters collected in real time, the system can dynamically construct a time series of "baseline" reference values that reflect the current line health status. In this process, the dynamic baseline model captures and learns the temporal variation characteristics under normal conditions, allowing it to flexibly adjust the predicted baseline as the environment and operating conditions change, significantly improving the adaptability and accuracy of monitoring. In this way, a scientific and objective benchmark can be provided for subsequent residual calculations and anomaly determinations, ensuring that the difference between the actual observation value and the dynamic baseline can truly reflect the potential aging or abnormal hazards of the line, rather than misjudgment caused by normal fluctuations.

[0025] In particular, the residual calculation module 330 is configured to input the time series of predicted line state parameter baseline values and the time series of line state parameters into the residual calculation module to obtain a time series of line state residual vectors. It should be understood that although the dynamic baseline model can reflect the normal dynamic characteristics of the line, in actual operation, line state parameters may still deviate from the normal fluctuation range due to aging, equipment failure, or other abnormal factors. This deviation is specifically reflected in the residuals. Therefore, in the technical solution of the present application, the time series of predicted line state parameter baseline values and the time series of line state parameters are input into the residual calculation module to obtain a time series of line state residual vectors. The residual calculation module is essentially a data processing unit whose function is to subtract the predicted baseline from the actual value corresponding to each time point in the form of a time series, capturing the variation pattern of the deviation between the two. Specifically, this step compares the ideal normal state parameters predicted based on the dynamic baseline model with the real-time observation data at each moment, calculates the difference between the two, i.e., the residual, thereby forming a time series of residual vectors reflecting the deviation of line operation from the normal state. In this way, the system can keenly capture subtle anomalies in line performance, improve the sensitivity and accuracy of anomaly detection, reduce false alarms and missed alarms, and thus provide a solid data foundation and technical guarantee for the stable operation of safety electrical equipment.

[0026] In particular, the residual sequence pattern feature extraction module 340 is used to extract residual sequence pattern features from the time series of the line state residual vector to obtain a line state residual sequence pattern feature encoding vector. It should be understood that the residual vector time series generated during the operation of the power line not only contains random fluctuations such as equipment inherent noise and environmental interference, but also implies specific evolution patterns formed by real aging processes such as insulation degradation and increased contact resistance. These patterns often manifest as non-stationary changes in specific spatiotemporal correlations in the residual sequence, such as the persistent diffusion of local anomalies and the coordinated drift of multi-parameter residuals. Traditional methods based on thresholds or simple statistics are difficult to effectively distinguish between random noise and real aging signals, and are even more unable to capture the complex dynamic coupling relationship between residual vectors. Therefore, in order to construct a deep feature representation that can characterize the inherent dynamic structure of the residual sequence, in the technical solution of the present application, residual sequence pattern feature extraction is performed on the time series of the line state residual vector to obtain a line state residual sequence pattern feature encoding vector. Those skilled in the art will recognize that when line aging occurs, its residual sequence often exhibits specific spatiotemporal propagation characteristics: for example, an abnormal temperature rise at a temperature measurement point can affect the residual changes of adjacent nodes through conductor heat conduction, and residual fluctuations caused by current harmonic distortion may exhibit specific phase delays. By collaboratively calculating dynamic transmission potential energy and message repulsion, the network can automatically identify potential patterns in the residual sequence that conform to the laws of aging evolution, such as continuously increasing positive deviations and time-correlated fluctuation diffusion. In the process of extracting residual sequence pattern features from the time series of line state residual vectors, each residual vector is regarded as a dynamic entity, and its potential influence transmission direction and content potential with the current residual vector are encoded through dynamic transmission potential energy. At the same time, the propagation repulsion coefficient is used to dynamically adjust the information penetration intensity between different time points, so that the system can adaptively strengthen abnormal propagation paths related to aging and suppress instantaneous fluctuation interference caused by environmental noise. Compared with traditional methods that only focus on the residual anomaly at the current moment, this application constructs a dynamic potential field across time points, enabling the system to capture early signs of the aging process, thereby significantly improving the spatiotemporal sensitivity of aging identification.

[0027] In a specific example of this application, Figure 3As shown, the residual sequence pattern feature extraction module 340 includes: a node extraction unit 341, which is used to extract the current line state residual vector from the time series of the line state residual vector and define other line state residual vectors in the time series of the line state residual vector as line state residual vectors to be propagated to obtain the time series of the line state residual vectors to be propagated; a simulated dynamic liquid message propagation unit 342, which is used to perform line state simulated dynamic liquid message propagation on each line state residual vector to be propagated in the time series of the line state residual vectors to be propagated based on the line state liquid interaction field between each line state residual vector to be propagated in the time series of the line state residual vectors to be propagated and the current line state residual vector to obtain a set of line state message dynamic coding vectors; and a fusion unit 343, which is used to fuse the set of line state message dynamic coding vectors and the current line state residual vector to obtain a line state residual sequence pattern feature coding vector.

[0028] Specifically, the node extraction unit 341 is used to extract the current line state residual vector from the time series of the line state residual vector and define other line state residual vectors in the time series of the line state residual vector as the line state residual vector to be propagated to obtain the time series of the line state residual vector to be propagated. It should be understood that in the residual sequence generated by the dynamic baseline model, the real aging defect is often manifested as a specific propagation mode formed by multi-parameter residuals on the time axis, such as the phase delay effect between the persistent positive offset of the temperature residual and the periodic fluctuation of the leakage current residual. If the entire residual sequence is directly analyzed globally, it is easy to overwhelm the key features due to noise interference and normal operating condition fluctuations. Therefore, in the technical solution of the present application, the current line state residual vector is extracted from the time series of the line state residual vector and other line state residual vectors in the time series of the line state residual vector are defined as the line state residual vector to be propagated to obtain the time series of the line state residual vector to be propagated.

[0029] Specifically, by using the current moment's residual vector as the information convergence center and defining residuals at historical and future moments as nodes to be propagated, a dynamic computational field centered on the current moment is constructed. This allows the system to focus on the current moment's dynamic position in the time field and simulate the propagation of abnormal signals in a liquid medium. This spatiotemporal deconstruction enables subsequent potential energy and repulsive force calculations to accurately quantify the path and intensity of the impact of residuals at different moments on the current state.

[0030] In a specific example of the present application, the current line state residual vector is extracted from the time series of line state residual vectors using the following formula, and other line state residual vectors in the time series of line state residual vectors are defined as line state residual vectors to be propagated to obtain a time series of line state residual vectors to be propagated; wherein the formula is:

[0031] H={h1,h2,...,h T}∈R T×d

[0032] h c =h T

[0033] H p ={h1,h2,...,h T-1}

[0034] Where H is the time series of the line state residual vector, h1,h2,h T are the line state residual vectors of the 1st, 2nd and Tth time steps in the time series of the line state residual vector, T represents the total length of the sequence, d represents the feature dimension of each time step, and h c is the current line state residual vector, H p is the time series of the state residual vector of the line to be propagated, h1,h2,h T-1 are respectively the state residual vectors of the line to be propagated at the 1st, 2nd and T-1th time steps in the time series of the state residual vectors of the line to be propagated.

[0035] Specifically, the simulated dynamic liquid message propagation unit 342 is configured to perform simulated dynamic liquid message propagation on each of the line state residual vectors to be propagated in the time sequence of the line state residual vectors to be propagated, based on the line state liquid interaction field between each of the line state residual vectors to be propagated and the current line state residual vector, to obtain a set of line state message dynamic encoding vectors. In an embodiment of the present application, the line state dynamic transmission potential energy of each of the line state residual vectors to be propagated in the time sequence of the line state residual vectors to be propagated relative to the current line state residual vector is first calculated to obtain a time sequence of line state dynamic transmission potential energy encoding vectors. It should be understood that when line aging occurs, the residual signals it generates do not exist in isolation, but rather follow physical laws such as conductor heat conduction and electromagnetic field distribution to form a temporal correlation network. For example, oxidative corrosion of a terminal can cause a slow increase in contact resistance. This change manifests itself in the residual sequence as a coordinated drift of temperature residuals and current residuals, and this drift pattern exhibits a specific propagation acceleration over time. Traditional methods use statistics calculated within a fixed time window and are unable to capture the changes in temporal correlation strength governed by physical laws. Therefore, in the technical solution of this application, the line state dynamic transmission potential energy of each to-be-transmitted line state residual vector in the time series of to-be-transmitted line state residual vectors relative to the current line state residual vector is calculated to obtain a time series of line state dynamic transmission potential energy encoding vectors.

[0036] By encoding the interaction potential between the residual vector to be propagated and the current residual vector into a high-dimensional vector, the system can capture the interaction patterns in the residual evolution process that conform to the physical laws of aging. For example, in the thermal aging scenario of insulating materials, a positive offset in the temperature residual will enhance the leakage current residual at subsequent time points through the thermal accumulation effect. This causal relationship is reflected in the transmission potential energy encoding as the activation of a high-dimensional vector in a specific direction. This encoding mechanism enables the system to identify potential regular signals that traditional time-domain analysis views as random fluctuations, providing key features for warning of poor contact risks weeks in advance. At the same time, the constraints of physical laws effectively suppress misjudgments caused by transient interference.

[0037] In a specific example of the present application, the line state dynamic transmission potential energy of each to-be-propagated line state residual vector relative to the current line state residual vector in the time series of the to-be-propagated line state residual vector is calculated using the following formula to obtain a time series of line state dynamic transmission potential energy encoding vectors; wherein the formula is:

[0038]

[0039] Among them, W a and W k denote the learnable query weight matrix and key weight matrix respectively, α is the line state dynamic transmission potential energy factor, Indicates the calculation of h c and h i The square difference between i A potential energy encoding vector for the line state dynamic transmission.

[0040] Next, the line state message propagation repulsion coefficient of each line state residual vector to be propagated in the time series of the line state residual vector to be propagated relative to the current line state residual vector is calculated to obtain a time series of the line state message propagation repulsion coefficient. It should be understood that when the line operates in a dynamic load environment, the instantaneous residual abnormalities caused by equipment start-up and shutdown or grid fluctuations often have similar amplitude characteristics to early aging signals, but there are essential differences in the time propagation characteristics of the two - noise interference usually presents a random distribution, while the real aging residual will form a propagation trajectory with physical regularity. Therefore, in order to establish an adaptive time correlation strength adjustment mechanism and realize intelligent attenuation of non-correlated noise, in the technical solution of the present application, the line state message propagation repulsion coefficient of each line state residual vector to be propagated in the time series of the line state residual vector to be propagated relative to the current line state residual vector is calculated to obtain a time series of the line state message propagation repulsion coefficient.

[0041] That is, by introducing a repulsive force mechanism, the system is able to simulate the obstructive effect of impurity particles in liquid media on information flow, and map the concept of spatial repulsion of particle interactions in physics to the field of time series data analysis, thereby constructing an information filtering field that conforms to the propagation characteristics of line aging. In one example, when the residual vector at a certain point in time lacks physical relevance to the current state (such as the instantaneous voltage residual spike caused by lightning induction), the corresponding repulsive force coefficient will automatically increase, suppressing the interference of the noise signal on the current state judgment. For example, in the gradual process of oxidation aging of cable joints, the temperature residuals at adjacent time points show a continuous gradient due to the physical characteristics of thermal inertia. This type of residual propagation that conforms to the law of heat conduction will trigger a lower repulsive force coefficient, allowing it to participate in the feature aggregation of the current node. By calculating the repulsive force coefficient, the system can identify and filter these abnormal propagation paths that violate the physical laws of line aging, significantly improving the system's anti-interference ability in complex environments.

[0042] In a specific example of the present application, the line state message propagation repulsive force coefficient of each to-be-propagated line state residual vector relative to the current line state residual vector in the time series of the to-be-propagated line state residual vector is calculated as follows to obtain a time series of the line state message propagation repulsive force coefficient; wherein the formula is:

[0043]

[0044] in, represents vector multiplication, time(·) is the timestamp extraction function, τ represents the time difference scaling factor, represents the square of the norm, β i A repulsive force coefficient is propagated for the line status message.

[0045] Furthermore, based on the time series of the line state message propagation repulsion coefficient and the time series of the line state dynamic transmission potential energy encoding vector, the line state dynamic liquid message propagation is performed on each of the line state residual vectors to be propagated in the time series to obtain a set of line state message dynamic encoding vectors. It should be understood that when a line has hidden defects, the propagation of its residual signal is not a simple linear superposition, but rather a multipath coupling effect caused by physical laws such as thermal fatigue of the conductor material and breakdown of the insulating dielectric. For example, the leakage current residual caused by localized carbonization of the cable insulation layer will form asymmetric fluctuations with phase delays with the temperature residual at adjacent time points. This complex relationship involves both the potential energy accumulation caused by thermal inertia and the nonlinear modulation of the material aging rate. Traditional time series modeling methods have difficulty capturing this dynamic coupling relationship of multidimensional parameters. Therefore, in order to transform discrete residual observations into spatiotemporal characteristic expressions that contain the essence of the physical process, in the technical solution of the present application, based on the time series of the line state message propagation repulsion coefficient and the time series of the line state dynamic transmission potential energy coding vector, each line state residual vector to be propagated in the time series of the line state residual vector to be propagated is subjected to line state simulated dynamic liquid message propagation to obtain a set of line state message dynamic coding vectors.

[0046] Here, by combining the interactive potential of dynamic transmission potential energy encoding with the modulation effect of the repulsive force coefficient, the system is able to simulate the propagation attenuation characteristics of aging signals in real scenarios. For example, in a scenario where resistance increases due to oxidation of the contact terminal, the conduction of the temperature residual is guided by both the thermal conductivity of the material (potential energy direction) and the thickness of the oxide layer on the contact surface (repulsive force strength). This composite effect is encoded into a high-dimensional vector of a specific form during the liquid message propagation process. This encoding mechanism enables the system to break through the limitations of traditional feature engineering and automatically extract deep patterns that conform to the physical evolution laws of line aging, significantly improving the system's accuracy in analyzing complex aging patterns.

[0047] In a specific example of the present application, based on the time series of the line state message propagation repulsive force coefficient and the time series of the line state dynamic transmission potential energy encoding vector, the line state simulated dynamic liquid message propagation is performed on each to-be-propagated line state residual vector in the time series of the to-be-propagated line state residual vector using the following formula to obtain a set of line state message dynamic encoding vectors; wherein, the formula is:

[0048] e i =h i ⊙sigmoid{h i -ReLU(s i ⊙(β i ·h i W v ))}

[0049] Among them, ⊙ represents the position point multiplication, sigmoid(·) represents the sigmoid function, ReLU(·) represents the ReLU function, W v represents the learnable value weight matrix, e i A set of dynamic encoding vectors for the line status message.

[0050] Specifically, the fusion unit 343 is configured to fuse the set of line state message dynamic encoding vectors and the current line state residual vector to obtain a line state residual sequence pattern feature encoding vector. In an embodiment of the present application, the set of line state message dynamic encoding vectors is first subjected to multi-body propagation collaborative renormalization to obtain a set of optimized line state message dynamic encoding vectors. It should be understood that while the set of line state message dynamic encoding vectors generated by simulating dynamic liquid message propagation can capture local interaction characteristics in the time dimension, in practical scenarios with multi-parameter coupling and frequent load fluctuations, the dynamic correlations between time points often exhibit nonlinear superposition effects. For example, the leakage current residual and temperature residual of a certain line section may form an asymmetric collaborative drift pattern on the time axis. Traditional simple vector superposition or weighted fusion methods are unable to effectively characterize the information propagation patterns under such multi-body synergy and can easily dilute or obfuscate key aging characteristics. Therefore, in a preferred embodiment of the present application, the set of line state message dynamic encoding vectors is subjected to multi-body propagation collaborative renormalization to obtain a set of optimized line state message dynamic encoding vectors. That is, based on the theory of multi-body effects, the system regards the set of coding vectors as a dynamic propagation network, introducing a product state structure and a global scale correction mechanism.

[0051] Specifically, by calculating the mean of the eigenvalues to construct a global scaling factor and dynamically adjusting the propagation weights of each vector based on partial derivative operations, the system simulates the scale covariance of multi-body interactions in physical systems. This renormalization process not only retains the nonlinear adaptability of the original liquid message propagation, but also strengthens the long-range correlation between the residual dynamics at different time points through the product structure, so that the early weak but continuously diffusing aging characteristics can be systematically enhanced, while the instantaneous fluctuations caused by random noise are suppressed due to the lack of synergistic effects. This renormalization mechanism based on manifold geometry significantly improves the system's characterization accuracy of complex aging patterns. Specifically, when the line is in a multi-stage composite aging process (such as insulation carbonization and metal creep occurring simultaneously), its residual propagation pattern will form nested cooperative fluctuations on the time axis. Through the joint optimization of global scale representation correction and local propagation bonds, the system can capture this cross-time scale cooperative evolution law.

[0052] In this example, the set of line state message dynamic coding vectors is subjected to multi-body propagation cooperative renormalization using the following formula to obtain a set of optimized line state message dynamic coding vectors; wherein the formula is:

[0053]

[0054] e' i =μ T-1 e i

[0055]

[0056] Wherein, μ is the mean of all eigenvalues of the set of dynamic coding vectors of the line state message, ln(·) represents the logarithm operation with e as the base, ω represents the hyperparameter weight, and e' i A set of dynamic encoding vectors for the optimized line state message.

[0057] Furthermore, the set of dynamic coding vectors of the line status message and the current line state residual vector are fused and optimized to obtain the line state residual sequence pattern feature coding vector. It should be understood that although the set of dynamic coding vectors obtained by multi-body propagation cooperative renormalization can characterize the long-term propagation pattern of the residual sequence, under complex working conditions such as load mutation or instantaneous interference, relying solely on historical dynamic information may weaken the key abnormal signal at the current moment. For example, when the line causes instantaneous degradation of the insulation material due to local overheating, the current residual vector may contain mutation features, and these features may be partially diluted by the smoothing effect in the time dimension during the dynamic propagation process. In addition, the real aging process often manifests itself as a dynamic propagation pattern and a current anomaly. Therefore, in order to construct a composite feature space that has both temporal dynamic correlation and immediate state sensitivity, in the technical solution of the present application, the set of dynamic coding vectors of the line status message and the current line state residual vector are fused and optimized to obtain the line state residual sequence pattern feature coding vector.

[0058] Here, by nonlinearly fusing the optimized dynamic coding vector set (which contains the propagation law of residuals at multiple time points) with the current residual vector (reflecting the latest state deviation), the system achieves dual capture of the historical evolution trajectory of the aging process and the current instantaneous state. In one example, when the line shows an early increase in contact resistance, the dynamic coding vector set may capture the periodic enhancement pattern of the residual fluctuation, while the current residual vector reflects the instantaneous amplitude characteristics of the resistance mutation. The fusion of the two enables the feature coding vector to characterize the continuous accumulation trend of the abnormal signal and lock the degree of deviation of the current state. In this way, the physical interpretability of the feature coding and the clarity of the abnormality discrimination boundary are significantly improved.

[0059] In a specific example of the present application, the set of line state message dynamic coding vectors and the current line state residual vector are fused and optimized using the following formula to obtain a line state residual sequence pattern feature coding vector, wherein the formula is:

[0060] z′ i =z i +r i

[0061]

[0062] Among them, r i represents the linear compression factor, g represents the normalized weight factor, softmax(·) represents the softmax function, W g represents the gating weight matrix, z' i A vector encoding a pattern feature of the line state residual sequence.

[0063] In particular, the abnormality judgment module 350 is used to obtain an abnormality judgment result based on the line state residual sequence pattern feature coding vector. In the technical solution of the present application, the line state residual sequence pattern feature coding vector is input into a pre-trained machine learning classifier to obtain an abnormality judgment result, and the abnormality judgment result includes normal fluctuations or real aging abnormalities. It should be understood that in the actual operation process, the electrical circuit is affected by various factors such as environmental changes and load fluctuations, and its state performance is extremely complex. It is difficult to effectively distinguish normal fluctuations from real aging abnormalities by relying solely on traditional threshold methods or simple rules. As a data-driven intelligent discrimination model, the machine learning classifier can automatically learn and extract deep temporal dynamic features and association rules between different categories through training on a large amount of historical labeled data (including normal and abnormal samples).

[0064] During the implementation process, the line state residual sequence pattern feature encoding vector is first input into a trained classifier. Based on the discriminant function it has learned, the classifier performs high-dimensional spatial mapping and pattern matching on the line state residual sequence pattern feature encoding vector. Combining probabilistic statistical methods, the classifier then makes a scientific judgment on whether the current line state represents "normal fluctuation" or "true aging anomaly." By leveraging the powerful nonlinear modeling and generalization capabilities of machine learning algorithms, the system improves the accuracy and response speed of anomaly detection under complex dynamic conditions, effectively reducing the risk of false positives and missed alarms, and providing early warning of potential aging hazards or failure risks. By pre-training the classifier, it can adapt to changes in data distribution across different equipment, environments, and operating conditions, giving the entire monitoring system greater data parsing power and adaptability. This significantly enhances the electrical line health monitoring system's ability to identify aging risks and safety hazards under various complex operating conditions.

[0065] As described above, the line aging monitoring system 300 for safety electrical devices according to an embodiment of the present application can be implemented in various wireless terminals, such as a server equipped with a line aging monitoring algorithm for safety electrical devices. In one possible implementation, the line aging monitoring system 300 for safety electrical devices according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the line aging monitoring system 300 for safety electrical devices can be a software module within the operating system of the wireless terminal, or an application developed specifically for the wireless terminal. Of course, the line aging monitoring system 300 for safety electrical devices can also be one of the many hardware modules of the wireless terminal.

[0066] Alternatively, in another example, the line aging monitoring system 300 applied to safety electrical devices and the wireless terminal may also be separate devices, and the line aging monitoring system 300 applied to safety electrical devices may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0067] While various embodiments of the present disclosure have been described above, the above descriptions are illustrative, non-exhaustive, and not intended to be limiting of the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A line aging monitoring system for safety electrical devices, characterized in that: include: A data acquisition module, configured to acquire line state parameters through a sensor component to obtain a time series of the line state parameters; A dynamic baseline construction module is used to input each line state parameter in the time series of the line state parameter into the trained dynamic baseline model to obtain a time series of predicted line state parameter baseline values; A residual calculation module, configured to input the time series of the predicted line state parameter baseline value and the time series of the line state parameter into the residual calculation module to obtain a time series of the line state residual vector; A residual sequence pattern feature extraction module is used to extract residual sequence pattern features from the time series of the line state residual vector to obtain a line state residual sequence pattern feature encoding vector; The abnormality judgment module is used to obtain an abnormality judgment result based on the line state residual sequence pattern feature coding vector.

2. The circuit aging monitoring system for safety electrical devices according to claim 1, characterized in that: The line status parameters include the temperature values of multiple temperature measurement points, the current and voltage values of key nodes, and the leakage current value.

3. The circuit aging monitoring system for safety electrical devices according to claim 1, characterized in that: The dynamic baseline model is trained using historical data of line status parameters that are labeled as operating normally.

4. The circuit aging monitoring system for safety electrical devices according to claim 1, characterized in that: Residual sequence pattern feature extraction module, including: a node extraction unit, configured to extract a current line state residual vector from the time series of line state residual vectors and define other line state residual vectors in the time series of line state residual vectors as line state residual vectors to be propagated to obtain a time series of line state residual vectors to be propagated; a simulated dynamic liquid message propagation unit configured to perform line state simulated dynamic liquid message propagation on each of the line state residual vectors to be propagated in the time sequence of the line state residual vectors to be propagated based on a line state liquid interaction field between each of the line state residual vectors to be propagated and the current line state residual vector to obtain a set of line state message dynamic coding vectors; The fusion unit is used to fuse the set of line state message dynamic coding vectors and the current line state residual vector to obtain a line state residual sequence pattern feature coding vector.

5. The circuit aging monitoring system for safety electrical devices according to claim 4, characterized in that: Dynamic liquid message transmission unit, used for: Calculating the line state dynamic transmission potential energy of each to-be-propagated line state residual vector in the time series of the to-be-propagated line state residual vector relative to the current line state residual vector to obtain a time series of the line state dynamic transmission potential energy encoding vector; Calculating a line state message propagation repulsive force coefficient of each line state residual vector to be propagated relative to the current line state residual vector in the time series of the line state residual vector to be propagated to obtain a time series of the line state message propagation repulsive force coefficients; Based on the time series of the line state message propagation repulsive force coefficient and the time series of the line state dynamic transmission potential energy encoding vector, line state simulated dynamic liquid message propagation is performed on each line state residual vector to be propagated in the time series of the line state residual vector to obtain a set of line state message dynamic encoding vectors.

6. The circuit aging monitoring system for safety electrical devices according to claim 5, characterized in that: Fusion unit for: Performing multi-body propagation cooperative renormalization on the set of line state message dynamic coding vectors to obtain a set of optimized line state message dynamic coding vectors; The set of line state message dynamic coding vectors and the current line state residual vector are fused and optimized to obtain a line state residual sequence pattern feature coding vector.

7. The circuit aging monitoring system for safety electrical devices according to claim 6, characterized in that: Abnormal judgment module, used for: The line state residual sequence pattern feature encoding vector is input into a pre-trained machine learning classifier to obtain an abnormality judgment result, which includes normal fluctuations or real aging abnormalities.

Images