Electric leakage early warning method and system based on electric behavior analysis

Abstract

The present application relates to the technical field of electric leakage detection, and particularly relates to an electric leakage early warning method and system based on electric behavior analysis. The method comprises collecting branch electric behavior characteristic parameters of each branch according to a unified clock to generate an electric behavior original sequence; identifying corresponding load stages of each branch to construct an electric behavior state atlas; extracting amplitude drift, phase offset, harmonic migration and peak duration of residual current in each stage to generate an electric leakage dynamic characteristic sequence; obtaining an electric leakage residual component to generate an electric leakage residual sequence; and calculating continuous cumulative intensity and stage regression deviation based on the electric leakage residual sequence to determine an electric leakage risk level of the corresponding branch. The present application can analyze the dynamic characteristics of residual current in different load stages in combination with the electric behavior stage changes of each branch, eliminate the explainable disturbance caused by normal load stage switching, and thus improve the accuracy and advance of electric leakage early warning in complex electric behavior scenarios.

Description

Technical Field

[0001] This invention relates to the field of leakage current detection technology, specifically to a leakage current early warning method and system based on electricity consumption behavior analysis. Background Technology

[0002] With the increasing number of residential and commercial electrical appliances, charging equipment, energy storage devices, and intelligent control loads connected to the distribution box, the power consumption status of each branch circuit exhibits significant dynamic changes. Different branches experience different load stages during operation, such as load startup, load shutdown, load power adjustment, and load charging. The switching between these load stages causes synchronous changes in branch current, load power transitions, and residual current. Especially in scenarios where multiple branches operate alternately, charging loads gradually enter a stable range, and power-adjusting loads repeatedly change their output power, the change in residual current no longer manifests as a single increase in amplitude but may exhibit multiple dynamic characteristics such as amplitude drift, phase shift, harmonic migration, and persistent spikes.

[0003] Existing leakage current detection methods typically rely on the instantaneous amplitude of the residual current in a branch, a fixed threshold, or the number of short-term exceedances as the basis for judgment. When the residual current significantly exceeds the protection threshold, such methods can trigger alarms or protection actions. However, in actual power consumption, load phase switching itself can cause short-term fluctuations in residual current. For example, the load startup phase may generate instantaneous spikes, the load power adjustment phase may generate harmonic changes, the load charging phase may generate slow phase changes, and the load shutdown phase may generate residual fluctuations. The disturbances caused by these normal power consumption behaviors can easily overlap with the early characteristics of actual leakage current, making it difficult for existing methods to accurately distinguish between "explainable disturbances caused by normal load phase switching" and "abnormal leakage current residuals that cannot be explained by load phase changes."

[0004] Furthermore, existing leakage current detection methods primarily focus on whether the residual current exceeds the limit at a single moment or within a single window, lacking analysis of the residual accumulation of the same branch during continuous load phases and the regression situation after the load phase ends. When leakage risk is in its early evolution stage, the residual current may not yet have reached the fixed alarm threshold, but its abnormal residual has been accumulating continuously over multiple load phases, or it may fail to return to normal levels after the load phase ends. Existing methods struggle to identify such continuously evolving hidden leakage risks in a timely manner, easily leading to problems such as delayed warnings, high false alarm rates, and inaccurate judgment of leakage risk levels. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a leakage current early warning method and system based on electricity consumption behavior analysis. This method can analyze the dynamic characteristics of residual current in different load stages by combining the changes in electricity consumption behavior stages of each branch, eliminating explainable disturbances caused by normal load stage switching, retaining leakage current residuals that cannot be explained by electricity consumption behavior states, and further determining the leakage current risk level based on the continuous accumulation and stage regression of leakage current residuals, thereby improving the accuracy and advance warning of leakage current in complex electricity consumption scenarios.

[0006] This invention employs the following technical solution: a leakage current early warning method based on electricity consumption behavior analysis, comprising:

[0007] The branch power consumption characteristic parameters of each branch are collected according to a unified clock. The load power transition is obtained based on the branch power consumption characteristic parameters, and an original sequence of power consumption behavior with branch identifier, sampling time identifier, branch power consumption characteristic parameters and load power transition is generated.

[0008] Based on the original sequence of electricity consumption behavior, the load stage corresponding to each branch is identified, and an electricity consumption behavior state map reflecting the changing relationship of the load stages of each branch is constructed.

[0009] Based on the power consumption behavior state map, the amplitude drift, phase shift, harmonic migration, and peak duration of the residual current at each stage are extracted to generate a dynamic feature sequence of leakage current.

[0010] Based on the leakage current dynamic feature sequence, explainable disturbances caused by load phase switching are eliminated, leakage current residual components are obtained, leakage current residual intensity is calculated based on the leakage current residual components, and the leakage current residual components and leakage current residual intensity corresponding to each load phase are arranged in time order to generate leakage current residual sequence.

[0011] Based on the leakage residual sequence, the continuous cumulative intensity and stage regression bias are calculated to generate the leakage risk evolution trajectory and determine the leakage risk level of the corresponding branch.

[0012] As a further description of the above technical solution: the branch power consumption characteristic parameters include branch voltage, branch current, branch residual current and switch status.

[0013] As a further description of the above technical solution: the method for obtaining the load power transition includes:

[0014] Branch voltage based on the same sampling time and branch current Calculate the intermediate power value of the β-th branch at the nth sampling time, and calculate the load power transition based on the intermediate power values ​​of adjacent sampling times.

[0015] As a further description of the above technical solution: the load phase includes a load start-up phase, a load shutdown phase, a load power adjustment phase, and a load charging phase;

[0016] Methods for identifying the load phase corresponding to each branch include:

[0017] Original sequence of electricity consumption behavior based on the β-th branch Branch currents were read in chronological order of sampling time. Switch status and load power transition Input the pre-trained sampling point type recognition model and output the sampling point type of the β-th branch at the nth sampling time;

[0018] By merging consecutive sampling times belonging to the same sampling point type, we obtain the load start-up stage, load shutdown stage, load power adjustment stage, and load charging stage.

[0019] As a further description of the above technical solution: the sampling point types include load start sampling point, load stop sampling point, load power adjustment sampling point, load charging sampling point and stage hold sampling point.

[0020] As a further description of the above technical solution: the method for merging consecutive sampling moments belonging to the same sampling point type to obtain the load start-up phase, load shutdown phase, load power adjustment phase, and load charging phase includes:

[0021] After continuous confirmation, the sampling point types are merged according to time order. The sampling times of consecutive load start sampling points in the β-th branch are merged into the load start stage; the sampling times of consecutive load stop sampling points are merged into the load stop stage; the sampling times of consecutive load power adjustment sampling points are merged into the load power adjustment stage; and the sampling times of consecutive load charging sampling points are merged into the load charging stage.

[0022] For a phase-hold sampling point, it is merged into the preceding load phase that is adjacent in time; if there is no confirmed load phase before the phase-hold sampling point, no load phase node is generated until a load start sampling point, load stop sampling point, load power adjustment sampling point, or load charging sampling point is subsequently identified.

[0023] As a further description of the above technical solution: the method for constructing the electricity consumption behavior state map includes:

[0024] The sampling times of consecutive load stages within the same branch are merged into one load stage node. For two adjacent load stage nodes within the same branch, a state switching relationship is established. All load stage nodes and their state switching relationships constitute an electricity consumption behavior state map.

[0025] As a further description of the above technical solution: the method for obtaining the amplitude drift of the residual current includes:

[0026] Based on the electricity consumption behavior state map, the r-th load stage node of the β-th branch is read sequentially. According to the start time of the phase and the end of the phase The original sequence of electricity consumption behavior from the β-th branch Extract the residual current of the branch within the corresponding time interval to obtain the residual current stage sequence of the branch corresponding to the load stage;

[0027] Select a continuous load at the beginning of the r-th load phase. Using 1 sampling point as the front-end analysis window, a continuous sampling point is selected at the end of the r-th load stage. Each sampling point serves as the backend analysis window;

[0028] Calculate the average absolute value of the residual current in the branch within the front-end analysis window and the average absolute value of the residual current in the branch within the back-end analysis window, and determine the absolute value of the difference between the two as the residual current amplitude drift for that load stage.

[0029] As a further description of the above technical solution: the method for obtaining the phase offset of the residual current includes:

[0030] Based on the electricity consumption behavior state map, the r-th load stage node of the β-th branch is read sequentially. According to the start time of the phase and the end of the phase The original sequence of electricity consumption behavior from the β-th branch Extract the branch current within the corresponding time interval to obtain the branch current stage sequence corresponding to the load stage;

[0031] The fundamental component of the branch current phase sequence within the r-th load phase is extracted to obtain the fundamental phase sequence of the branch current; the fundamental component of the residual current phase sequence within the r-th load phase is extracted to obtain the fundamental phase sequence of the residual current.

[0032] Calculate the average value of the fundamental phase sequence of the branch current and the average value of the fundamental phase sequence of the branch residual current, and determine the absolute value of the difference between the two as the phase offset of the residual current in this load stage.

[0033] As a further description of the above technical solution: the method for obtaining the leakage residual component includes:

[0034] Based on the leakage current dynamic feature sequence D, each leakage current dynamic feature is read. The corresponding branch identifier β, load phase number r, and load phase type ;

[0035] Based on branch identifier β and load phase type Extract historical normal operation phases that belong to the same branch and load phase type as the current load phase from the electricity consumption behavior state map;

[0036] Based on the leakage current dynamic characteristics corresponding to the historical normal operation phase, the explainable disturbance range of the branch under the load phase type is established.

[0037] Current leakage current dynamic characteristics The k-th type eigenvalue in the eigenvalue is denoted as and will By comparing with the corresponding interpretable disturbance range, the k-th type of leakage current anomaly deviation is generated. The leakage current anomaly deviation is then normalized to obtain the k-th type of leakage current residual component.

[0038] As a further description of the above technical solution: the method for calculating continuous cumulative intensity includes:

[0039] For the β-th branch in the r-th load stage, the formula for calculating the continuous cumulative intensity is:

[0040] ;

[0041] in, This represents the continuous cumulative intensity of the load in the r-th stage of the β-th branch; This represents the continuous cumulative intensity of the load in the (r-1)th stage of the βth branch; This represents the residual leakage current strength of the β-th branch during the r-th load phase; This represents the historical cumulative retention coefficient.

[0042] As a further description of the above technical solution: the method for determining the leakage risk level of the corresponding branch includes:

[0043] when ≤ and ≤ When the leakage risk level of the βth branch is determined to be normal;

[0044] when ≤ and > When the leakage risk level of the βth branch is determined to be of concern level;

[0045] when > and ≤ When the leakage risk level of the βth branch is determined to be of concern level;

[0046] when > and > When the leakage risk level of the βth branch is determined to be the warning level;

[0047] in, This represents the continuous cumulative intensity threshold of the β-th branch; This represents the stage regression bias threshold for the β-th branch; This represents the continuous cumulative intensity of the load in the r-th stage of the β-th branch; This represents the stage regression bias of the β-th branch at the r-th load stage.

[0048] A leakage current early warning system based on electricity consumption behavior analysis is used to implement the aforementioned leakage current early warning method based on electricity consumption behavior analysis. The system includes:

[0049] The original sequence construction module collects the branch power consumption characteristic parameters of each branch according to a unified clock, obtains the load power transition based on the branch power consumption characteristic parameters, and generates an original sequence of power consumption behavior with branch identifier, sampling time identifier, branch power consumption characteristic parameters and load power transition.

[0050] The state map construction module identifies the load stage corresponding to each branch based on the original sequence of electricity consumption behavior and constructs an electricity consumption behavior state map that reflects the changing relationship of the load stages of each branch.

[0051] The feature sequence construction module extracts the amplitude drift, phase shift, harmonic migration, and peak duration of the residual current at each stage based on the power consumption behavior state map, and generates a dynamic feature sequence of leakage current.

[0052] The residual sequence construction module removes interpretable disturbances caused by load phase switching based on the leakage current dynamic feature sequence, obtains leakage current residual components, calculates leakage current residual intensity based on leakage current residual components, and arranges the leakage current residual components and leakage current residual intensity corresponding to each load phase in chronological order to generate leakage current residual sequence.

[0053] The risk level generation module calculates the continuous cumulative intensity and stage regression bias based on the leakage residual sequence, generates the leakage risk evolution trajectory, and determines the leakage risk level of the corresponding branch.

[0054] The beneficial effects of this application are: it can combine the changes in the electricity consumption behavior of each branch to analyze the dynamic characteristics of the residual current in different load stages, eliminate the explainable disturbances caused by the switching of normal load stages, retain the leakage residual that cannot be explained by the electricity consumption behavior state, and further determine the leakage risk level based on the continuous accumulation and stage regression of the leakage residual, thereby improving the accuracy and advance warning of leakage in complex electricity consumption scenarios. Attached Figure Description

[0055] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0056] Figure 1 A flowchart of the leakage current early warning method based on electricity consumption behavior analysis provided in Embodiment 1 of the present invention;

[0057] Figure 2 A flowchart of the method for obtaining the amplitude drift of the residual current provided in Embodiment 1 of the present invention;

[0058] Figure 3 This is a flowchart of a method for obtaining the phase offset of the residual current provided in Embodiment 1 of the present invention;

[0059] Figure 4 This is a module connection diagram of the leakage current early warning system based on electricity consumption behavior analysis provided in Embodiment 2 of the present invention. Detailed Implementation

[0060] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0061] Example 1

[0062] Please see Figures 1-3 This invention provides a technical solution: a leakage current early warning method based on electricity consumption behavior analysis, comprising:

[0063] S1. Collect the branch power consumption characteristic parameters of each branch according to a unified clock, obtain the load power transition based on the branch power consumption characteristic parameters, and generate the original sequence of power consumption behavior with branch identifier, sampling time identifier, branch power consumption characteristic parameters and load power transition.

[0064] The branch circuit power consumption characteristic parameters include branch circuit voltage, branch circuit current, branch circuit residual current, and switch status.

[0065] It should be noted that the voltage sampling device, current sampling device, residual current sampling device, and switch status sampling device corresponding to each branch are synchronously triggered by the clock trigger signal output by the same sampling controller, or the sampling time is corrected separately based on the same time synchronization signal. Each sampling result is written to the corresponding branch identifier and sampling time identifier. The branch voltage, branch current, and branch residual current are instantaneous sampled values ​​obtained at the same sampling time, and the unified sampling period is denoted as Δt. The unified sampling frequency is higher than twice the highest harmonic frequency to be analyzed, and multiple consecutive sampling points are included within one power frequency cycle for subsequent fundamental phase extraction and harmonic decomposition.

[0066] In some implementation methods, the implementation steps include:

[0067] Let the branch voltage of the β-th branch at the nth sampling time be denoted as... Let the branch current of the β-th branch at the nth sampling time be denoted as Let the residual current of the β-th branch at the nth sampling time be denoted as Let the switching state of the β-th branch at the nth sampling time be denoted as ,in =1 indicates that the corresponding branch is in a closed state. =0 indicates that the corresponding branch is in an open state;

[0068] The method for obtaining the load power transition includes:

[0069] Branch voltage based on the same sampling time and branch current Calculate the intermediate power value of the β-th branch at the nth sampling time. For the β-th branch, the number of sampling points NF contained in one power frequency cycle is used as the power calculation window. The average value of the product of the branch voltage and the branch current of each sampling point in the window is determined as the intermediate power value. The difference between the intermediate power value corresponding to the current power calculation window and the intermediate power value corresponding to the previous power frequency cycle is determined as the load power transition amount.

[0070] ;

[0071] in, This represents the load power transition of the β-th branch at the nth sampling time; This represents the intermediate power value of the β-th branch at the nth sampling time; This represents the intermediate power value of the β-th branch at the (n-1)th sampling time.

[0072] The raw data of the electricity consumption behavior of the β-th branch at the nth sampling time is denoted as:

[0073] ;

[0074] in, This represents the raw data of the power consumption behavior of the β-th branch at the nth sampling time; β represents the branch identifier. This represents the nth sampling time. Indicates the branch voltage; Indicates the branch current; Indicates the residual current in the branch; Indicates the switch status; Indicates the load power transition;

[0075] Arrange the raw power consumption data of the β-th branch at N consecutive sampling times according to the chronological order of the sampling times to obtain the raw power consumption sequence of the β-th branch. :

[0076]

[0077] in, This represents the original sequence of electricity consumption behavior in the β-th branch; This represents the total number of sampling times.

[0078] It should be noted that the method for obtaining the branch circuit power consumption characteristic parameters includes: setting up voltage sampling devices, current sampling devices, residual current sampling devices, and switch status sampling devices on each branch circuit of the distribution box. The voltage sampling devices employ voltage transformers, isolated voltage sensors, or resistive voltage divider sampling circuits to collect the branch voltage of the corresponding branch; the current sampling devices employ current transformers, Hall effect current sensors, or shunt sampling resistors to collect the branch current of the corresponding branch; the residual current sampling devices employ zero-sequence current transformers sleeved outside the phase and neutral lines of the corresponding branch to collect the residual current of the branch caused by the imbalance between the phase and neutral currents; the switch status sampling devices employ circuit breaker auxiliary contacts, relay auxiliary contacts, or optocoupler isolated input circuits to collect the switch closed and open states of the corresponding branch. Thus, at each sampling moment, the branch voltage, branch current, branch residual current, and branch switch status of the same branch are obtained.

[0079] In this embodiment, the power consumption data of each branch can be analyzed under a unified time reference, avoiding misalignment in load stage identification and residual current feature extraction due to inconsistent sampling times, and improving the data consistency of subsequent leakage current early warning analysis.

[0080] S2. Based on the original sequence of electricity consumption behavior, identify the load stage corresponding to each branch and construct an electricity consumption behavior state map that reflects the relationship between the load stages of each branch.

[0081] The load phase includes the load start-up phase, the load shutdown phase, the load power adjustment phase, and the load charging phase.

[0082] In some implementation methods, the implementation steps include:

[0083] Methods for identifying the load phase corresponding to each branch include:

[0084] Original sequence of electricity consumption behavior based on the β-th branch Branch currents were read in chronological order of sampling time. Switch status and load power transition The load stage is identified based on changes in switch status, branch current, and load power transition.

[0085] Branch current Switch status and load power transition Input the pre-trained sampling point type recognition model and output the sampling point type of the β-th branch at the n-th sampling time; then merge the sampling times that belong to the same sampling point type to obtain the load start-up stage, load shutdown stage, load power adjustment stage and load charging stage.

[0086] The sampling point types include load start sampling point, load stop sampling point, load power adjustment sampling point, load charging sampling point, and phase hold sampling point;

[0087] Specifically, the training methods for the sampling point type recognition model include:

[0088] Pre-collected time-series sampling training data under multiple branches and operating conditions. Each data set includes branch current, switch status, and load power transition. Simultaneously, the actual type labels of the corresponding sampling points are manually calibrated based on the on-site operating conditions. The sampling points for load start-up, load shutdown, load power adjustment, load charging, and phase hold are sequentially numerically encoded. The input feature vector is constructed by combining branch current, switch status, and load power transition, and the encoded sampling point type labels are used as the model training target.

[0089] The sampling point type identification model can be implemented using gradient boosting tree classification model, random forest classification model, or support vector machine classification model.

[0090] When using a gradient boosting tree classification model, branch current, switch state, and load power transitions constitute the input features. Manually calibrated sampling point types serve as training labels. A multi-class loss function is used to calculate the deviation between the predicted and true types, and decision trees are generated sequentially in the direction that decreases the loss function. The model outputs the class probabilities corresponding to the five types of sampling points and determines the sampling point type with the highest class probability as the sampling point type at the current sampling time. The training data is divided into training, validation, and test sets. The number of decision trees, tree depth, and learning rate are selected based on the classification results of the validation set, and the recognition results of each type of sampling point are verified using the test set. It should be noted that the focus of this application's improvement is not on the sampling point classification model itself, but on using its output sampling point types to form load stages and combining these load stages with subsequent leakage current dynamic features, leakage current residuals, and risk evolution processing chains.

[0091] It should be noted that during manual calibration, the actual type is determined based on the load control record and switch status record: when the switch changes from open to closed and the branch current moves from the zero current range to the non-zero current range, it is marked as a load start sampling point; when the switch changes from closed to open and the branch current falls back to the zero current range, it is marked as a load stop sampling point; when the switch remains closed and the power switches from one non-zero stable range to another, it is marked as a load power adjustment sampling point; when the charging control command is valid, the switch remains closed, and the power changes towards the charging stable value, it is marked as a load charging sampling point; the rest are marked as stage hold sampling points.

[0092] The method for merging consecutive sampling moments belonging to the same sampling point type to obtain the load startup phase, load shutdown phase, load power adjustment phase, and load charging phase includes:

[0093] After continuous confirmation, the sampling point types are merged according to time order. The sampling times of continuous load start sampling points in the β-th branch are merged into the load start stage; the sampling times of continuous load stop sampling points are merged into the load stop stage; the sampling times of continuous load power adjustment sampling points are merged into the load power adjustment stage; and the sampling times of continuous load charging sampling points are merged into the load charging stage.

[0094] Among them, continuous confirmation refers to the same sampling point type in continuous Maintain consistency across all sampling times. Take at least one number of sampling points within a power frequency cycle. The power frequency cycle is the time required for the fundamental voltage or fundamental current of the AC power supply system to complete a full sine wave cycle.

[0095] For a phase-hold sampling point, it is merged into the preceding load phase that is adjacent in time; if there is no confirmed load phase before the phase-hold sampling point, no load phase node is generated until a load start sampling point, load stop sampling point, load power adjustment sampling point, or load charging sampling point is subsequently identified.

[0096] Methods for constructing electricity consumption behavior state maps include:

[0097] The sampling times of consecutive load stages within the same branch are merged into one load stage node. For two adjacent load stage nodes within the same branch, a state switching relationship is established. All load stage nodes and their state switching relationships constitute an electricity consumption behavior state map. The switching relationship includes the switching direction, switching time, and switching interval of switching from the previous load stage type to the next load stage type.

[0098] Specifically, the branch power consumption behavior state diagram of the β-th branch is represented as follows:

[0099] Here, the r-th load stage node of the β-th branch is denoted as:

[0100]

[0101] in, This represents the r-th load stage node of the β-th branch; β represents the branch identifier; r represents the load stage number. This indicates the load phase type, which includes load start-up phase, load shutdown phase, load power adjustment phase, and load charging phase. Indicates the start time of the corresponding load phase; This indicates the end time of the corresponding load phase.

[0102] The state transition relationship expression is:

[0103]

[0104] in, This represents the state transition relationship between the r-th load stage node and the (r+1)-th load stage node in the β-th branch; β represents the branch identifier; and r represents the state transition relationship number. Indicates the node of the previous load phase; Indicates the node of the next load phase; This indicates the switching direction from the previous load phase type to the next load phase type; Indicates the moment of state transition; Indicates the state transition interval.

[0105] Among them, the next load phase node Phase start time This is determined to be the state transition moment; among which, = - .

[0106] In this embodiment, the load start-up phase, load shutdown phase, load power adjustment phase, and load charging phase of each branch are identified based on the original sequence of power consumption behavior. Sampling times belonging to the same load phase consecutively are merged into load phase nodes. Furthermore, the state switching relationship between adjacent load phase nodes is established, constructing a power consumption behavior state map. In this way, the system can obtain the sequential relationship of the load phases of each branch over time, enabling subsequent leakage current analysis to no longer observe residual current changes in isolation, but rather to determine whether abnormal residual current has a reasonable source of power consumption behavior in conjunction with specific load phases.

[0107] S3. Based on the electricity consumption behavior state spectrum, extract the amplitude drift, phase shift, harmonic migration, and peak duration of the residual current at each stage to generate a leakage current dynamic characteristic sequence; in some implementations, the steps include:

[0108] Based on the electricity consumption behavior state map, the r-th load stage node of the β-th branch is read sequentially. According to the start time of the phase and the end of the phase The original sequence of electricity consumption behavior from the β-th branch Extract the branch current data and branch residual current within the corresponding time interval to obtain the branch current stage sequence and branch residual current stage sequence corresponding to the load stage.

[0109] The sequence of residual current stages in the branch during this load phase is represented as follows:

[0110]

[0111] After obtaining the branch residual current stage sequence, a continuous phase is selected at the starting point of the r-th load stage. Using a sampling point as the front-end analysis window, a continuous sampling point is selected at the end of this load phase. Each sampling point serves as the backend analysis window;

[0112] The length of both the front-end analysis window and the back-end analysis window is the number of sampling points corresponding to one power frequency cycle.

[0113] Calculate the average absolute value of the residual current in the branch within the front-end analysis window and the average absolute value of the residual current in the branch within the back-end analysis window, and determine the absolute value of the difference between the two as the residual current amplitude drift for that load stage.

[0114] The fundamental component of the branch current phase sequence within the r-th load stage is extracted to obtain the fundamental phase sequence of the branch current; the fundamental component of the branch residual current phase sequence within the r-th load stage is also extracted to obtain the fundamental phase sequence of the branch residual current. The average value of the fundamental phase sequence of the branch current and the average value of the fundamental phase sequence of the branch residual current are calculated respectively, and the absolute value of the difference between the two is determined as the phase offset of the residual current in that load stage.

[0115] It should be noted that existing fundamental phase extraction methods can be used to extract the fundamental component. These methods can be implemented using Fourier analysis, phase-locked loop analysis, or power frequency periodic synchronization analysis. These are conventional processing methods in electrical signal phase analysis and are existing technologies, so they will not be elaborated on here.

[0116] Harmonic decomposition is performed on the residual current of the branch in the front-end analysis window and the residual current of the branch in the back-end analysis window for the load stage to obtain the harmonic energy of each harmonic in the front-end analysis window and the harmonic energy of each harmonic in the back-end analysis window. The harmonic centroid of the front-end analysis window and the harmonic centroid of the back-end analysis window are calculated based on the harmonic energy of each harmonic, and the absolute value of the difference between the two is determined as the harmonic migration of the residual current in the load stage.

[0117] It should be noted that existing current harmonic analysis methods can be used to process the branch residual current sampling sequences in the front-end analysis window and the branch residual current sampling sequences in the back-end analysis window respectively to obtain the harmonic amplitudes corresponding to each harmonic. The current harmonic analysis method can be implemented using Fast Fourier Transform, Discrete Fourier Transform, or power frequency periodic synchronous harmonic analysis, which are existing technologies and will not be elaborated upon here.

[0118] Specifically, the formula for calculating the harmonic migration of the residual current is: ;

[0119] in, This represents the residual current harmonic migration in the r-th load stage of the β-th branch; This indicates the centroid of the residual current harmonics within the front-end analysis window; This indicates the centroid of the residual current harmonics within the backend analysis window.

[0120] Residual current harmonic centroid in the front-end analysis window The calculation formula is: ;

[0121] Residual current harmonic centroid in the back-end analysis window The calculation formula is: ;

[0122] Where h represents the harmonic order; This represents the h-th harmonic energy within the front-end analysis window of the r-th load stage of the β-th branch. This represents the h-th harmonic energy within the back-end analysis window of the r-th load stage of the β-th branch; This indicates that the summation of the harmonics involved in the analysis is performed. When the sum of the energies of the harmonics involved in the analysis is 0, the harmonic centroid of this window is set to 0.

[0123] The residual current harmonic migration is used to characterize the degree to which the residual current spectrum distribution migrates from lower harmonics to higher harmonics or from higher harmonics to lower harmonics within the same load phase.

[0124] The average value of the residual current of the branch within the front-end analysis window is used as the reference value of the residual current at the beginning of the load stage. The standard deviation of the residual current of the branch within the front-end analysis window is used as the residual current fluctuation value at the beginning of the load stage. The sum of the residual current reference value and three times the residual current fluctuation value is determined as the residual current peak identification threshold for the load stage.

[0125] The absolute value of the residual current in the branch during the load phase is traversed. When it is not less than the residual current peak identification threshold, it is determined as a peak sampling point. The number of peak sampling points is counted, and the peak duration of the residual current is obtained based on the number of peak sampling points and the uniform sampling period.

[0126] Specifically, ;in, This represents the duration of the residual current spike during the r-th load phase of the β-th branch; This indicates the number of sampling points where the residual current of the branch reaches the residual current spike identification threshold during this load phase. This indicates a uniform sampling period. The duration of residual current spikes is used to characterize the length of time that abnormal residual current spikes persist within this load phase.

[0127] The method for generating a leakage current dynamic feature sequence includes: combining the acquired amplitude drift, phase shift, harmonic migration, and peak duration with the branch identifier, load stage number, and load stage type to generate leakage current dynamic features; sorting the leakage current dynamic features corresponding to all branches and all load stages according to the chronological order of the load stage nodes in the power consumption behavior state graph to generate a leakage current dynamic feature sequence.

[0128] Specifically, the dynamic characteristics of leakage current are expressed as follows: ;in, This represents the dynamic characteristics of leakage current in the β-th branch during the r-th load phase; Indicates the corresponding load phase type; This represents the phase offset of the residual current in the r-th load stage of the β-th branch.

[0129] In this embodiment, based on the load stage time interval defined by the electricity consumption behavior state spectrum, the amplitude drift, phase shift, harmonic migration, and peak duration of the residual current in each load stage are extracted to generate a dynamic feature sequence of leakage current. Therefore, this invention not only focuses on whether the residual current exceeds a fixed threshold, but also characterizes the dynamic changes of the residual current in different load stages from multiple dimensions such as amplitude, phase, spectrum, and duration, improving the ability to characterize early leakage current, latent leakage current, and staged leakage current anomalies.

[0130] S4. Based on the leakage current dynamic characteristic sequence, explainable disturbances caused by load phase switching are eliminated to obtain leakage current residual components. The leakage current residual intensity is calculated based on these components. The leakage current residual components and residual intensities corresponding to each load phase are arranged in chronological order to generate a leakage current residual sequence. In some implementations, the steps include:

[0131] Based on the leakage current dynamic feature sequence D, each leakage current dynamic feature is read. The corresponding branch identifier β, load phase number r, and load phase type .

[0132] Based on branch identifier β and load phase type In the electricity consumption behavior state map, historical normal operation phases belonging to the same branch and load phase type as the current load phase are extracted. These historical normal operation phases refer to load phases in which no leakage current alarms, no manually confirmed leakage current faults, and insulation test results are qualified in historical operation data. Based on the leakage current dynamic characteristics corresponding to these historical normal operation phases, the interpretable disturbance range of the branch under this load phase type is established.

[0133] Specifically, the dynamic characteristics of leakage current are divided into four categories: the amplitude drift A of the residual current, the phase shift of the residual current, and so on. The harmonic shift of the residual current H and the peak duration of the residual current Q.

[0134] For the dynamic characteristics of leakage current of type k, k takes the value of A, H or Q, read the k-th type of leakage current dynamic characteristic value corresponding to multiple historical normal operation stages under load stage type s of the β-th branch, and calculate the historical normal average value and historical normal standard deviation of the k-th type of leakage current dynamic characteristic.

[0135] Based on the historical normal average value and the historical normal standard deviation, the explainable disturbance range of the k-th type of leakage current dynamic characteristics of the β-th branch under load phase type s is determined. The explainable disturbance range is used to characterize the range of residual current characteristic changes caused by normal load phase switching under the same branch and the same load phase type.

[0136] The explainable perturbation range includes a lower limit and an upper limit, which are determined as follows:

[0137] ;

[0138] ;

[0139] in This represents the lower limit of the explainable disturbance for the dynamic characteristics of leakage current of the β-th branch under load phase type s; This represents the upper limit of the explainable disturbance for the dynamic characteristics of leakage current of the β-th branch under load phase type s; This represents the historical normal average value of the dynamic leakage current characteristics of the β-th branch under the load phase type s; The historical normal standard deviation of the dynamic leakage current characteristic of the β-th branch under the load phase type s represents the type of leakage current dynamic characteristic of the k-th branch; s represents the load phase type; k represents the leakage current dynamic characteristic type.

[0140] Current leakage current dynamic characteristics The k-th type eigenvalue in the eigenvalue is denoted as and will By comparing with the corresponding interpretable disturbance range, the k-th type of leakage current anomaly deviation is generated;

[0141] Specifically, when Not less than and not greater than When the current dynamic characteristic of the k-th type of leakage current is determined to be an interpretable disturbance caused by load phase switching, the residual component of the k-th type of leakage current is recorded as 0.

[0142] when Greater than When the current dynamic characteristic of the k-th type of leakage current exceeds the upper limit of the explainable disturbance, the part exceeding the upper limit is determined as the abnormal deviation of the k-th type of leakage current.

[0143] when Less than When the current dynamic characteristic of the k-th type of leakage current is determined to be lower than the lower limit of the explainable disturbance, the leakage current residual is not included.

[0144] To enable comparison of residual current amplitude drift, residual current phase shift, harmonic migration, and peak duration on the same scale, the leakage current anomaly deviation is normalized to obtain the k-th type of leakage current residual component.

[0145] Specifically, when When within the explainable disturbance range, the type k leakage residual component =0;

[0146] when When the perturbation exceeds the upper limit of explainable disturbance:

[0147] ;

[0148] In the formula, This represents the residual component of leakage current corresponding to the k-th type of leakage current dynamic characteristic in the r-th load stage of the β-th branch; This represents the current characteristic value of the k-th type of leakage dynamic characteristic in the r-th load stage of the β-th branch; It represents the smallest effective scale for the dynamic characteristics of leakage current of type k, used to prevent the residual components from being abnormally amplified due to the historical normal standard deviation being too small.

[0149] Where, when k = A, The residual current amplitude drift The corresponding leakage residual component is denoted as ;

[0150] When k = hour, Residual current phase offset The corresponding leakage residual component is denoted as ;

[0151] When k = hour, Residual current harmonic migration The corresponding leakage residual component is denoted as ;

[0152] When k = hour, Duration of residual current spike The corresponding leakage residual component is denoted as .

[0153] The four types of leakage residual components are weighted and summed according to preset weighting coefficients to obtain the leakage residual intensity of the β-th branch at the r-th load stage. ,

[0154] Specifically,

[0155] in, This represents the residual leakage current strength of the β-th branch during the r-th load phase; This represents the residual leakage component corresponding to the magnitude drift of the residual current. This represents the residual leakage component corresponding to the phase offset of the residual current; This represents the residual leakage current component corresponding to the residual current harmonic migration. This represents the residual leakage component corresponding to the duration of the residual current spike. This represents the weighting coefficient corresponding to the residual current amplitude drift. This represents the weighting coefficient corresponding to the phase offset of the residual current; This represents the weighting coefficient corresponding to the residual current harmonic migration. This represents the weighting coefficient corresponding to the duration of the residual current spike.

[0156] In this embodiment, = = = =0.25, the sum of the four weighting coefficients is 1.

[0157] In this embodiment, an interpretable disturbance range is established based on historical normal operation phases under the same branch and load phase type. The current leakage current dynamic characteristics are compared with the corresponding interpretable disturbance range. Interpretable disturbances caused by normal load phase switching are eliminated, and leakage current residual components exceeding the interpretable disturbance range are retained. In this way, the interference of normal disturbances during load startup, load power adjustment, load charging, and load shutdown on leakage current warning results can be reduced, and the probability of misjudging normal power consumption behavior as leakage current anomaly can be decreased.

[0158] Furthermore, the residual components corresponding to the residual current amplitude drift, phase shift, harmonic migration, and peak duration are fused to obtain the residual current intensity, and a residual current sequence is generated according to the time sequence of the load stage. Thus, the system can convert residual current anomaly characteristics of different types and dimensions into residual expressions that can be used for continuous analysis, transforming leakage risk assessment from single-point exceedance judgment to continuous residual analysis based on the load stage.

[0159] S5. Calculate the continuous cumulative intensity and stage regression bias based on the leakage residual sequence to generate the leakage risk evolution trajectory; in some implementation methods, the steps include:

[0160] Based on the leakage residual sequence, the continuous cumulative intensity of each branch is calculated in the order of the load stages.

[0161] For the β-th branch in the r-th load stage, the formula for calculating the continuous cumulative intensity is:

[0162]

[0163] in, This represents the continuous cumulative intensity of the load in the r-th stage of the β-th branch; This represents the continuous cumulative intensity of the load in the (r-1)th stage of the βth branch; This represents the residual leakage current strength of the β-th branch during the r-th load phase; This represents the historical cumulative retention coefficient, with a value greater than 0 and less than 1. Continuous cumulative intensity is used to characterize the cumulative increase in leakage residual during continuous load periods. When r=1, The historical cumulative retention coefficient α is set to 0 based on the degree of retention of leakage risk across load stages. It can be optionally set to 0.8, and can be set to 0.5 to 0.7 when it is necessary to shorten the historical impact, and to 0.8 to 0.9 when it is necessary to enhance the continuous cumulative memory.

[0164] Based on the residual leakage strength of the β-th branch at the r-th load stage And the residual leakage strength of the adjacent load phase after the load phase. Calculate the regression bias during the calculation phase;

[0165]

[0166] in, This represents the stage regression bias of the (r+1)th load stage of the βth branch; This represents the residual leakage current strength of the (r+1)th load stage of the βth branch; represents the baseline value of the residual regression for the β-th branch; max indicates taking the larger value. It is determined by the average residual leakage current intensity during the historical normal operation phase of the β-th branch. The phase regression bias is used to characterize whether the residual leakage current returns to the normal level after the load phase switching.

[0167] The continuous cumulative intensity and stage regression deviation of the β-th branch at the r-th load stage are combined to form the leakage risk evolution point, which is expressed as:

[0168] ;

[0169] in, This represents the leakage risk evolution point corresponding to the r-th load stage of the β-th branch; the leakage risk evolution points of all branches are arranged in chronological order to generate the leakage risk evolution trajectory.

[0170] S6. Based on the leakage risk evolution trajectory, determine the continuous cumulative intensity threshold and the stage regression deviation threshold for each branch, and compare the continuous cumulative intensity and stage regression deviation of the corresponding branch in the current load stage with the continuous cumulative intensity threshold and the stage regression deviation threshold, respectively, to determine the leakage risk level of the corresponding branch. In some embodiments, the implementation steps include:

[0171] Based on the leakage risk evolution trajectory formed during the historical normal operation phase of the βth branch, the continuous cumulative intensity threshold and the stage regression deviation threshold are determined respectively.

[0172] Specifically, the methods for obtaining the continuous cumulative intensity threshold and the stage regression bias threshold include:

[0173]

[0174]

[0175] in, This represents the continuous cumulative intensity threshold of the β-th branch; This represents the average cumulative intensity during the historical normal operation phase of the β-th branch; This represents the continuous cumulative strength standard deviation during the historical normal operation phase of the β-th branch; This represents the stage regression bias threshold for the β-th branch; This represents the average stage regression bias during the historical normal operation phase of the β-th branch; This represents the standard deviation of the stage regression bias during the historical normal operation phase of the β-th branch.

[0176] The leakage risk levels include normal level, attention level, and warning level.

[0177] when ≤ and ≤ When the leakage risk level of the βth branch is determined to be normal, it means that the leakage residual of the βth branch has not accumulated continuously, and the leakage residual after the load phase ends can return to the normal level.

[0178] when ≤ and > When the leakage risk level of the βth branch is determined to be of concern level, it means that the leakage residual of the βth branch has not yet accumulated continuously, but has failed to return to the normal level after the load phase ends.

[0179] when > and ≤ When the leakage risk level of the βth branch is determined to be of concern level, it means that the leakage residual of the βth branch has accumulated and increased, but it can still return to the normal level after the load phase ends.

[0180] when > and > When the leakage risk level of the β-th branch is determined to be the warning level, it means that the β-th branch has both leakage residual accumulation and stage regression anomaly.

[0181] When the βth branch is determined to be at the warning level, the corresponding branch's early warning signal is output.

[0182] The early warning signal includes the branch identifier β and the time of warning occurrence. Leakage risk level, continuous cumulative intensity and stage regression bias .

[0183] Specifically, the end time of the corresponding load phase is determined as the time when the warning occurs.

[0184] In this embodiment, the continuous cumulative intensity and stage regression deviation are calculated based on the leakage residual sequence to generate the leakage risk evolution trajectory. The continuous cumulative intensity can characterize the degree of enhancement of the leakage residual in the same branch during the continuous load stage, and the stage regression deviation can characterize whether the leakage residual returns to the normal level after the load stage ends. By analyzing the above two indicators at the same time, the present invention can identify the situation where the residual current has not yet reached the traditional protection threshold but the leakage risk has continued to evolve, and realize early warning of hidden leakage risk.

[0185] Example 2

[0186] Please see Figure 4 This invention provides a technical solution: a leakage current early warning system based on electricity consumption behavior analysis, which is used to implement the leakage current early warning method based on electricity consumption behavior analysis. The system includes:

[0187] The original sequence construction module collects the branch power consumption characteristic parameters of each branch according to a unified clock, obtains the load power transition based on the branch power consumption characteristic parameters, and generates an original sequence of power consumption behavior with branch identifier, sampling time identifier, branch power consumption characteristic parameters and load power transition.

[0188] The state map construction module identifies the load stage corresponding to each branch based on the original sequence of electricity consumption behavior and constructs an electricity consumption behavior state map that reflects the changing relationship of the load stages of each branch.

[0189] The feature sequence construction module extracts the amplitude drift, phase shift, harmonic migration, and peak duration of the residual current at each stage based on the power consumption behavior state map, and generates a dynamic feature sequence of leakage current.

[0190] The residual sequence construction module removes interpretable disturbances caused by load phase switching based on the leakage current dynamic feature sequence, obtains leakage current residual components, calculates leakage current residual intensity based on leakage current residual components, and arranges the leakage current residual components and leakage current residual intensity corresponding to each load phase in chronological order to generate leakage current residual sequence.

[0191] The risk level generation module calculates the continuous cumulative intensity and stage regression bias based on the leakage residual sequence, generates the leakage risk evolution trajectory, and determines the leakage risk level of the corresponding branch.

[0192] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for leakage pre-warning based on power consumption behavior analysis, characterized in that, include: The branch power consumption characteristic parameters of each branch are collected according to a unified clock. The load power transition is obtained based on the branch power consumption characteristic parameters, and an original sequence of power consumption behavior with branch identifier, sampling time identifier, branch power consumption characteristic parameters and load power transition is generated. Based on the original sequence of electricity consumption behavior, the load stage corresponding to each branch is identified, and an electricity consumption behavior state map reflecting the changing relationship of the load stages of each branch is constructed. Based on the power consumption behavior state map, the amplitude drift, phase shift, harmonic migration, and peak duration of the residual current at each stage are extracted to generate a dynamic feature sequence of leakage current. Based on the leakage current dynamic feature sequence, explainable disturbances caused by load phase switching are eliminated, leakage current residual components are obtained, leakage current residual intensity is calculated based on the leakage current residual components, and the leakage current residual components and leakage current residual intensity corresponding to each load phase are arranged in time order to generate leakage current residual sequence. Based on the leakage residual sequence, the continuous cumulative intensity and stage regression bias are calculated to generate the leakage risk evolution trajectory and determine the leakage risk level of the corresponding branch.

2. The method for leakage pre-warning based on power utilization behavior analysis according to claim 1, characterized in that, The branch circuit power consumption characteristic parameters include branch circuit voltage, branch circuit current, branch circuit residual current, and switch status.

3. The method for leakage pre-warning based on power utilization behavior analysis according to claim 2, characterized in that, The method for obtaining the load power transition includes: Branch voltage based on the same sampling time and branch current Calculate the intermediate power value of the β-th branch at the nth sampling time, and calculate the load power transition based on the intermediate power values ​​of adjacent sampling times.

4. The leakage current early warning method based on electricity consumption behavior analysis according to claim 1, characterized in that, The load phase includes the load start-up phase, the load shutdown phase, the load power adjustment phase, and the load charging phase. Methods for identifying the load phase corresponding to each branch include: Original sequence of electricity consumption behavior based on the β-th branch Branch currents were read in chronological order of sampling time. Switch status and load power transition Input the pre-trained sampling point type recognition model and output the sampling point type of the β-th branch at the nth sampling time; By merging consecutive sampling times belonging to the same sampling point type, we obtain the load start-up stage, load shutdown stage, load power adjustment stage, and load charging stage.

5. The leakage current early warning method based on electricity consumption behavior analysis according to claim 4, characterized in that, The sampling point types include load start sampling point, load stop sampling point, load power adjustment sampling point, load charging sampling point, and phase hold sampling point.

6. The leakage current early warning method based on electricity consumption behavior analysis according to claim 5, characterized in that, The method for merging consecutive sampling moments belonging to the same sampling point type to obtain the load startup phase, load shutdown phase, load power adjustment phase, and load charging phase includes: After continuous confirmation, the sampling point types are merged according to time order. The sampling times of consecutive load start sampling points in the β-th branch are merged into the load start stage; the sampling times of consecutive load stop sampling points are merged into the load stop stage; the sampling times of consecutive load power adjustment sampling points are merged into the load power adjustment stage; and the sampling times of consecutive load charging sampling points are merged into the load charging stage. For a phase-hold sampling point, it is merged into the preceding load phase that is adjacent in time; if there is no confirmed load phase before the phase-hold sampling point, no load phase node is generated until a load start sampling point, load stop sampling point, load power adjustment sampling point, or load charging sampling point is subsequently identified.

7. The leakage current early warning method based on electricity consumption behavior analysis according to claim 1, characterized in that, Methods for constructing electricity consumption behavior state maps include: The sampling times of consecutive load stages within the same branch are merged into one load stage node. For two adjacent load stage nodes within the same branch, a state switching relationship is established. All load stage nodes and their state switching relationships constitute an electricity consumption behavior state map.

8. The leakage current early warning method based on electricity consumption behavior analysis according to claim 1, characterized in that, The method for obtaining the amplitude drift of the residual current includes: Based on the electricity consumption behavior state map, the r-th load stage node of the β-th branch is read sequentially. According to the start time of the phase and the end of the phase The original sequence of electricity consumption behavior from the β-th branch Extract the residual current of the branch within the corresponding time interval to obtain the residual current stage sequence of the branch corresponding to the load stage; Select a continuous load at the beginning of the r-th load phase. Using 1 sampling point as the front-end analysis window, a continuous sampling point is selected at the end of the r-th load stage. Each sampling point serves as the backend analysis window; Calculate the average absolute value of the residual current in the branch within the front-end analysis window and the average absolute value of the residual current in the branch within the back-end analysis window, and determine the absolute value of the difference between the two as the residual current amplitude drift for that load stage.

9. The leakage current early warning method based on electricity consumption behavior analysis according to claim 8, characterized in that, Methods for obtaining the phase offset of the residual current include: Based on the electricity consumption behavior state map, the r-th load stage node of the β-th branch is read sequentially. According to the start time of the phase and the end of the phase The original sequence of electricity consumption behavior from the β-th branch Extract the branch current within the corresponding time interval to obtain the branch current stage sequence corresponding to the load stage; The fundamental component of the branch current phase sequence within the r-th load phase is extracted to obtain the fundamental phase sequence of the branch current; the fundamental component of the residual current phase sequence within the r-th load phase is extracted to obtain the fundamental phase sequence of the residual current. Calculate the average value of the fundamental phase sequence of the branch current and the average value of the fundamental phase sequence of the branch residual current, and determine the absolute value of the difference between the two as the phase offset of the residual current in this load stage.

10. The leakage current early warning method based on electricity consumption behavior analysis according to claim 1, characterized in that, The method for obtaining the leakage residual component includes: Based on the leakage current dynamic feature sequence D, each leakage current dynamic feature is read. The corresponding branch identifier β, load phase number r, and load phase type ; Based on branch identifier β and load phase type Extract historical normal operation phases that belong to the same branch and load phase type as the current load phase from the electricity consumption behavior state map; Based on the leakage current dynamic characteristics corresponding to the historical normal operation phase, the explainable disturbance range of the branch under the load phase type is established. Current leakage current dynamic characteristics The k-th type eigenvalue in the eigenvalue is denoted as and will By comparing with the corresponding interpretable disturbance range, the k-th type of leakage current anomaly deviation is generated. The leakage current anomaly deviation is then normalized to obtain the k-th type of leakage current residual component.

11. The leakage current early warning method based on electricity consumption behavior analysis according to claim 1, characterized in that, The method for calculating continuous cumulative intensity includes: For the β-th branch in the r-th load stage, the formula for calculating the continuous cumulative intensity is: in, This represents the continuous cumulative intensity of the load in the r-th stage of the β-th branch; This represents the continuous cumulative intensity of the load in the (r-1)th stage of the βth branch; This represents the residual leakage current strength of the β-th branch during the r-th load phase; This represents the historical cumulative retention coefficient.

12. The leakage current early warning method based on electricity consumption behavior analysis according to claim 1, characterized in that, The method for determining the leakage risk level of the corresponding branch includes: when ≤ and ≤ When the leakage risk level of the βth branch is determined to be normal; when ≤ and > When the leakage risk level of the βth branch is determined to be of concern level; when > and ≤ When the leakage risk level of the βth branch is determined to be of concern level; when > and > When the leakage risk level of the βth branch is determined to be the warning level; in, This represents the continuous cumulative intensity threshold of the β-th branch; This represents the stage regression bias threshold for the β-th branch; This represents the continuous cumulative intensity of the load in the r-th stage of the β-th branch; This represents the stage regression bias of the β-th branch at the r-th load stage.

13. A leakage current early warning system based on electricity consumption behavior analysis, used to implement the leakage current early warning method based on electricity consumption behavior analysis as described in any one of claims 1-12, characterized in that, The system includes: The original sequence construction module collects the branch power consumption characteristic parameters of each branch according to a unified clock, obtains the load power transition based on the branch power consumption characteristic parameters, and generates an original sequence of power consumption behavior with branch identifier, sampling time identifier, branch power consumption characteristic parameters and load power transition. The state map construction module identifies the load stage corresponding to each branch based on the original sequence of electricity consumption behavior and constructs an electricity consumption behavior state map that reflects the changing relationship of the load stages of each branch. The feature sequence construction module extracts the amplitude drift, phase shift, harmonic migration, and peak duration of the residual current at each stage based on the power consumption behavior state map, and generates a dynamic feature sequence of leakage current. The residual sequence construction module removes interpretable disturbances caused by load phase switching based on the leakage current dynamic feature sequence, obtains leakage current residual components, calculates leakage current residual intensity based on leakage current residual components, and arranges the leakage current residual components and leakage current residual intensity corresponding to each load phase in chronological order to generate leakage current residual sequence. The risk level generation module calculates the continuous cumulative intensity and stage regression bias based on the leakage residual sequence, generates the leakage risk evolution trajectory, and determines the leakage risk level of the corresponding branch.

Images