Intelligent leakage current real-time monitoring and diagnosis system and method

Abstract

The present application relates to the technical field of measuring electric variable, in particular to a kind of intelligent leakage current real-time monitoring and diagnosis system and method, including the real-time current signal of target area power grid and corresponding current signal time stamp, real-time electric signal is intelligently handled to obtain corresponding current signal sequence;Current signal sequence is carried out wavelet packet transform and intelligent processing, obtain current signal subsequence;For each current signal subsequence dynamic window sequence set, dynamic window sequence set is intelligently analyzed, and the steady-state characteristic and transient characteristic of current signal are obtained;Establish current diagnosis model, and based on current diagnosis model, real-time current is diagnosed and early warning.The present application solves the situation that single frequency band is decomposed, it is difficult to accurately distinguish high-frequency transient noise and low-frequency steady-state anomaly, leading to missed detection or false alarm, and threshold is adjusted according to actual risk duration, real-time dynamic prediction is realized, and the health management level of regional power grid is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of measuring electrical variables, in particular to an intelligent leakage current real-time monitoring and diagnosis system and method. BACKGROUND

[0002] With the accelerated development of electrification, the number of electrical equipment connected to the industrial area power grid also increases sharply, resulting in an increase in the risk of leakage. Therefore, it is necessary to monitor and diagnose the current of the power grid to prevent leakage from occurring. Leakage monitoring and diagnosis is the "first line of defense" for the safe operation of the industrial area power grid, and is also a key technology for improving energy efficiency, reducing operation and maintenance costs, and promoting the level of health management of the industrial area power grid.

[0003] In the existing technology, the traditional method adopts single-band decomposition, which is difficult to accurately distinguish high-frequency transient noise and low-frequency steady-state anomalies, resulting in missed detection or false positives. The traditional system relies on fixed thresholds or static models and cannot capture the dynamic changes of the current signal in real time, resulting in delayed early warning or poor adaptability. The environmental noise caused by the start and stop of electrical equipment is easily confused with the leakage signal, and the traditional method is difficult to distinguish, which is not good for the health management level of the regional power grid. SUMMARY

[0004] The present application aims to solve the problems in the background art and provides an intelligent leakage current real-time monitoring and diagnosis system and method.

[0005] The technical solution of the present application is an intelligent leakage current real-time monitoring and diagnosis method, comprising the following methods:

[0006] Obtain the real-time current signal of the target area power grid and the corresponding current signal timestamp, and intelligently process the real-time current signal to obtain the corresponding current signal sequence;

[0007] Perform wavelet packet transform processing on the current signal sequence to decompose the current signal processing sequence into current signal sub-sequences of different frequency bands, and intelligently process the current signal sub-sequences to obtain the current signal sub-sequences;

[0008] For each current signal sub-sequence, a dynamic window is established, and the current signal sub-sequence located at the center of the dynamic window is marked as the center sequence C. The current signal sub-sequences contained in the dynamic window are taken as the dynamic sequence of the center sequence C, a dynamic window sequence set is established, and intelligent feature analysis is performed on the dynamic window sequence set to obtain the steady-state feature Sav and the transient feature Tsv of the current signal;

[0009] A current diagnosis model is established based on the steady-state analysis value Sav and the transient risk value Tsv, and the real-time current is diagnosed and warned based on the current diagnosis model.

[0010] Preferably, the wavelet packet transform processing of the current signal processing sequence comprises:

[0011] A first decomposition layer number is selected, and the current signal processing sequence is decomposed into a plurality of first current signal subsequences with a first frequency band width based on the first decomposition layer number;

[0012] A second decomposition layer number is selected, and the current signal processing sequence is decomposed into a plurality of second current signal subsequences with a second frequency band width based on the second decomposition layer number;

[0013] The first decomposition layer number is less than the second decomposition layer number.

[0014] Preferably, the method for intelligently analyzing the dynamic window sequence set is:

[0015] The rated current of the historical electrical equipment of the intervention current signal and the corresponding historical current signal sequence are obtained, the historical electrical equipment information and the historical current signal sequence are intelligently analyzed, and the fitting degree of the electrical equipment information to the current signal sequence is determined.

[0016] The method for intelligently analyzing the historical electrical equipment information and the historical current signal sequence is:

[0017] The historical current signal sequence is decomposed based on the first decomposition layer number to obtain historical subsequences, the sampling point current mean of each historical subsequence is obtained, and the current standard deviation of each historical subsequence is calculated based on the current value of the sampling point of each historical subsequence to obtain the dispersion degree of the current of the sampling point of each historical subsequence.

[0018] Preferably, the current standard deviation difference value of adjacent historical subsequences is calculated, a historical subsequence current standard deviation difference value sequence is established, the historical subsequence current standard deviation difference value sequence is traversed, elements in the historical subsequence current standard deviation difference value sequence that are less than a standard deviation value threshold are marked as target elements, and the continuity of a plurality of target elements is identified.

[0019] The elements of the current standard deviation difference value sequence are numbered, and if there are at least m target elements whose numbers are continuous, it is determined that the plurality of target elements with continuous numbers have continuity, and the target element with the smallest number is marked as a selected element.

[0020] The historical subsequence corresponding to the current standard deviation obtained by subtracting the selected element is marked as a target sequence, the historical current signal sequence is segmented at the starting sequence of the target sequence to obtain two segmented sequences, the segmented sequence to which the target sequence belongs is deleted, and the remaining segmented sequence is taken as a noise sequence N.

[0021] Preferably, the noise sequence N is processed:

[0022] The extreme value of each historical sub-sequence in the noise sequence N is obtained, and the average energy value of each historical sub-sequence is obtained by calculating the absolute value and the average value of the extreme value of each historical sub-sequence.

[0023] The fitting degree Gof of the noise sequence N and the rated current of the electrical equipment is calculated by the formula

[0024] In the formula, |E i is the average energy value of each historical sub-sequence; i is the historical sub-sequence number, i = [1, 2, …, n], and n is the total number of historical sub-sequences; is the standard deviation of the average energy value of the noise sequence N; Pc is the rated current of the electrical equipment; t is the sequence length of the noise sequence N; λ1 and λ2 are weight coefficients of the standard deviation of the average energy value and the sequence length, respectively, and are obtained based on historical big data test.

[0025] Preferably, the method for intelligently analyzing the dynamic window sequence set further comprises intelligently processing the first current signal sub-sequence, and the method is as follows:

[0026] The average current and the average electric signal frequency of the center sequence C, and the average current and the average electric signal frequency of the first k acquisition time periods and the last k acquisition time periods of the center sequence C are obtained respectively, and the average current, the rated current and the average electric signal frequency are normalized, and the steady-state analysis value of the dynamic window is calculated based on a steady-state analysis model, and the expression of the steady-state analysis model is:

[0027]

[0028] In the formula, Sav is the steady-state analysis value of the dynamic window; If j and Ff j are the average current and the average electric signal frequency of the first j acquisition time periods of the center sequence C, Ib (k-j) and Fb (k-j) are the average current and the average electric signal frequency of the last k-j acquisition time periods of the center sequence C; j is the acquisition time period number, j = [1, 2, …, k]; Im and Fm are the average current and the average electric signal frequency of the center sequence C.

[0029] Preferably, the method for intelligently analyzing the dynamic window sequence set further comprises intelligently processing the second current signal sub-sequence, and the method is as follows:

[0030] ​The risk sequence of the previous risk period generating the electric leakage protection in the historical current signal sequence is obtained, a plurality of risk sequences are determined, and the instantaneous frequency and the instantaneous current are taken as the first transient parameter and the second transient parameter of the risk sequence; the first transient parameter and the second transient parameter corresponding to all collection points in the risk sequence are obtained, and the first transient parameter set and the second transient parameter set are obtained;

[0031] The same transient parameter set is subjected to parameter analysis, and the method is as follows:

[0032] The mean value of the transient parameters of the last collection point of the plurality of risk sequences is calculated as the risk parameter of the transient parameter set;

[0033] The transient parameters in the same transient parameter set are sorted in ascending order to obtain a parameter sorting sequence, the parameter sorting sequence is subjected to sequence splitting based on the second decomposition layer number to obtain a plurality of parameter sorting sub-sequences with the same number as the second decomposition layer number, and the median value of the parameter sorting sub-sequence is obtained;

[0034] The number of transient parameters contained in each parameter sorting sub-sequence is counted to obtain the total number of sub-sequence transient parameters, and the weighted average parameter in the transient parameter set is calculated based on the parameter sorting sub-sequence median value and the total number of sub-sequence transient parameters by using the weighted average algorithm, and is taken as an abnormal parameter;

[0035] The time point corresponding to the weighted average parameter in the risk sequence is obtained by using the interpolation algorithm and is marked as the target collection point, the rated power of the electrical equipment corresponding to the target collection point is obtained and is marked as the corrected power, and the interval time length from the time point corresponding to the target collection point to the end time point of the risk period is taken as the risk time length of the abnormal parameter corresponding to the risk sequence;

[0036] Based on the above method of parameter analysis on the same transient parameter set, the abnormal parameters and risk time lengths corresponding to the first transient parameter set and the second parameter set are obtained, respectively, and are marked as the first risk parameter, the second risk parameter, the first abnormal parameter, the second abnormal parameter, and the first risk time length and the second risk time length.

[0037] Preferably, the real-time current and the real-time current signal frequency are determined based on the second current signal sub-sequence and are taken as the first real-time parameter and the second real-time parameter, and the corresponding real-time electrical equipment power is obtained; if any real-time parameter reaches the corresponding risk parameter, the actual risk time length of the corresponding real-time parameter is counted, and the first actual risk time length corresponding to the first transient parameter set and the second actual risk time length corresponding to the second transient parameter set are obtained;

[0038] The transient risk value Tsv of the second current signal sub-sequence is calculated by the formula

[0039] In the formula, I t ​is a real-time current; F t is a real-time frequency; I1 is a first abnormal parameter; F1 is a second abnormal parameter; T I is a first actual risk duration; T F is a second actual risk duration; T t is an actual risk duration; min() is a minimum comparison function; P t is a real-time electrical equipment power; P1 is a corrected power.

[0040] Preferably, the method for establishing a current diagnosis model based on a steady-state analysis value Sav and a transient risk value Tsv is as follows:

[0041] The steady-state analysis value Sav of the current first current signal sub-sequence is compared with a preset steady-state threshold value S0;

[0042] If Sav>S0, it is determined that the real-time current has an abnormal steady-state characteristic and a first abnormal label is generated; otherwise, no operation is performed.

[0043] And the transient risk value Tsv of the current second current signal sub-sequence is compared with a preset transient threshold value T0;

[0044] If Tsv≤0.8T0, no operation is performed; if 0.8T0<Tsv≤T0, the current second current signal sub-sequence is marked as an abnormal sub-sequence, and it is determined whether there is at least a continuous number of z abnormal sub-sequences; if there is at least a continuous number of z abnormal sub-sequences, it is determined that the real-time current has a first abnormal transient characteristic and a second abnormal label is generated; if there is no at least a continuous number of z abnormal sub-sequences, no operation is performed.

[0045] If T0<Tsv, it is determined that the real-time current has a second abnormal transient characteristic and a third abnormal label is generated.

[0046] According to the first abnormal label, the second abnormal label and the third abnormal label, it is determined that there is a leakage phenomenon, and a technician is notified to perform maintenance.

[0047] The application also discloses an intelligent leakage current real-time monitoring and diagnosis system, which applies the intelligent leakage current real-time monitoring and diagnosis method.

[0048] A data acquisition module is configured to acquire a real-time current signal of a target area power grid and a corresponding current signal timestamp, and to intelligently process the real-time current signal to acquire a corresponding current signal sequence.

[0049] A data preprocessing module is configured to perform wavelet packet transform processing on the current signal sequence, decompose the current signal processing sequence into current signal sub-sequences of different frequency bands, and intelligently process the current signal sub-sequences to acquire the current signal sub-sequences.

[0050] a data analysis module for establishing a dynamic window for each current signal subsequence, marking the current signal subsequence located at the center of the dynamic window as a center sequence C, taking the current signal subsequences contained in the dynamic window as the center sequence C dynamic sequence, establishing a dynamic window sequence set, intelligently analyzing the dynamic window sequence set to obtain the steady-state characteristic Sav and the transient characteristic Tsv of the current signal;

[0051] an intelligent monitoring and diagnosis module for establishing a current diagnosis model based on the steady-state analysis value Sav and the transient risk value Tsv, diagnosing and warning the real-time current based on the current diagnosis model.

[0052] Compared with the prior art, the above technical solutions of the present application have the following beneficial technical effects:

[0053] (1) By combining the first decomposition layer and the second decomposition layer, the current signal is decomposed into sub-sequences of different frequency bands, the frequency resolution is significantly improved, based on dynamic window sequence analysis, a dynamic window is established for each sub-sequence, the steady-state characteristic and the transient risk value of the center sequence are combined to realize accurate separation of high-frequency transient and low-frequency steady-state signals, solve the problem of single frequency band decomposition, difficult to accurately distinguish high-frequency transient noise and low-frequency steady-state abnormality, resulting in missed detection or false alarm, and improve the health management level of regional power grid;

[0054] (2) Through real-time updating and analysis of the dynamic window sequence set, the average current and frequency of the previous and subsequent k acquisition periods are combined to quickly identify current fluctuations: based on the transient parameters of the historical risk sequence and the weighted average algorithm, the transient risk value is dynamically generated, and the threshold value is adjusted according to the actual risk duration, real-time dynamic prediction is realized, and the problem of delay or poor adaptability caused by relying on fixed threshold and static model is solved;

[0055] (3) According to the threshold comparison of the steady-state analysis value and the transient risk value, a hierarchical abnormality label is generated, and the false positive rate is reduced by combining the continuous z abnormal sub-sequences, realizing multi-dimensional accurate diagnosis of the leakage phenomenon. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The flowchart of the embodiment one of the present application is shown. DETAILED DESCRIPTION

[0057] Embodiment one, as shown in the figure, the present application proposes an intelligent leakage current real-time monitoring and diagnosis method, including the following methods: Figure 1

[0058] Obtain the real-time current signal of the target regional power grid and the corresponding current signal timestamp, intelligently process the real-time current signal to obtain the corresponding current signal sequence; ​

[0059] The wavelet packet transform is performed on the current signal sequence, the current signal processing sequence is decomposed into current signal subsequences of different frequency bands, and the current signal subsequences are intelligently processed to obtain the current signal subsequences;

[0060] A dynamic window is established for each current signal subsequence, and a current signal subsequence located at the center of the dynamic window is marked as a center sequence C. The current signal subsequences contained in the dynamic window are taken as the dynamic sequence of the center sequence C, a dynamic window sequence set is established, and intelligent feature analysis is performed on the dynamic window sequence set to obtain the steady-state feature and the transient feature of the current signal;

[0061] The wavelet packet transform performed on the current signal processing sequence includes:

[0062] A first decomposition layer number is selected, and the current signal processing sequence is decomposed into a plurality of first current signal subsequences with a first frequency band width based on the first decomposition layer number;

[0063] A second decomposition layer number is selected, and the current signal processing sequence is decomposed into a plurality of second current signal subsequences with a second frequency band width based on the second decomposition layer number;

[0064] The first decomposition layer number is less than the second decomposition layer number;

[0065] The method for intelligently analyzing the dynamic window sequence set is:

[0066] The rated current of the historical electrical equipment of the intervention current signal and the corresponding historical current signal sequence are obtained, and intelligent analysis is performed on the historical electrical equipment information and the historical current signal sequence to determine the fitting degree of the electrical equipment information on the current signal sequence;

[0067] The method for intelligently analyzing the historical electrical equipment information and the historical current signal sequence is:

[0068] The historical current signal sequence is decomposed based on the first decomposition layer number to obtain historical subsequences, the current mean value of each historical subsequence is obtained, and the current standard deviation of each historical subsequence is calculated based on the current value of each sampling point of the historical subsequence to obtain the dispersion degree of the current of each sampling point of the historical subsequence;

[0069] The current standard deviation difference values of adjacent historical subsequences are calculated, a historical subsequence current standard deviation difference value sequence is established, the historical subsequence current standard deviation difference value sequence is traversed, elements in the historical subsequence current standard deviation difference value sequence that are less than a standard deviation value threshold are marked as target elements, and the continuity of a plurality of target elements is identified;

[0070] The elements of the current standard deviation difference sequence are numbered, and if there are at least m consecutive target element numbers, it is determined that the consecutive target elements have continuity, and the target element with the smallest number is marked as a selected element;

[0071] The historical subsequence corresponding to the current standard deviation of the selected element and as the deduction is marked as the target sequence, and the historical current signal sequence is segmented at the starting sequence of the target sequence to obtain two segmentation sequences. The segmentation sequence to which the target sequence belongs is deleted, and the remaining segmentation sequence is taken as a noise sequence N;

[0072] The noise sequence N is processed:

[0073] The extreme value of each historical subsequence in the noise sequence N is obtained, and the absolute value and mean value of each historical subsequence extreme value are calculated to obtain the historical subsequence average energy value. It should be noted that the extreme value of the historical subsequence is the maximum and minimum value in a short sequence length, and the extreme value of the historical subsequence can be multiple;

[0074] The fitting degree Gof of the noise sequence N and the rated current of the electrical equipment is calculated by the formula

[0075] In the formula, |E i | is the average energy value of each historical subsequence; i is the historical subsequence number, i = [1, 2, …, n], n is the total number of historical subsequences; The average energy value standard deviation of the noise sequence N is represented; Pc is the rated current of the electrical equipment; t is the sequence length of the noise sequence N; λ1 and λ2 are weight coefficients of the average energy value standard deviation and the sequence length respectively, which are obtained based on historical big data test;

[0076] The method for intelligently analyzing the dynamic window sequence set also includes intelligently processing the first current signal subsequence, and the method is as follows:

[0077] The average current and average electrical signal frequency of the center sequence C are obtained, as well as the average current and average electrical signal frequency of the front k acquisition time periods and the rear k acquisition time periods of the center sequence C. The average current, rated current and average electrical signal frequency are normalized, and the steady-state analysis value of the dynamic window is calculated based on the steady-state analysis model. The expression of the steady-state analysis model is:

[0078]

[0079] In the formula, Sav is the steady-state analysis value of the dynamic window; If j and Ff j are the average current and average electrical signal frequency of the front j acquisition time periods of the center sequence C, Ib (k-j) ​and Fb (k-j) respectively the average current and the average electric signal frequency of the k-jth acquisition period of the central sequence C; j is the number of the acquisition period, j=[1, 2, …, k]; Im and Fm are respectively the average current and the average electric signal frequency of the central sequence C;

[0080] The method for intelligently analyzing the dynamic window sequence set further comprises intelligently processing the second current signal subsequence, and the method is as follows:

[0081] A risk sequence of a previous risk period generating the electric leakage protection in the historical current signal sequence is obtained, a plurality of risk sequences are determined, and the instantaneous frequency and the instantaneous current are taken as the first transient parameter and the second transient parameter of the risk sequence; the first transient parameter and the second transient parameter corresponding to all acquisition points in the risk sequence are obtained, and a first transient parameter set and a second transient parameter set are obtained;

[0082] The same transient parameter set is subjected to parameter analysis, and the method is as follows:

[0083] The mean value of the transient parameters of the last acquisition point of the plurality of risk sequences is calculated as the risk parameter of the transient parameter set;

[0084] The transient parameters in the same transient parameter set are sorted in ascending order to obtain a parameter sorting sequence, the parameter sorting sequence is subjected to sequence splitting based on the second decomposition layer number, a plurality of parameter sorting sub-sequences with the same number as the second decomposition layer number are obtained, and the median value of the parameter sorting sub-sequence is obtained;

[0085] The number of transient parameters contained in each parameter sorting sub-sequence is counted to obtain the total number of sub-sequence transient parameters, and the weighted average parameter in the transient parameter set is obtained based on the parameter sorting sub-sequence median value and the total number of sub-sequence transient parameters, and is taken as an abnormal parameter by using the weighted average algorithm;

[0086] The time point corresponding to the weighted average parameter in the risk sequence is obtained by using the interpolation algorithm and is marked as a target acquisition point, the rated power of the electrical equipment corresponding to the target acquisition point is obtained and is marked as a corrected power, and the interval time length from the time point corresponding to the target acquisition point to the end time point of the risk period is taken as the risk time length of the abnormal parameter corresponding to the risk sequence;

[0087] Based on the above method for parameter analysis of the same transient parameter set, the abnormal parameters and the risk time lengths corresponding to the first transient parameter set and the second parameter set are obtained, and are respectively marked as the first risk parameter, the second risk parameter, the first abnormal parameter, the second abnormal parameter, and the first risk time length and the second risk time length;

[0088] determining real-time current and real-time electric signal frequency based on the second current signal subsequence as the first real-time parameter and the second real-time parameter, and obtaining the corresponding real-time electrical equipment power; if any real-time parameter reaches the corresponding risk parameter, the actual risk duration of the corresponding real-time parameter is counted, and the first actual risk duration corresponding to the first transient parameter set and the second actual risk duration corresponding to the second transient parameter set are obtained respectively;

[0089] The transient risk value Tsv of the second current signal subsequence is calculated by the formula

[0090] In the formula, I t is the real-time current; F t is the real-time frequency; I1 is the first abnormal parameter; F1 is the second abnormal parameter; T I is the first actual risk duration; T F is the second actual risk duration; T t is the actual risk duration; min() is the minimum value comparison function; P t is the real-time electrical equipment power; P1 is the corrected power.

[0091] A current diagnosis model is established based on the steady-state analysis value Sav and the transient risk value Tsv, and the real-time current is diagnosed and warned based on the current diagnosis model.

[0092] The method for establishing the current diagnosis model based on the steady-state analysis value Sav and the transient risk value Tsv is:

[0093] The steady-state analysis value Sav of the current first current signal subsequence is compared with the preset steady-state threshold S0.

[0094] If Sav>S0, it is determined that the real-time current has abnormal steady-state characteristics and a first abnormal label is generated; otherwise, no operation is performed.

[0095] And the transient risk value Tsv of the current second current signal subsequence is compared with the preset transient threshold T0.

[0096] If Tsv≤0.8T0, no operation is performed; if 0.8T0<Tsv≤T0, the current second current signal subsequence is marked as an abnormal subsequence, and it is determined whether there is at least a continuous number of z abnormal subsequences; if there is at least a continuous number of z abnormal subsequences, it is determined that the real-time current has abnormal transient characteristics and a second abnormal label is generated; if there is no at least a continuous number of z abnormal subsequences, no operation is performed; it should be noted that z is obtained according to big data analysis; for example, if z is 2, it means that there is at least a continuous number of 2 abnormal subsequences, and it is determined that the real-time current has abnormal transient characteristics.

[0097] ​If T0 < Tsv, a second real-time current transient characteristic abnormality is judged, and a third abnormality label is generated;

[0098] According to the first abnormality label, the second abnormality label and the third abnormality label, it is determined that there is a leakage phenomenon, and a technician is notified to perform maintenance.

[0099] In embodiment two, the present application proposes an intelligent leakage current real-time monitoring and diagnosis system, which is applied to the intelligent leakage current real-time monitoring and diagnosis method proposed in embodiment one, and specifically includes:

[0100] A data acquisition module is configured to acquire real-time current signals of a target area power grid and corresponding current signal time stamps, and to acquire corresponding current signal sequences by intelligently processing the real-time current signals.

[0101] A data preprocessing module is configured to perform wavelet packet transform processing on the current signal sequences, decompose the current signal processing sequences into current signal sub-sequences of different frequency bands, and intelligently process the current signal sub-sequences to acquire the current signal sub-sequences.

[0102] A data analysis module is configured to establish a dynamic window for each current signal sub-sequence, mark a current signal sub-sequence located at the center of the dynamic window as a center sequence C, take current signal sub-sequences contained in the dynamic window as a dynamic sequence of the center sequence C, establish a dynamic window sequence set, intelligently analyze the dynamic window sequence set, and acquire steady-state characteristics Sav and transient characteristics Tsv of the current signal.

[0103] An intelligent monitoring and diagnosis module is configured to establish a current diagnosis model based on the steady-state analysis value Sav and the transient risk value Tsv, diagnose and warn the real-time current based on the current diagnosis model.

[0104] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited thereto, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the present application.

Claims

1. A method for real-time monitoring and diagnosis of intelligent leakage current, characterized in that, include The system acquires the real-time current signal and corresponding current signal timestamp of the power grid in the target area, and performs intelligent processing on the real-time electrical signal to obtain the corresponding current signal sequence. Wavelet packet transform is performed on the current signal sequence to decompose the current signal sequence into current signal sub-sequences of different frequency bands, and intelligent processing is performed on the current signal sub-sequences to obtain the current signal sub-sequences. The wavelet packet transform processing of the current signal processing sequence includes: Select a first decomposition level, and based on the first decomposition level, decompose the current signal processing sequence into several first current signal sub-sequences with a frequency band width of a first frequency band; A second decomposition layer is selected, and based on the second decomposition layer, the current signal processing sequence is decomposed into several second current signal subsequences with a frequency band width of the second frequency band; The number of the first decomposition level is less than the number of the second decomposition level; A dynamic window is established for each current signal subsequence, and the current signal subsequence located at the center of the dynamic window is marked as the center sequence C. The current signal subsequences contained in the dynamic window are used as the dynamic sequence of the center sequence C to establish a dynamic window sequence set. Intelligent feature analysis is performed on the dynamic window sequence set to obtain the steady-state feature Sav and transient feature Tsv of the current signal. The method for intelligent feature analysis of dynamic window sequence sets is as follows: The rated current of historical electrical equipment and the corresponding historical current signal sequence are obtained from the intervention current signal. The historical electrical equipment information and historical current signal sequence are intelligently analyzed to determine the degree of fit between the electrical equipment information and the current signal sequence. The method for intelligent analysis of historical electrical equipment information and historical current signal sequences is as follows: The historical current signal sequence is decomposed based on the first decomposition level to obtain historical subsequences. The mean current of the sampling points of each historical subsequence is obtained, and the standard deviation of the current is calculated based on the current value of the sampling points of each historical subsequence to obtain the dispersion of the current of the sampling points of each historical subsequence. A current diagnostic model is established based on the steady-state analysis value Sav and the transient risk value Tsv. The real-time current is then diagnosed and an early warning is issued based on the current diagnostic model.

2. The intelligent leakage current real-time monitoring and diagnosis method according to claim 1, characterized in that, Calculate the difference in standard deviation of current between adjacent historical subsequences, establish a sequence of differences in standard deviation of current between historical subsequences, traverse the sequence of differences in standard deviation of current between historical subsequences, mark the elements in the sequence of differences in standard deviation of current between historical subsequences that are less than the standard deviation threshold as target elements, and identify the continuity of several target elements. For the element numbering of the current standard deviation difference sequence, if there are at least m target elements with consecutive numbers, then the target elements with consecutive numbers are determined to be continuous, and the target element with the smallest number is marked as the selected element. The historical subsequence corresponding to the current standard deviation calculated from the selected element and used as the subtrahend is marked as the target sequence. The historical current signal sequence is segmented at the starting sequence of the target sequence to obtain two segmented sequences. The segmented sequence to which the target sequence belongs is deleted, and the remaining segmented sequence is used as the noise sequence N.

3. The intelligent leakage current real-time monitoring and diagnosis method according to claim 2, characterized in that, Process the noise sequence N: Obtain the extreme value of each historical subsequence in the noise sequence N, and calculate the average energy value of the historical subsequence by taking the absolute value and mean of the extreme value of each historical subsequence; Through formula The goodness of fit between the noise sequence N and the rated current of the electrical equipment was calculated. ; In the formula, The average energy value for each historical subsequence; i is the historical subsequence number, i=[1,2,...,n], and n is the total number of historical subsequences; The standard deviation of the average energy value of the noise sequence N is represented; Pc is the rated current of the electrical equipment; t is the sequence duration of the noise sequence N; and These are the weighting coefficients for the standard deviation of the average energy value and the sequence duration, respectively, obtained based on historical big data testing.

4. The intelligent leakage current real-time monitoring and diagnosis method according to claim 3, characterized in that, The method for intelligent feature analysis of dynamic window sequence sets also includes intelligent processing of the first current signal subsequence, as follows: The average current and average electrical signal frequency of the central sequence C are obtained, as well as the average current and average electrical signal frequency of the first k and last k acquisition periods of the central sequence C. The average current, rated current, and average electrical signal frequency are then normalized. Based on the steady-state analysis model, the steady-state analysis value of the dynamic window is calculated. The expression of the steady-state analysis model is as follows: ； In the formula, Sav is the steady-state analysis value of the dynamic window; and These are the average current and average electrical signal frequency during the first j acquisition periods of the central sequence C, respectively. and , , and , respectively, represent the average current and average electrical signal frequency of the last kj acquisition periods of the central sequence C; j is the acquisition period number, j=[1,2,...,k]; Im and Fm are the average current and average electrical signal frequency of the central sequence C, respectively.

5. The intelligent leakage current real-time monitoring and diagnosis method according to claim 3, characterized in that, The method for intelligent feature analysis of dynamic window sequence sets also includes intelligent processing of the second current signal subsequence, as follows: Obtain the risk sequence from the historical current signal sequence for the previous risk period before the leakage current protection was generated, determine several risk sequences, and use the instantaneous frequency and instantaneous current as the first and second transient parameters of the risk sequence; Obtain the first transient parameters and the second transient parameters corresponding to all collection points in the risk sequence to obtain the first transient parameter set and the second transient parameter set; The following method is used to perform parameter analysis on the same transient parameter set: Calculate the mean of transient parameters at the last collection point of several risk sequences, and use it as the risk parameter of the transient parameter set; The transient parameters in the same transient parameter set are sorted in ascending order to obtain a parameter sorting sequence. The parameter sorting sequence is then split based on the second decomposition level to obtain several parameter sorting subsequences with the same number as the second decomposition level. The values ​​in the parameter sorting subsequences are then obtained. The number of transient parameters contained in each parameter sorted subsequence is counted to obtain the total number of transient parameters in the subsequence. Based on the median of the parameter sorted subsequence and the total number of transient parameters in the subsequence, a weighted average parameter is calculated from the transient parameter set using a weighted average algorithm and used as an outlier parameter. The weighted average parameter is obtained by using an interpolation algorithm to obtain the time point corresponding to the risk sequence and marked as the target collection point. The rated power of the electrical equipment corresponding to the target collection point is obtained and marked as the corrected power. The time interval from the time point corresponding to the target collection point to the end time point of the risk period is used as the risk duration of the abnormal parameter corresponding to the risk sequence. Based on the above method for parameter analysis of the same transient parameter set, the abnormal parameters and risk durations corresponding to the first transient parameter set and the second parameter set are obtained respectively, and are marked as the first risk parameter, the second risk parameter, the first abnormal parameter, the second abnormal parameter, the first risk duration, and the second risk duration, respectively.

6. The intelligent leakage current real-time monitoring and diagnosis method according to claim 5, characterized in that, Based on the second current signal subsequence, the real-time current and real-time electrical signal frequency are determined and used as the first real-time parameter and the second real-time parameter to obtain the corresponding real-time electrical equipment power; if any real-time parameter reaches the corresponding risk parameter, the actual risk duration of the corresponding real-time parameter is calculated, and the first actual risk duration corresponding to the first transient parameter set and the second actual risk duration corresponding to the second transient parameter set are obtained respectively. Using the following formula: ； The transient risk value Tsv of the second current signal subsequence was calculated; In the formula, For real-time current; For real-time frequency; This is the first abnormal parameter; This is the second abnormal parameter; The duration of the first actual risk; This refers to the second actual risk duration; This refers to the actual duration of the risk. min() is a minimum value comparison function; Real-time power of electrical equipment; To correct the power.

7. The intelligent leakage current real-time monitoring and diagnosis method according to claim 6, characterized in that, The method for establishing a current diagnosis model based on the steady-state analysis value Sav and the transient risk value Tsv is as follows: Compare the steady-state analysis value Sav of the current first current signal subsequence with the preset steady-state threshold S0; If Sav > S0, determine that the real-time current steady-state characteristic is abnormal and generate a first abnormal label; Otherwise, do nothing; And compare the transient risk value Tsv of the current second current signal subsequence with the preset transient threshold T0; If Tsv ≤ 0.8T0, do nothing; if 0.8T0 < Tsv ≤ T0, mark the current second current signal subsequence as an abnormal subsequence, and determine whether there are at least z consecutive abnormal subsequences; if there are at least z consecutive abnormal subsequences, determine that the real-time current transient characteristic is first abnormal and generate a second abnormal label; if there are not at least z consecutive abnormal subsequences, do nothing; If T0 < Tsv, determine that the real-time current transient characteristic is second abnormal and generate a third abnormal label; Determine that there is a leakage phenomenon according to the first abnormal label, the second abnormal label and the third abnormal label, and notify the technical personnel for maintenance.

Images