A State-Based Monitoring Method for Distribution Cable Branch Boxes

Abstract

This invention discloses a state monitoring method for distribution cable branch boxes based on state recognition, specifically relating to the field of power monitoring technology. It involves collecting multi-dimensional operational data from multiple branch boxes over a continuous time period to construct a state evolution sequence; calculating state deviation scores and adjacent equipment state consistency scores to determine whether an assessment process is triggered; after triggering, constructing a state influence probability map, identifying abnormal influence paths, extracting state disturbance diffusion indices and cooperative behavior deviation indices, inputting these into a pre-trained risk identification model, generating risk level intervals and causal probability distribution vectors, executing influence control measures, and updating the state recognition logic. This invention achieves dynamic perception of branch box states by constructing a state evolution sequence, combines state deviation scores and adjacent equipment state consistency scores to achieve joint judgment of individual and group behaviors, and drives control strategies and map updates using risk level intervals and causal probability distribution vectors, thereby improving monitoring accuracy and system adaptability.

Description

Technical Field

[0001] This invention relates to the field of power monitoring technology, and more specifically, to a method for monitoring the status of distribution cable branch boxes based on status identification. Background Technology

[0002] As a critical node in medium-voltage power distribution networks, distribution cable branch boxes play a vital role in power distribution and transfer in urban power grids, industrial parks, rail transit, and large-scale infrastructure power supply systems. Due to their wide geographical distribution, complex structural connections, and frequent outdoor or semi-enclosed environments, branch boxes are highly susceptible to the combined effects of factors such as temperature and humidity fluctuations, cable aging, abnormal current carrying capacity, environmental interference, and poor equipment contact during long-term operation. This leads to the gradual accumulation of latent faults such as partial discharge, insulation degradation, and loose electrical connections, ultimately triggering system-level failures or cascading tripping events.

[0003] Traditional methods for monitoring the condition of branch boxes mainly rely on periodic manual inspections, infrared thermography, and discrete data acquisition equipment to obtain status data. These methods have the following significant shortcomings:

[0004] Response lag and coverage blind spots: Manual inspections make it difficult to continuously monitor all equipment, and status changes can only be recorded after an event occurs, making it impossible to capture abnormal behavior during the incubation period;

[0005] Single-point threshold monitoring model: Existing online monitoring systems mostly adopt alarm strategies based on fixed thresholds to detect over-limits of single physical quantities such as current, voltage, temperature, humidity or partial discharge. However, they are difficult to adapt to the natural fluctuation characteristics of different equipment under different seasons and load conditions, resulting in false alarms or missed alarms.

[0006] Lack of behavioral evolution perspective and causal reasoning ability: Current technology usually judges the collected data according to the "single point state at the current moment", ignoring the trend information of the device state evolution over time, and fails to establish a state transmission mechanism between devices, and cannot identify the propagation and regional coordination of abnormal states.

[0007] In recent years, with the development of the power Internet of Things and edge computing technologies, more and more distribution cable boxes have been equipped with multi-dimensional sensing terminals, capable of collecting various state data in real time, including current, voltage, temperature, humidity, and partial discharge, forming massive amounts of equipment operation sequence data. This provides a foundation for modeling and intelligent identification based on equipment state change trajectories. However, how to effectively extract representative behavioral features from this high-dimensional time-series data, identify the evolution trend of abnormal states, and establish a cross-device causal model in conjunction with a distributed structure remains one of the current technical challenges in the industry. Therefore, this invention proposes a state monitoring method for distribution cable distribution boxes based on state recognition, aiming to solve the above problems. Summary of the Invention

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] The method for monitoring the status of distribution cable branch boxes based on status recognition includes the following steps:

[0010] Data on current, voltage, temperature, humidity, and partial discharge of multiple branch boxes were collected over a continuous time period, and combined with historical data to obtain the state evolution sequence of each branch box.

[0011] Based on each state evolution sequence, calculate the state offset score and the state consistency score of adjacent devices. If any score exceeds the corresponding preset threshold, the state assessment process is triggered. If none of them exceed the threshold, real-time updates and monitoring are maintained.

[0012] After triggering the assessment, a probability map of the state impact between devices is constructed based on the historical state event sequence to identify potential abnormal impact paths. Two types of indices are extracted from this map: the state disturbance diffusion index, which measures the intensity of the current state anomaly spreading to other devices through causal paths; and the cooperative behavior deviation index, which measures the degree of difference in temporal characteristics between the current device behavior and the behavior of its neighboring group.

[0013] Input the state disturbance diffusion index and the cooperative behavior deviation index into the pre-trained risk identification model to generate the risk level range and cause probability distribution vector of the current state, which are used to characterize the risk scenario type to which the current state belongs and the corresponding abnormal cause.

[0014] Based on the generated risk level range and causal probability distribution vector, corresponding impact control measures are implemented according to the preset strategy, and the causal path map in the state identification logic is iteratively updated according to the causal distribution.

[0015] In a preferred embodiment, the state evolution sequence includes current characteristic values, change slope, and trajectory deviation, which are used to express the evolution trend of the device. The current characteristic value is the raw physical quantity data collected at the current time point, including the instantaneous values ​​of current, voltage, temperature, humidity, and partial discharge. The change slope is the rate of change obtained by differentiating the current characteristic values ​​of two adjacent time points and dividing by the time interval, which is used to reflect the dynamic growth or decay trend of the physical quantity. The trajectory deviation is the result of calculating the distance between the current characteristic value and the mean or median value of the device in the same historical time period, which is used to assess whether the current state of the device deviates from its long-term behavioral trajectory. Both the change slope and the trajectory deviation are numerically normalized in a standardized manner to ensure comparability.

[0016] In a preferred embodiment, the trajectory deviation is calculated based on any one of Euclidean distance, cosine similarity, or Mahalanobis distance.

[0017] In a preferred embodiment, the state deviation score is calculated by combining the current feature value, the slope of change, and the trajectory deviation in the state evolution sequence to obtain a specific numerical value. This value is used to assess the overall deviation of the current state from the historical behavior model. The current feature value and the slope of change represent the instantaneous state and trend change, respectively, while the trajectory deviation represents the historical consistency over time. These three values ​​are first normalized to form a standard feature vector. Then, this standard feature vector is matched with the set of historical normal state sample feature vectors of the device in the corresponding season and time period. The calculation method involves using the Mahalanobis distance function to evaluate the covariance matrix of the current standard feature vector and the historical feature distribution. The statistical distance of the current state from the center of this distribution is output as the state deviation score. This score is a single numerical value that represents the degree to which the device's state deviates from its historical normal distribution. The higher the state deviation score, the greater the difference between the current state and the historical stable state, indicating a potential anomaly or trend evolution risk. The state deviation score is generated solely based on the device's own state evolution sequence, making it targeted and traceable.

[0018] In a preferred embodiment, the adjacent device state consistency score is calculated by comparing the state evolution sequence of the current device with the state evolution sequences of multiple other branch boxes that are structurally or physically adjacent to it. This score measures the consistency between the current device's state evolution trend and the behavior of neighboring devices. Each device's state evolution sequence includes its current eigenvalue, change slope, and trajectory deviation. The score calculation involves two steps:

[0019] The first step is to select a set of adjacent devices that are directly connected in spatial location or power supply path or whose physical distance is less than a set spatial threshold;

[0020] The second step is to calculate the dynamic time-normalized distance between the current device and each neighboring device in the same time period, and take the average distance value of all neighboring devices as the preliminary consistency measurement result. Then, the reciprocal of this result is normalized to obtain the neighboring device state consistency score. The higher the value, the higher the consistency between the current device state and the neighboring device state. The lower the value, the more discrete the current device state evolution trend is relative to the neighboring group. This scoring mechanism can identify possible deviation behaviors in local areas, provide a horizontal comparison basis for whether to trigger the state assessment process, and ensure that the identification process is not only based on individual characteristics, but also combined with the group structure and behavior characteristics for analysis.

[0021] In a preferred embodiment, after triggering the evaluation, a probability map of the state influence between devices is constructed based on the historical state event sequence. The construction process includes the following steps:

[0022] First, collect the state evolution sequences of multiple branch boxes at different time periods and their corresponding risk level ranges, and organize them into a set of historical state event sequences.

[0023] Statistical learning methods are used to analyze the frequency of events in the set, and a directed relation structure is constructed based on the temporal relationship of state evolution between different devices to form an initial event association network.

[0024] Using conditional probability calculation, the joint probability of neighboring devices exhibiting abnormal state offset scores after a certain number of moments is evaluated, given that the current device's state offset score exceeds a threshold, thus establishing a cross-device state causal relationship.

[0025] Based on the correlation strength, i.e. the joint probability, the edge weights are selected to form a sparse graph structure that retains only high-probability paths that exceed the preset probability standard value. This is the inter-device state influence probability map. The map uses the state propagation probability as the edge weight and the historical event time series as the path basis to represent the set of paths that a specific state evolution trajectory may propagate from one device to other devices.

[0026] In a preferred embodiment, the process of identifying potential abnormal impact paths based on the construction of an inter-device state influence probability map includes the following operations:

[0027] Starting with the current device as the starting node, traverse all directed edges in the state influence probability graph that originate from the device, filter out target nodes with edge weights higher than the preset propagation threshold, and form the first layer of affected node set.

[0028] For each first-level node, repeat the above filtering operation to expand the influence link outward layer by layer until the traversal depth reaches the preset propagation step size or the propagation strength is lower than the dynamic termination condition, forming a set of all high-probability state influence paths starting from the current device.

[0029] During the identification process, each path records the device sequence and corresponding propagation probability in its propagation link, and sorts them by the cumulative probability of the paths. The top-ranked paths are selected as the final output results of the abnormal impact paths. This process ensures that starting from a single point of state anomaly, the most likely multi-hop propagation links can be reasonably identified, providing a causal chain basis for the subsequent extraction of the state disturbance diffusion index. At the same time, it ensures that the impact path identification has convergence and judgment accuracy in complex multi-device scenarios.

[0030] In a preferred embodiment, the process of extracting the state perturbation diffusion index and the cooperative behavior deviation index after identifying potential anomalous impact paths includes the following steps:

[0031] For the state disturbance diffusion index, an exponential expression model is constructed based on the propagation probability, path hop count, and state disturbance intensity of each path in the abnormal impact path. The state disturbance diffusion index is defined as the exponential decay function of disturbance intensity multiplied by total path length. That is, the disturbance diffusion index is equal to the disturbance intensity multiplied by the path probability weighting coefficient, and then multiplied by the negative exponential function of hop count. The calculation formula is: D=S×(∑p_i×w_i)×exp(-λ×h), where D is the state disturbance diffusion index, S is the current state offset score, p_i is the propagation probability of path i, w_i is the normalized weight of the state transfer factor in the path, h is the path hop count, and λ is the path decay coefficient, the value of which is set according to the experience of the equipment hierarchy structure.

[0032] For the cooperative behavior deviation index, the state evolution sequences of multiple neighboring devices are regarded as distributed sequences. A deviation measurement model based on temporal feature entropy weight distribution is constructed. The dynamic weighted information entropy between the current device state evolution sequence and the first-layer neighboring device sequence in three dimensions: current feature value, change slope, and trajectory deviation is calculated. The cooperative behavior deviation index is defined as: C=∑(E_j×δ_j), where C is the cooperative behavior deviation index, E_j is the relative information entropy of the j-th feature dimension, and δ_j is the mean square error of the current device and neighboring devices in this dimension. The three dimensions are normalized using historical means and then the sequence is calculated. This method expresses behavioral fluctuations and retains the correlation between evolutionary features between sequences by transferring the composite modeling strategy of "distribution discreteness and change amplitude". Both types of indices only depend on the obtained state evolution sequence, state deviation score and abnormal influence path parameters, and have high endogenousness, evolvability and generalization ability.

[0033] In a preferred embodiment, the pre-trained risk identification model is a fuzzy logic device, which receives the state disturbance diffusion index and the cooperative behavior deviation index as input variables, and uses a fuzzy inference mechanism based on membership function to construct the input-output mapping relationship, which is used to generate the risk level interval and causal probability distribution vector of the current state.

[0034] The state disturbance diffusion index and the cooperative behavior deviation index are defined as input dimensions, and are divided into three level intervals (low, medium, and high) by a three-segment membership function. Each input variable corresponds to three fuzzy linguistic terms, and the input space is thus combined to form nine input state rule combinations. The fuzzy logic device adopts an inference engine based on a fuzzy rule table to map each combination state to a five-level risk level output. The risk level interval is determined by the interval with the largest fuzzy membership value. At the same time, based on the position of the fuzzy aggregation center point of the output layer, and combined with the statistical frequency of the event cause label of each rule in the training samples, a causal probability distribution vector is constructed to represent the probability of each type of abnormal cause belonging to this input state. The risk level interval is used to describe the urgency of the state evolution, and the causal probability distribution vector is used to identify which historical cause mechanisms the current state may be attributed to, thereby assisting in the subsequent implementation of impact control measures and the updating of the causal path map.

[0035] The technical effects and advantages of this invention are as follows:

[0036] This invention collects current, voltage, temperature, humidity, and partial discharge data from multiple branch boxes over a continuous time period, and combines this data with historical data to construct a state evolution sequence for each device, thus achieving a temporal continuous description of key operating states. Compared to traditional methods that primarily rely on single-point static detection, this invention introduces the temporal dimension of state evolution, enabling the system to capture subtle changes that gradually accumulate during device operation. This significantly improves the sensitivity and depth of identifying early abnormal trends, enhancing the timeliness and early warning capabilities of power distribution system operation monitoring.

[0037] This invention not only calculates state offset scores based on the state evolution sequence of a single device, but also simultaneously assesses the state consistency with other devices that are structurally or physically adjacent to it. By using the state consistency scores of adjacent devices as a horizontal comparison indicator, a framework for comparing behavioral consistency among multiple devices is constructed. This mechanism can effectively identify whether a single device deviates from the evolutionary trend within a group, avoiding false triggers due to individual misjudgments or abnormal diffusion. Thus, it achieves collaborative judgment of individual trends and group behavior in monitoring strategies, significantly improving the accuracy and robustness of monitoring results.

[0038] This invention identifies the state disturbance diffusion index and the cooperative behavior deviation index, then inputs them into a pre-trained model to generate the risk level range and causal probability distribution vector of the current state. Based on the results, it automatically executes impact control measures and iteratively optimizes the established path structure in the state identification logic. This process forms a feedback loop between state identification, risk assessment, control execution, and model updates. This not only enables intelligent response capabilities for state monitoring but also allows the system to continuously evolve and optimize during actual operation, enhancing its adaptability and sustainability in dealing with various complex state evolution scenarios. Attached Figure Description

[0039] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0040] Figure 1 This is a schematic diagram of the power distribution cable branch box status monitoring method based on status recognition in this invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0042] Reference Figure 1 The following examples were obtained:

[0043] Example 1: A method for monitoring the status of distribution cable branch boxes based on status recognition, comprising the following steps:

[0044] Data on current, voltage, temperature, humidity, and partial discharge from multiple branch boxes over a continuous time period are collected, and combined with historical data to obtain the state evolution sequence of each branch box. This step is the fundamental data preparation process for the state identification and monitoring function of this invention. By collecting real-time multidimensional state data from multiple branch boxes during a continuous operating cycle, covering key physical indicators such as current, voltage, temperature, humidity, and partial discharge, the operating characteristics and micro-electrical behavior of the equipment can be comprehensively reflected. Combining the historical monitoring data of each branch box, time series alignment and data structure standardization are performed at a unified time scale to obtain a continuous multidimensional state evolution sequence, which is used to express the evolution trend of the equipment state over time in multiple physical dimensions. This state evolution sequence plays a core role in the subsequent feature extraction, trend modeling, and dynamic evaluation, ensuring the continuity and predictability of state judgment.

[0045] Based on each state evolution sequence, a state deviation score and a neighboring device state consistency score are calculated. If either score exceeds a corresponding preset threshold, a state assessment process is triggered; otherwise, real-time updates and monitoring continue. This step is to make a preliminary judgment on whether the current state of the device is abnormal and to decide whether to initiate a higher-level risk identification process. By mathematically modeling the state evolution sequence of each device, a state deviation score reflecting its deviation from the normal operating trajectory and a neighboring device state consistency score measuring the similarity between its state and the evolution trend of neighboring devices are calculated. These two types of scores comprehensively judge whether there are potential state anomalies from two dimensions: "individual behavioral anomalies" and "group consistency disruption." When any score exceeds the corresponding threshold, it indicates that the device has an abnormal evolution trend or has become disconnected from neighboring devices, triggering the assessment process; otherwise, it indicates that the current state is in a stable operating range, and the original monitoring strategy can be maintained. This mechanism is dynamic and targeted, effectively balancing monitoring resources and computational efficiency.

[0046] After triggering the assessment, a probability map of the state impact between devices is constructed based on the historical state event sequence to identify potential abnormal impact paths. Two types of indices are extracted from this map: a state disturbance diffusion index, which measures the strength of the current state anomaly spreading to other devices through causal paths; and a cooperative behavior deviation index, which measures the degree of temporal difference between the behavior of the current device and the behavior of its neighboring group. This step addresses the systemic level of state anomaly identification by constructing a state propagation impact relationship map between devices, i.e., an inter-device state impact probability map, using the temporal relationship and joint probability of historical state events after the assessment is triggered. This map not only includes which devices the current device may affect but also represents the diffusion probability of different paths with probability values. Based on this, abnormal impact paths with higher state propagation risk are further identified. Subsequently, two key assessment indices are extracted: the state disturbance diffusion index reflects the potential for the current state to cause a chain reaction to other devices and measures the degree of threat to system stability; the cooperative behavior deviation index measures whether the behavior characteristics of the current device deviate significantly from those of its neighboring group, helping to identify potential mismatches or individual mutation behaviors. These two indices provide a quantitative basis for subsequent risk level determination.

[0047] The state disturbance diffusion index and the cooperative behavior deviation index are input into a pre-trained risk identification model to generate the risk level interval and causal probability distribution vector of the current state. These vectors characterize the risk scenario type to which the current state belongs and the corresponding abnormal causes. This step introduces an intelligent decision-making mechanism to output the state identification results in a structured form. The pre-trained risk identification model is built based on a large amount of prior state data and actual fault events, and has the ability to establish a mapping relationship between the input indices and risk levels and causes. When the state disturbance diffusion index and the cooperative behavior deviation index are used as input, the model calculates the corresponding risk level interval through a fuzzy inference mechanism and further derives the causal probability distribution vector. The risk level interval can be used to determine whether the current state risk falls into the normal, minor, moderate, severe, or emergency category, while the causal probability distribution vector reflects the weight distribution of various risk causes that may correspond to the current state, such as electrical failure, environmental impact, or structural aging. This result constitutes the core judgment output of intelligent monitoring.

[0048] Based on the generated risk level intervals and causal probability distribution vectors, corresponding impact control measures are implemented according to a preset strategy, and the causal path graph in the state identification logic is iteratively updated according to the causal distribution. This step achieves closed-loop linkage between risk identification results, monitoring strategies, and operational behaviors. Depending on the different levels of the risk level intervals and the main causal types reflected in the causal probability distribution vectors, the preset strategy will target different control behaviors, such as increasing data sampling frequency, prioritizing maintenance work orders, and activating neighboring devices to enhance monitoring. Furthermore, the causal types in the current identification results can also be used to correct the causal path graph used by the state identification model, i.e., adjusting the weights of the impact chains, thereby improving the model's accuracy in identifying similar future states and achieving adaptive evolution of the model. This update mechanism ensures continuous optimization of the monitoring logic, forming continuous learning and risk adaptation capabilities.

[0049] In this invention, to achieve dynamic identification and evolution monitoring of the operating status of power distribution cable branch boxes, it is first necessary to construct a state evolution sequence that can truly reflect the operating trend of the equipment. The state evolution sequence refers to a data sequence composed of multiple physical state characteristics of the branch box in chronological order over a continuous time period, used to express the changing trend and behavioral trajectory of the equipment in multidimensional indicators. The state evolution sequence includes three types of core temporal feature data: current feature value, change slope, and trajectory deviation. These three types of features are uniformly constructed into a multidimensional temporal feature structure, which is used together to express the evolution trend of the equipment.

[0050] Here, the current characteristic value refers to the original physical quantity value collected for each monitoring indicator at the current monitoring time point. The monitoring indicators involved in this invention include, but are not limited to, current, voltage, temperature, humidity, and partial discharge intensity, all of which are physical quantities that can reflect the electrical and environmental operating status of the equipment in real time. Taking temperature as an example, the current characteristic value is the temperature reading measured by the sensor at a specific moment, usually in degrees Celsius.

[0051] The slope of change is used to quantify the rate of increase or decrease of a physical quantity in a device at two adjacent time points. Specifically, for each type of current characteristic value, the slope of change is calculated as follows: subtract the characteristic value of the previous time point from the characteristic value at the current time point, and divide by the sampling interval between the two time points to obtain the rate of change of the physical quantity per unit time. For example, if the temperature of a branch box is 48℃ at time t1 and 45℃ at time t0, with an interval of 10 minutes between the two times, then the slope of the temperature change is (48-45) / 10 = 0.3℃ / min. This feature reflects the dynamic evolution trend of the state and can be used to identify risk characteristics such as rapid temperature rise and sudden voltage drop.

[0052] Trajectory deviation is used to represent the degree of deviation of the current state from the device's behavior pattern under historical normal conditions. The calculation of trajectory deviation requires constructing a "baseline behavior model" based on state data collected within similar historical time periods. "Similar time periods" can include historical records under dimensions such as the same season, similar environmental load, and similar runtime. When comparing the current feature value with the historical sample set, the deviation calculation method can choose one of the following three models:

[0053] Euclidean distance: Calculates the straight-line distance between the current feature vector and the historical mean vector in multidimensional space;

[0054] Cosine similarity: The cosine value of the angle between vectors reflects the consistency of their evolution direction. The smaller the angle (i.e., the larger the cosine value), the smaller the deviation.

[0055] Mahalanobis distance: A multivariate distance metric that considers the covariance relationship between features. It is suitable for high-dimensional feature spaces with strong correlations and can avoid the error amplification problem caused by different distributions of features.

[0056] For example, during a low-load nighttime period in winter, the current fluctuation range of a certain branch box under normal conditions is [70, 75] A, and the voltage is stable at [380, 385] V. At the current moment, the current is 85 A and the voltage is 390 V. The Euclidean distance can be used to calculate the distance between the current value and the historical vector formed by the mean center mentioned above, yielding a numerical deviation. To improve the comparability of various indicators and avoid the influence of dimensions and scale on the analysis results, the slope of change and trajectory deviation are standardized. That is, the original values ​​of different feature dimensions are normalized by mean zero center and standard deviation normalization to ensure that their distribution has a uniform scale. The standardized results are used for subsequent scoring and model input, ensuring that the state recognition model has consistency and numerical stability when processing multiple types of physical quantities. These three types of features together constitute the state evolution sequence of the equipment. The sequence can be continuously updated in a sliding time window manner, providing the input basis for subsequent key steps such as state offset scoring, adjacent device consistency scoring, abnormal path identification, and index generation.

[0057] State deviation score is a crucial quantitative parameter used in this invention to quantify whether the operating state of an individual device deviates from its historical normal behavioral trajectory. This score is calculated based on the time dimension of a single device and its constructed state evolution sequence. The state evolution sequence includes three key sub-items: current characteristic value, change slope, and trajectory deviation. Each sub-item expresses the device's current instantaneous state, change trend, and long-term behavioral consistency, respectively.

[0058] Before calculating the score, these three feature dimensions need to be standardized. Standardization refers to converting the original physical quantities (such as current, voltage, temperature, etc.) into dimensionless relative values, making features of different units and scales comparable within the same evaluation space. The standardization method uses mean centering and standard deviation normalization, that is, subtracting the historical mean from the original value and then dividing by the standard deviation to form a standard feature vector. Subsequently, the standardized feature vector at the current moment is matched with the feature distribution of the equipment under historical normal conditions. The selection method for the set of historical normal state sample feature vectors is as follows: from historical records that have temporal similarity and load condition matching with the current operating environment, several standard sample data representing "normal state" are selected. For example, the interval with small fluctuation range and stable score in the equipment status records under the same season and the same operating time period is selected as the sample set. This set can represent the typical state behavior distribution of the equipment under normal conditions.

[0059] In the matching calculation, the Mahalanobis distance function is used for evaluation. Mahalanobis distance is a multivariate distance metric that considers the covariance relationship between features across dimensions, and is suitable for calculating deviations when multiple features are correlated. The core idea of ​​this function is to measure the distance between the standard feature vector and the mean vector of the historical sample set, while simultaneously weighting the features across different dimensions based on the covariance matrix. The specific calculation formula is as follows: ;

[0060] Where x is the current feature standard vector, μ is the mean vector of the historical sample set, and Σ is the covariance matrix of the sample set. Let represent its inverse matrix, and T represent the transpose operation. The statistical distance D is the current device state offset score. The score is a single numerical value, representing the degree to which the current device state deviates from its historical stable distribution in the multidimensional feature space. The larger the score, the more abnormal or unstable the current state is, and the more likely it is to be in an atypical operating state or an early stage of fault evolution.

[0061] For example, at a certain moment, the standard feature vector of a branch box is [1.2, 0.8, 1.5], while its historical normal behavior mean vector is [0, 0, 0], and the covariance matrix is ​​an identity matrix. The Mahalanobis distance is approximately 2.1, indicating a significant shift in the current state, possibly near the warning state boundary. It should be noted that this score is calculated entirely based on the device's own state evolution sequence, without relying on the behavior of adjacent devices, thus possessing individual independence, traceability, and high sensitivity. In subsequent state assessment processes, the state shift score and the adjacent device state consistency score together constitute the triggering condition, ensuring that state identification considers both individual biases and incorporates a group behavior comparison mechanism, thereby enhancing the intelligent identification capability of the monitoring system.

[0062] The adjacent device state consistency score is an important indicator used to measure whether the state evolution trend of the current branch box device during its operation remains synchronized or consistent with that of its neighboring devices. As a horizontal reference mechanism in the state identification system, this score is mainly used to identify whether a device exhibits "behavioral anomalies" or "feature isolation" within a group of devices. It can effectively assist the state offset score in multi-dimensional collaborative judgment, thereby improving the overall accuracy and robustness of identification.

[0063] This score is calculated based on the state evolution sequence of the current equipment and its adjacent branch boxes. The state evolution sequence, as defined earlier, consists of three types of features: current characteristic value, change slope, and trajectory deviation. It is used to express the trajectory and evolution trend of the equipment's operating state over a certain period. The score calculation process includes the following two steps:

[0064] Step 1: Determining the set of adjacent devices. First, it's necessary to determine the adjacent objects of the current device. Adjacency relationships can be established using two criteria:

[0065] Spatial adjacency: refers to other branch boxes that are physically within a certain set spatial threshold of the current device. For example, if the spatial threshold is set to 100 meters, other devices within a 100-meter radius of the current device will be considered adjacent devices.

[0066] Power supply path adjacency: This refers to other devices that have a direct cable connection to the current device or belong to the same power supply link, forming an electrical structural adjacency relationship. By determining any one of the above conditions or a combination of both, a set of adjacent devices for the current device can be formed, which is used for subsequent state comparison calculations.

[0067] Step Two: Calculation of State Consistency. After determining the set of adjacent devices, it is necessary to evaluate whether the state evolution trends of the current device and its neighboring devices are consistent. To this end, the state evolution sequences of each device within the same time window are selected for comparison, and the Dynamic Time Warped Distance (DTW) is used to calculate the differences. DTW is a method for measuring the similarity of two time series under non-strictly aligned time scales, and is suitable for situations where there are sampling offsets or different rhythms between devices.

[0068] The specific method is as follows: Using the current device's state evolution sequence as a baseline sequence, calculate the DTW distance between each device and its neighboring state sequences, obtaining multiple distance values. Averaging all distance values ​​yields the average dynamic time-warped distance between the current device and its neighboring group. This average value represents the degree of consistency between the current device and its neighboring devices in terms of multidimensional feature evolution trends. To ensure the score aligns with the representation logic of "higher consistency, higher score," the reciprocal of the average distance value is taken, and then normalized to a score between 0 and 1, which is the final neighboring device state consistency score. A value closer to 1 indicates a high degree of consistency in state evolution between the current device and its neighboring devices; a lower value indicates a significant difference in the current device's evolution trajectory, potentially indicating localized mutations, early anomalies, or passive interference.

[0069] Application Example: For instance, the state evolution sequence of current, temperature, and partial discharge characteristics collected from the current branch box A and its three adjacent devices B, C, and D within the last 15 minutes is used to calculate the matching distances with B, C, and D using DTW, which are 1.2, 1.0, and 1.5 respectively. The average distance is 1.23, and the reciprocal is 0.813. After normalization, the consistency score is approximately 0.85, indicating that the current state trend is generally synchronized with the neighborhood and there are no obvious anomalies. The consistency score of adjacent device states provides a horizontal comparison basis for whether to trigger the state assessment process. It provides an important supplementary judgment when the state deviation score cannot independently determine anomalies. At the same time, by introducing the idea of ​​collaborative analysis of group behavior, this score enhances the ability of this invention to identify sudden isolated events or regional anomalies in actual engineering environments, and improves the overall intelligence level of the state recognition mechanism.

[0070] In the state identification method of this invention, to identify the potential propagation paths of abnormal states in the distribution cable branch box network, a state influence probability map between devices needs to be constructed after the state assessment is triggered. This map is used to express the historical probability structure of the propagation of specific state anomalies between devices and is one of the core foundations for subsequent extraction of state disturbance diffusion indices, auxiliary risk assessment, and dynamic control strategy generation. The construction process of the state influence probability map includes the following disclosed steps:

[0071] Step 1: Constructing the historical state event sequence set. First, it's necessary to collect state evolution data from multiple branch box devices during their past operation. The structure of the state evolution sequence has been defined previously, including three types of time features: current characteristic value, change slope, and trajectory deviation. Based on this, the evolution sequence within each time period is paired with the risk level interval corresponding to the system monitoring results for that period, forming a structured historical state event sample. The risk level interval is a classification output representing the degree of risk of the equipment state at a specific time period, typically divided into five levels: normal, minor anomaly, moderate anomaly, severe anomaly, and emergency. This data is then archived and organized along a timeline, forming a set of "state-risk" pairs arranged chronologically, constituting the historical state event sequence set. This set is used for statistical learning and causal path analysis.

[0072] Step Two: Construct an initial event association network. Based on the historical set of state event sequences, statistical learning methods are used to analyze the order of state anomalies among devices. This involves statistically analyzing whether adjacent devices experience similar shifts shortly after a device's state offset score increases (i.e., enters an abnormal range). If such joint events occur frequently in history, it indicates a possible causal relationship between the two. In the specific implementation, directed edges are established to point "first abnormal" devices to "later abnormal" devices, forming an initial event association network based on chronological order. In this network structure, nodes represent devices, directed edges represent the chronological relationship of state influence, and the existence of edges is based solely on frequency statistics.

[0073] Step 3: Establishing State-Cause Relationships. To further extract the true causal impact from statistical linkages, a conditional probability calculation mechanism is introduced based on the initial network. Specifically, the conditional probability of the following joint event is calculated: If the state offset score of device A exceeds a threshold at time t (entering an abnormal state), what is the probability that the state offset score of adjacent device B also exceeds the threshold at time t+Δt? This conditional probability can be statistically analyzed using a counter to count the frequency of device B's abnormality under the precondition (device A is abnormal) in all historical event sequences. This joint probability is used to quantify the probability of cross-device abnormality propagation. The core of this step is to supplement the previous temporal sequence mapping with a probability-weighted structure, thereby realizing the transformation model from "correlation" to "causation".

[0074] Step 4: Select edge weights to construct a sparse graph structure. After obtaining the joint probability among all devices, assign a corresponding conditional probability as the edge weight to each directed edge, which represents the probability value of the state propagating from the source node to the target node.

[0075] To improve recognition efficiency and model sparsity, only edges with a joint probability exceeding a preset probability standard (such as 0.3 or 0.5) are retained, while other edges are deleted. The resulting directed graph is the sparse structured state influence probability graph. This graph satisfies the following technical attributes: nodes are device numbers; edges are state anomaly propagation paths; edge weights are state propagation probabilities (ranging from 0 to 1); the directed structure represents the state propagation direction; and only paths with actual propagation evidence in history are retained.

[0076] Application Example: For instance, in historical operational data, device A's state offset score showed anomalies five times consecutively. Four of these anomalies occurred within 15 minutes of its physical neighbor device B. The conditional probability is 4 / 5 = 0.8, significantly higher than the preset standard value of 0.5. Device A → B is thus considered a state propagation path with an edge weight of 0.8. Ultimately, this inter-device state influence probability map is formed after the assessment trigger but before the risk identification model input. It serves as the core bridge connecting the anomaly identification logic and behavioral path deduction. Its structure not only serves the generation of the state disturbance diffusion index but can also be used for risk transmission prediction and resource allocation priority ranking in control strategies.

[0077] Based on the constructed probability map of inter-device state impact, to further quantify the propagation trend of state anomalies, it is necessary to identify potential anomaly impact paths for devices currently in an abnormal state. This identification process aims to start from devices whose current state offset scores have triggered assessment conditions, combine this with the historical experience structure in the state impact probability map, identify the downstream device links that may be affected, and quantify the propagation probability of each path. This operation provides the path basis for generating the state disturbance diffusion index and is used for subsequent risk linkage response, local early warning zoning, and resource allocation control decisions. The identification process includes the following steps:

[0078] Step 1: Initialize the starting point and first-level node filtering. The device currently in an abnormal state is used as the starting point node. This device's evaluation process is triggered by its state offset score or the state consistency score of its adjacent devices. Extract all directed edges originating from this starting point node from the state influence probability graph. Each directed edge represents a path that historically started from the current device's abnormal state and may have affected other devices; the edge weight represents the propagation probability.

[0079] Target nodes with edge weights (i.e., propagation probabilities) higher than a preset propagation threshold are selected from this set. These target nodes constitute the first layer of affected nodes for the device. The propagation threshold is a preset control parameter used to filter low-impact paths and ensure that the path identification results are representative of actual risks. For example, if the propagation threshold is set to 0.4, then a path is only included in the analysis path if the propagation probability of a certain edge is ≥0.4.

[0080] Step Two: Multi-layer recursive path expansion. For the first-layer affected node, repeat the above operation: treat it as the current node and continue searching downwards for target nodes whose outgoing edges have a weight greater than the propagation threshold, forming the second layer of affected nodes. This process is repeated layer by layer to form a path tree structure. To prevent infinite expansion or path depth exceeding the analysis capability, the following two types of termination conditions are introduced: Preset propagation step size: Limits the maximum number of hops in the path (e.g., no more than 3 layers) to control the path structure within a reasonable complexity; Dynamic propagation strength limit: If the propagation probability of all outgoing edges of a node is lower than a certain dynamically adjusted threshold, the propagation trend is considered weakened, and the path expansion is automatically terminated. Through this layer-by-layer expansion mechanism, a complete set of paths originating from the current device and meeting the propagation conditions can be constructed, representing the potential range of abnormal influence in the probabilistic graph.

[0081] Step 3: Path Recording and Sorting Output. During the expansion process described above, the system records the device sequence and corresponding propagation probability for each path. To facilitate path effectiveness evaluation, the cumulative probability of each path is calculated, which is the product of the propagation probabilities of each edge in the path, representing the connectivity and propagation probability of the entire path. For example, if the propagation probabilities of the three edges in path A→B→C are 0.8, 0.7, and 0.6 respectively, then the cumulative propagation probability of this path is 0.8 × 0.7 × 0.6 = 0.336.

[0082] All paths are sorted in descending order of their cumulative probabilities, and the top few paths (e.g., the top 5 or those with cumulative probabilities covering the top 80%) are selected as the final output of the anomaly impact paths. This result not only reflects the scope of the current anomalous device's impact but also clarifies the structural chain of its high-risk impact paths.

[0083] Example Explanation: Assume device A triggers state assessment. A→B (edge ​​weight 0.8), A→C (edge ​​weight 0.2), B→D (edge ​​weight 0.6), C→D (edge ​​weight 0.5). Let the propagation threshold be 0.3 and the maximum step size be 2. Then: First step, B is eliminated; second step, the propagation from B to D is expanded (because edge weight 0.6 > 0.3); the cumulative probability of path A→B→D is 0.8 × 0.6 = 0.48, which can be included in the final output; path A→C is excluded (0.2 < 0.3). This identification process realizes multi-hop propagation link identification starting from a single point of anomaly, derived from historical statistical probability structure. This enables the system to complete convergent path search even in complex multi-device scenarios, and possesses strong logical interpretability and risk orientation. Simultaneously, this path structure provides key factors such as path number, depth, and propagation weight for the subsequent calculation of the "state disturbance diffusion index," playing a crucial role in the overall state identification mechanism.

[0084] After identifying the abnormal impact paths of the current equipment, two core quantitative indicators need to be extracted from the identified path results for subsequent risk identification modeling: the state disturbance diffusion index and the cooperative behavior deviation index. These two indices, as a fusion expression of the equipment in multi-dimensional state analysis and group behavior analysis dimensions, are important foundational variables for constructing the risk identification system in this invention. The calculation processes for both are as follows:

[0085] I. Extraction Method of State Disturbance Diffusion Index: The state disturbance diffusion index is used to measure the probability and intensity of the current device state anomaly spreading to other devices in the graph structure, reflecting the trend assessment of the current anomaly's impact on the overall system. This index is constructed based on the identified set of anomaly impact paths. Each path includes key attributes such as device sequence, path propagation probability, and path hop count. The index expression model is also constructed by referencing the current device state offset score. The definition is as follows: The calculation formula is: D = S × (∑p_i × w_i) × exp(-λ × h), where D is the state disturbance diffusion index, which is the final output result; S is the current state offset score, representing the degree of anomaly in the current state, obtained based on Mahalanobis distance using the method described above; p_i is the propagation probability of path i, derived from the edge weights in the state influence probability graph; w_i is the normalized weight of the state transmission factor in the path, representing the relative importance of propagation capability in each path, which can be set proportionally according to the path propagation strength and structural complexity, assigned by experts, or set by other alternative methods; h is the path hop count, i.e., the total number of device nodes in the path, reflecting the length of the anomaly propagation chain; λ is the path attenuation coefficient, an empirical value determined based on network topology complexity or hierarchical structure. The larger the value, the more emphasis is placed on the hop count attenuation effect. In this formula, the weighted sum of the propagation probability and the transmission weight represents the concentrated propagation capability of the entire path. Multiplying it by the state offset score expresses the initial anomaly strength multiplied by the propagation trend. The negative exponential decay function of hop count is used to reflect that the longer the path length, the greater the decay of its overall impact, thereby controlling the exaggerated expression of abnormal impacts on remote nodes.

[0086] II. Extraction Method of Cooperative Behavior Deviation Index: The Cooperative Behavior Deviation Index measures whether the behavior pattern of a current device in its state evolution sequence deviates significantly from that of its neighboring group of devices, emphasizing the importance of consistency in individual behavior within the group structure. This index models the state evolution sequences of multiple neighboring devices and constructs a temporal feature entropy weight distribution model to measure behavioral differences. The cooperative behavior deviation index is defined as: C = ∑(E_j × δ_j), where C is the cooperative behavior deviation index, E_j is the relative information entropy of the j-th feature dimension, reflecting the dispersion of the behavioral pattern distribution in this dimension, and δ_j is the mean square error of the current device and neighboring devices in this dimension, representing the degree of outlier in the behavioral trend on the order of magnitude. Feature dimension j includes three items: current feature value, change slope, and trajectory deviation, all of which are the core components of the state evolution sequence defined above. The three dimensions are normalized using historical means and then used for sequence calculation. This method, by transferring the composite modeling strategy of "distribution dispersion and change amplitude", expresses both behavioral fluctuations and retains the correlation of evolutionary features between sequences. Both types of indices depend only on the derived state evolution sequence, state deviation score, and abnormal influence path parameters, and have high endogeneity, evolvability, and generalization ability. Information entropy is calculated using the Shannon entropy method. If the state changes of adjacent devices are relatively concentrated in the current feature value dimension, the entropy value is low, indicating strong group consistency; if the behavior is widely distributed, the entropy value is high, indicating the existence of unstable or noisy behavior.

[0087] In the state monitoring method of this invention, after identifying the state disturbance diffusion index and the cooperative behavior deviation index, a pre-trained fuzzy logic device is used as the risk identification model to further determine the risk level of the current equipment state and its potential causes. This model has the ability to drive multi-level fuzzy rule reasoning with uncertain inputs, and can combine the mapping relationship between state evolution characteristics and risk labels in historical samples to output interpretable risk intervals and cause types, providing key basis for the implementation of state control measures.

[0088] In this fuzzy logic model, the state disturbance diffusion index and the cooperative behavior deviation index are used as input variables, representing the risk trends of the current device in the direction of abnormal influence propagation and the direction of group behavior deviation, respectively. Both indices have been calculated using variables such as path propagation probability, hop count, information entropy, and mean square error, as previously disclosed. They possess evolvability and a normalized structure, making them convenient as fuzzy logic inputs. For each input variable, a set of three-segment membership functions is constructed, dividing its value range into three intervals: "low," "medium," and "high." The three-segment membership functions can take the following forms: Low: corresponding to the interval below a certain threshold (e.g., 0.3), using a Z-shaped function; Medium: with the center of the interval as the peak value (e.g., 0.5), using a trapezoidal or Gaussian function; High: corresponding to the interval above a certain threshold (e.g., 0.7), using a S-shaped function.

[0089] Each interval is mapped to a fuzzy linguistic term, such as "weak diffusion," "medium diffusion," "strong diffusion," or "small deviation," "medium deviation," "large deviation." In this way, the input space consists of three combinations of two variables, forming 3×3=9 combinations of input state rules, which represent all possible fuzzy conditional intersections.

[0090] The fuzzy logic engine internally constructs a rule-based inference engine, with each rule taking the following form:

[0091] If the "State Disturbance Diffusion Index" is high and the "Cooperative Behavior Deviation Index" is medium, then the "Risk Level" is severely abnormal, and the "Cause of Abnormality" is Category A. During model training, based on the actual risk level and event label corresponding to each state evolution feature in the historical dataset, the above rule set is constructed through statistical learning. Each rule is associated with a set of training frequency weights for subsequent inference. The inference process employs a Mamdani-type fuzzy inference mechanism, including: calculating the membership degree of the current input index value under three membership functions; activating the corresponding rules according to the membership degree combination to obtain multiple fuzzy outputs; aggregating the fuzzy output sets and using a weighted average method to obtain the final output.

[0092] The output consists of two parts:

[0093] Risk level range: Based on a five-segment division, such as "normal", "mildly abnormal", "moderately abnormal", "severely abnormal" and "urgent", the level corresponding to the fuzzy language item with the highest membership degree is the main output;

[0094] Causal probability distribution vector: By statistically analyzing the frequency of historical sample labels corresponding to each activation rule and combining it with the membership degree weighted summation, a normalized vector is formed, representing the relative probability of various abnormal causes under the current input combination.

[0095] The risk level range is used to characterize the urgency of the current state. If the output is "severe anomaly" or "urgent," high-priority impact control measures in the preset strategy are triggered, such as forced power switching or scheduling of backup branches. The causal probability distribution vector is a probability vector of length n, where n is the number of preset abnormal scenario categories in the system (such as abnormal temperature, sudden humidity changes, load fluctuations, discharge increases, etc.). Each dimension value pk∈[0,1] represents the probability that the current state is attributed to this abnormal cause.

[0096] Implementation example: Input state disturbance diffusion index D=0.65, cooperative behavior deviation index C=0.42; D is 0.6 and 0.4 on the "medium" and "high" membership functions respectively, and C is 0.9 on the "medium" membership function; activation rules such as "medium-medium" and "high-medium" correspond to outputs "moderate anomaly" and "severe anomaly"; after aggregation, the fuzzy center falls in the "severe anomaly" interval; causal vector [temperature fluctuation: 0.1, humidity anomaly: 0.3, voltage disturbance: 0.5, discharge enhancement: 0.1].

[0097] The output causal probability distribution vector can be used as a feedback signal to participate in the adaptive update mechanism of the causal path graph. That is, if a certain state causal edge in the path graph is associated with the same causal label multiple times, the statistical weight of the path will be enhanced, thereby improving the self-learning and generalization ability of the state propagation model.

[0098] The mechanism for influencing the implementation of control measures and the iterative update of the causal path map is as follows: In this invention, after completing the risk identification of the current state, two types of key information outputs are obtained: risk level interval and causal probability distribution vector. The former describes the urgency of the current equipment state and is used to quantify which of the five levels—normal, mildly abnormal, moderately abnormal, severely abnormal, or urgent—the current state belongs to; the latter expresses the various possible causes of abnormality behind this state and their probability distribution, providing a basis for anomaly tracing and decision-making.

[0099] Impact control measures are implemented according to preset strategies: Impact control measures refer to the control behaviors automatically executed by the system based on preset strategies when a device's status is identified as having a high risk level or is attributed to a certain type of high-risk anomaly. These strategies are predefined and mapped to risk level intervals and causal probability distribution vectors.

[0100] The control measures may include: Strategy A: Increased data sampling frequency, used for situations where the risk level is in a "mildly abnormal" state, to increase monitoring granularity; Strategy B: Automatic alarm issuance, used for "moderately abnormal" and above levels, pushing real-time alarms with risk level and cause tags to the operation and maintenance platform; Strategy C: Temporary load switching, used for "severely abnormal" and above levels, automatically performing local power distribution load adjustments to avoid further cascading risks; Strategy D: Physical isolation operations, used for "emergency" levels, implementing safety isolation measures such as power outage protection to prevent further spread of the condition; Strategy E: Operation and maintenance command push, combined with the anomaly mechanism with the highest probability of cause, for example, if the probability of "cable insulation deterioration" is 0.78, then a special insulation status investigation task is issued. This strategy library and the output results of the fuzzy logic unit constitute a "rule-action" mapping set. After each identification is completed, the control execution process is immediately entered to achieve adaptive intervention control.

[0101] Iterative Update of the Causal Path Graph Based on Causal Distribution: The causal path graph in this invention is a structure formed based on the previously constructed inter-device state propagation model. This graph contains directed edges for state propagation between device nodes, with edge weights representing propagation probabilities and paths representing possible state diffusion trajectories. To enhance the timeliness and adaptability of the model, the graph needs to be iteratively updated based on the identified causal probability distribution vector. This update process includes the following mechanism:

[0102] Causal label attribution reinforcement: The currently identified primary causal factor (i.e., the one with the highest probability) is mapped to all edges associated with the hit propagation paths in the graph path. If an edge is repeatedly determined to be highly correlated with the same causal factor, the "causal label strength" of that edge is weighted and increased.

[0103] Dynamic correction of propagation probability: If a path is continuously identified as an abnormal propagation path, and the probability of the same cause category is high in each output, the propagation probability of the path is adjusted upward, reflecting the "data-driven correction" capability of the actual propagation trend.

[0104] The new path discovery mechanism: If a skip propagation path exists in the current identification path that was not previously present in the graph structure, but its corresponding edge factor is reasonable (e.g., similar anomalies appear in adjacent devices after a short-term large disturbance), then this edge is added to the graph, forming an expansion mechanism for "potential causal paths". This update process operates based on a closed-loop approach of "causal feedback → graph reconstruction → weight redistribution", ensuring that the causal graph can continuously learn and evolve with state changes, improving the overall adaptive capability and long-term accuracy of the state recognition system.

[0105] Specific implementation example: For instance, the current device number is A, the state disturbance diffusion index D=0.71, and the cooperative behavior deviation index C=0.58; the risk identification result shows that the risk level range is "severe anomaly", and the "cable partial discharge degradation" label accounts for 0.63% in the cause probability distribution vector; Execution strategy C: Immediately switch branch box A to the backup branch to avoid concentrated current load; at the same time, increase the weight of the edge factors marked with "partial discharge degradation" in the A→B→C three-hop propagation path; if the next round of evaluation of device B also confirms the existence of the same cause, then automatically increase the propagation probability of this path by 10%, forming a path reinforcement learning mechanism. Through the above mechanism, this invention realizes a closed-loop intelligent identification and control system of risk identification-control decision-model update, which not only has the ability to respond instantly, but also supports data-driven model self-evolution, significantly improving the response accuracy and operation and maintenance efficiency to complex abnormal states in dynamic power distribution environments.

[0106] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0107] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0108] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0109] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0110] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for monitoring the status of distribution cable branch boxes based on status recognition, characterized in that, Includes the following steps: Data on current, voltage, temperature, humidity, and partial discharge of multiple branch boxes were collected over a continuous time period, and combined with historical data to obtain the state evolution sequence of each branch box. Based on each state evolution sequence, calculate the state offset score and the state consistency score of adjacent devices. If any score exceeds the corresponding preset threshold, the state assessment process is triggered. If none of them exceed the threshold, real-time updates and monitoring are maintained. After triggering the assessment, a probability map of the state impact between devices is constructed based on the historical state event sequence to identify potential abnormal impact paths. Two types of indices are extracted from this map: the state disturbance diffusion index, which measures the intensity of the current state anomaly spreading to other devices through causal paths; and the cooperative behavior deviation index, which measures the degree of difference in temporal characteristics between the current device behavior and the behavior of its neighboring group. The process of constructing the probability map of the state influence between devices includes the following steps: First, collect the state evolution sequences of multiple branch boxes at different time periods and their corresponding risk level ranges, and organize them into a set of historical state event sequences. Statistical learning methods are used to analyze the frequency of events in the set, and a directed relation structure is constructed based on the temporal relationship of state evolution between different devices to form an initial event association network. Using conditional probability calculation, the joint probability of neighboring devices exhibiting abnormal state offset scores after a certain number of moments is evaluated, given that the current device's state offset score exceeds a threshold, thus establishing a cross-device state causal relationship. Based on the correlation strength, i.e. the joint probability, the edge weights are filtered to form a sparse graph structure that retains only high-probability paths that exceed the preset probability standard value, which is the probability map of the state influence between devices. Input the state disturbance diffusion index and the cooperative behavior deviation index into the pre-trained risk identification model to generate the risk level range and cause probability distribution vector of the current state, which are used to characterize the risk scenario type to which the current state belongs and the corresponding abnormal cause. Based on the generated risk level range and causal probability distribution vector, corresponding impact control measures are implemented according to the preset strategy, and the causal path map in the state identification logic is iteratively updated according to the causal distribution.

2. The method for monitoring the status of distribution cable branch boxes based on status recognition according to claim 1, characterized in that, The state evolution sequence includes the current characteristic value, the slope of change, and the trajectory deviation, which are used to express the evolution trend of the device. The current characteristic value is the raw physical quantity data collected at the current time point, including the instantaneous values ​​of current, voltage, temperature, humidity, and partial discharge. The slope of change is the rate of change obtained by differentiating the current characteristic values ​​of two adjacent time points and dividing by the time interval, which is used to reflect the dynamic growth or decay trend of the physical quantity. The trajectory deviation is the result of calculating the distance between the current characteristic value and the mean or median value of the device in the same historical time period, which is used to assess whether the current state of the device deviates from its long-term behavioral trajectory. The slope of change and the trajectory deviation are both numerically normalized in a standardized manner to ensure comparability.

3. The method for monitoring the status of distribution cable branch boxes based on status recognition according to claim 2, characterized in that, The trajectory deviation is calculated based on any one of Euclidean distance, cosine similarity, or Mahalanobis distance.

4. The method for monitoring the status of distribution cable branch boxes based on status recognition according to claim 3, characterized in that, The state offset score is calculated by combining the current feature value, the slope of change, and the trajectory deviation in the state evolution sequence to obtain a specific value. It is used to evaluate the overall degree of deviation of the current state relative to the historical behavior model. When calculating, the three are first normalized to form a standard feature vector. Then, the standard feature vector is matched with the set of historical normal state sample feature vectors of the device in the corresponding season and time period. The calculation method is to use the Mahalanobis distance function to evaluate the covariance matrix of the current standard feature vector and the historical feature distribution, and output the statistical distance of the current state from the center of the distribution as the state offset score.

5. The method for monitoring the status of distribution cable branch boxes based on status recognition according to claim 4, characterized in that, The adjacent device state consistency score is calculated by comparing the state evolution sequence of the current device with the state evolution sequences of multiple other branch boxes that are structurally or physically adjacent to it. This score measures the consistency between the current device's state evolution trend and the behavior of its neighboring devices. Each device's state evolution sequence includes its current eigenvalue, change slope, and trajectory deviation. The score calculation involves two steps: The first step is to select a set of adjacent devices that are directly connected in spatial location or power supply path or whose physical distance is less than a set spatial threshold; The second step is to calculate the dynamic time-normalized distance between the current device and each neighboring device in the same time period, and take the average distance value of all neighboring devices as the preliminary consistency measurement result. Then, take the reciprocal of the result and normalize it to obtain the neighboring device state consistency score.

6. The method for monitoring the status of distribution cable branch boxes based on status recognition according to claim 5, characterized in that, Based on the construction of the inter-device state influence probability map, the process of identifying potential abnormal influence paths includes the following operations: Starting with the current device as the starting node, traverse all directed edges in the state influence probability graph that originate from the device, filter out target nodes with edge weights higher than the preset propagation threshold, and form the first layer of affected node set. For each first-level node, repeat the above filtering operation to expand the influence link outward layer by layer until the traversal depth reaches the preset propagation step size or the propagation strength is lower than the dynamic termination condition, forming a set of all high-probability state influence paths starting from the current device. During the identification process, each path records the device sequence and corresponding propagation probability in its propagation link, and sorts them by the cumulative probability of the paths. The top-ranked paths are selected as the final output results of the abnormal impact paths.

7. The method for monitoring the status of distribution cable branch boxes based on status recognition according to claim 6, characterized in that, After identifying potential anomalous impact paths, the process of extracting the state perturbation diffusion index and the cooperative behavior deviation index includes the following steps: For the state disturbance diffusion index, an exponential expression model is constructed based on the propagation probability, path hop count, and state disturbance intensity of each path in the abnormal impact path. The state disturbance diffusion index is defined as the exponential decay function of disturbance intensity multiplied by total path length. That is, the disturbance diffusion index is equal to the disturbance intensity multiplied by the path probability weighting coefficient, and then multiplied by the negative exponential function of hop count. The calculation formula is: D=S×(∑p_i×w_i)×exp(-λ×h), where D is the state disturbance diffusion index, S is the current state offset score, p_i is the propagation probability of path i, w_i is the normalized weight of the state transfer factor in the path, h is the path hop count, and λ is the path decay coefficient, the value of which is set according to the experience of the equipment hierarchy structure. For the cooperative behavior deviation index, the state evolution sequences of multiple neighboring devices are regarded as distributed sequences. A deviation measurement model based on the temporal feature entropy weight distribution is constructed. The dynamic weighted information entropy between the current device state evolution sequence and the first-layer neighboring device sequence in three dimensions: current feature value, change slope, and trajectory deviation is calculated. The cooperative behavior deviation index is defined as: C=∑(E_j×δ_j), where C is the cooperative behavior deviation index, E_j is the relative information entropy of the j-th feature dimension, and δ_j is the mean square error between the current device and the neighboring devices in this dimension. The three dimensions are normalized using historical mean and then calculated for the sequence.

8. The method for monitoring the status of a distribution cable branch box based on status recognition according to claim 7, characterized in that, The pre-trained risk identification model is a fuzzy logic device. It receives the state disturbance diffusion index and the cooperative behavior deviation index as input variables and uses a fuzzy inference mechanism based on the membership function to construct the mapping relationship from input to output, which is used to generate the risk level interval and causal probability distribution vector of the current state. The state disturbance diffusion index and the cooperative behavior deviation index are defined as input dimensions, and are divided into three level intervals (low, medium, and high) by a three-segment membership function. Each input variable corresponds to three fuzzy linguistic terms, and the input space is thus combined to form nine input state rule combinations. The fuzzy logic device adopts an inference engine based on the fuzzy rule table to map each combination state to a five-level risk level output. The risk level interval is determined by the interval with the largest fuzzy membership value. At the same time, based on the position of the fuzzy aggregation center point of the output layer, and combined with the statistical frequency of the event cause label of each rule in the training samples, a causal probability distribution vector is constructed to represent the probability of each type of abnormal cause belonging to this input state.

Images