A panoramic intelligent perception method for the state of a distribution network

Abstract

The present invention discloses a method for panoramic intelligent perception of the state of a distribution network, which relates to the technical field of distribution network monitoring and management and is used to solve the problem of poor abnormal monitoring of the distribution network. Through precise acquisition and preprocessing of the node signals of the distribution network, the present invention realizes multi-scale decomposition and baseline construction of the signals, effectively extracts the steady-state and abnormal dynamic characteristics of the distribution network, uses variational mode decomposition technology, constructs a baseline by screening low-frequency modes through the central frequency, and then combines the difference between the baseline and the original signal to construct robust residuals and abnormal energy indicators to accurately determine abnormal events. After an abnormal trigger, the distribution network nodes switch to the ultra-high frequency sampling mode, use time-domain difference to extract local disturbance characteristics, and perform noise reduction through robust filtering, and adaptively adjust the filtering parameters according to the local edge ratio, so as to fully retain the abnormal edge information, thereby realizing real-time and high-precision perception of the abnormal state of the distribution network and improving the response speed and accuracy of the panoramic perception of the distribution network.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network monitoring and management. More specifically, the present invention relates to a method for panoramic intelligent perception of the state of a distribution network. Background Art

[0002] In recent years, with the acceleration of the urbanization process and the increasing popularity of new energy access, as the key terminal link of the power system, the stable operation and intelligent monitoring of the distribution network are particularly important. The distribution network undertakes the task of distributing the electric energy in the high-voltage power grid to each terminal node user. Its operating environment is affected by multiple factors such as load fluctuations, equipment aging, and external interference, and problems such as voltage fluctuations, short-term load mutations, and local faults are likely to occur.

[0003] Deficiencies of the prior art: It is difficult to capture medium-short-term and minute dynamic changes during the monitoring process of the distribution network, resulting in the difficulty of timely identification of transient anomalies and subtle disturbances, which in turn affects the accurate detection of local abnormal events. At the same time, the signals of each distribution node are often affected by noise interference during the transmission process, and there are problems of data discontinuity. While the existing processing schemes achieve signal smoothing, they often sacrifice the integrity of key abnormal edge information, which not only reduces the sensitivity of fault warning but also restricts the accurate response of the intelligent dispatching system to the actual operating state, thereby affecting the safe and stable operation of the distribution network. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the following solutions are provided to solve the problem of poor abnormal monitoring of the distribution network in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for panoramic intelligent perception of the state of a distribution network, comprising the following steps:

[0007] Collect and process the signals of the distribution network nodes to obtain the original signal, perform multi-scale decomposition on the original signal using variational mode decomposition, and screen out the low-frequency modes through the central frequency and use them to reconstruct the baseline;

[0008] According to the difference between the reconstructed baseline signal and the original signal, construct a robust residual and an abnormal energy index to determine the abnormal event;

[0009] After triggering the abnormal event, switch the sampling mode to the ultra-high-frequency sampling mode to collect data for the abnormal event;

[0010] Perform time-domain difference on the data sampled at ultra-high frequency, extract the local disturbance characteristics, and perform noise reduction through robust filtering. Adjust the parameters of the robust filtering according to the local edge ratio, and perform perception analysis of the distribution network on the adjusted collected signal.

[0011] In a preferred embodiment, the node signals of the distribution network are collected and processed to obtain the original signal, and variational mode decomposition is used to perform multi-scale decomposition on the original signal. The specific steps are as follows:

[0012] Collect the node signals of the distribution network, including voltage signals, current signals, power signals, and auxiliary signals;

[0013] Calibrate and normalize all the collected signals according to a unified timestamp to obtain the original signal;

[0014] Perform multi-scale signal decomposition on the normalized original signal to obtain the modes and the corresponding set of central frequencies.

[0015] In a preferred embodiment, the low-frequency modes are screened out by the central frequency and used to reconstruct the baseline. The specific steps include:

[0016] Compare the central frequency with the spectral threshold, screen out the low-frequency modes, and determine the set of low-frequency mode indices;

[0017] Perform baseline reconstruction operation on the low-frequency modes obtained by multi-scale signal decomposition, and output the baseline signal.

[0018] In a preferred embodiment, according to the difference between the reconstructed baseline signal and the original signal, a robust residual and an abnormal energy index are constructed to determine the abnormal event. The specific steps are as follows:

[0019] Define the residual signal according to the deviation between the baseline and the original signal;

[0020] Construct the robust residual and the abnormal index function according to the Huber loss function, and use the local time window integration method to construct the abnormal energy index. The calculated abnormal energy index is used as the basis for real-time abnormal judgment in the real-time monitoring link to determine the abnormal event.

[0021] In a preferred embodiment, after the abnormal event is triggered, the sampling mode is switched to the ultra-high frequency sampling mode to collect data for the abnormal event. The specific steps are as follows:

[0022] After the abnormal event is triggered, compare the real-time calculated abnormal energy index with the preset abnormal energy threshold:

[0023] When the abnormal energy index is greater than the preset abnormal energy threshold, it is calibrated as an abnormal event, and the abnormal trigger time is recorded and the sampling time window is determined;

[0024] Perform continuous data sampling using the ultra-high frequency sampling mode within the sampling time window for which the abnormal event has been determined.

[0025] In a preferred embodiment, time-domain difference is performed on the ultra-high frequency sampled data to extract local disturbance characteristics, and noise reduction is performed through robust filtering. The parameters of the robust filtering are adjusted according to the local edge ratio, and the adjusted acquired signal is subjected to perception analysis of the distribution network, including the following steps:

[0026] Perform local change analysis on the continuously sampled data using time-domain difference operation, and determine the local disturbance measure according to the difference between adjacent sampling points;

[0027] Perform robust filtering on the local disturbance measure determined by ultra-high frequency sampling and the original signal data to determine the signal after de-filtering;

[0028] Define the local edge index, perform comparison and verification of local high-frequency components, determine the differential amplitude ratio between the original signal data and the filtered signal in the local area, and adjust the robust filtering parameters according to the differential amplitude ratio;

[0029] Perform perception analysis on the state of the distribution network for the adjusted acquired signal.

[0030] Technical effects and advantages of a panoramic intelligent perception method for the state of a distribution network according to the present invention:

[0031] Through precise acquisition and preprocessing of the signals of the distribution network nodes, the present invention realizes multi-scale decomposition and baseline construction of the signals, effectively extracts the steady-state and abnormal dynamic characteristics of the distribution network, uses variational mode decomposition technology, constructs a baseline by screening the low-frequency mode through the central frequency, and then constructs a robust residual and abnormal energy index in combination with the difference between the baseline and the original signal to accurately determine abnormal events. After an abnormal event is triggered, the distribution network nodes switch to the ultra-high frequency sampling mode, use time-domain difference to extract local disturbance characteristics, and perform noise reduction through robust filtering, and adaptively adjust the filtering parameters according to the local edge ratio, so as to fully retain the abnormal edge information, thereby realizing real-time and high-precision perception of the abnormal state of the distribution network, providing reliable data support for fault warning and intelligent scheduling, and significantly improving the response speed and accuracy of the panoramic perception of the distribution network. Description of the Drawings

[0032] Figure 1 It is a schematic flow chart of a panoramic intelligent perception method for the state of a distribution network according to the present invention. Detailed Embodiments

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0034] To achieve the above object, Figure 1 The structural schematic diagram of a method for panoramic intelligent perception of the state of a distribution network according to the present invention is given, and the specific steps are as follows;

[0035] Collect and process the node signals of the distribution network to obtain the original signal, perform multi-scale decomposition on the original signal using variational mode decomposition, and screen out the low-frequency modes through the central frequency and use them to reconstruct the baseline;

[0036] According to the difference between the reconstructed baseline signal and the original signal, construct a robust residual and an abnormal energy index, and determine the abnormal event;

[0037] After triggering the abnormal event, switch the sampling mode to the ultra-high frequency sampling mode to collect data for the abnormal event;

[0038] Perform time-domain difference on the data sampled at ultra-high frequency, extract the local disturbance characteristics, and perform noise reduction through robust filtering. Adjust the parameters of the robust filtering according to the local edge ratio, and perform perception analysis of the distribution network on the collected signal after adjustment.

[0039] In the panoramic intelligent perception of the distribution network, the collected original signals (such as voltage, current, power, etc.) contain both low-frequency components describing the steady-state operation of the distribution network and high-frequency fluctuations that may indicate sudden abnormalities. Therefore, the original signal often mixes different frequency components, where the low-frequency part reflects the stable operation trend of the distribution network, and the high-frequency part may contain fault precursors or transient disturbances. Using the variational mode decomposition (VMD) method, the mixed signal can be adaptively decomposed into a series of modal components, and each modal has the characteristic of local concentration in the frequency spectrum, thus providing a basis for subsequent baseline construction;

[0040] Step 1, perform multi-scale baseline modeling and real-time residual calculation, and the specific steps are as follows:

[0041] Use high-precision sensors installed at key nodes (such as transformers, distribution cabinets, branch nodes, etc.) to collect the node signals of the distribution network, including voltage signals, current signals, active / reactive power signals, and other auxiliary signals (such as frequency, temperature, etc., which can be determined according to the specific application scenario);

[0042] It should be noted that the node signals of the distribution network include data sets related to each part of the distribution network, including but not limited to numerical data, character data, etc. For example, the outdoor environmental temperature of Node 1 is 27 degrees Celsius, where Node 1 is character data and 27 degrees Celsius is numerical data.

[0043] Calibrate all the acquired signals according to a unified timestamp to eliminate transmission delays and device clock biases, enabling signals from different acquisition points to be compared at the same moment. Then, use interpolation or timestamp calibration algorithms to resample the data to a unified time reference;

[0044] Conduct preliminary filtering on the acquired signals to remove noise and outliers caused by sensor errors, transient interferences, or occasional faults. Meanwhile, according to the device calibration information, perform amplitude calibration and normalization on the signals to eliminate differences between sensors and possible DC offsets;

[0045] Use the normalized original signal as the input for multi-scale signal decomposition.

[0046] Carry out the solution constraint optimization of the original signal f(t), that is, decompose f(t) into K intrinsic mode components and their central frequencies; the constructed optimization is as follows: , where f(t) represents the original signal to be decomposed, reflecting the real-time state data of the distribution network; represents the k-th mode component, and the goal is to capture a certain frequency range component in the signal; K is the preset number of modes, and its value is based on the understanding of the signal complexity (for example, determined through prior analysis or experiments); is the central frequency of the k-th mode, which is used to down-convert this mode to the low-frequency region for bandwidth evaluation; the operator represents the derivative with respect to time, reflecting the instantaneous change of the mode in the time domain; the expression is the Hilbert transform kernel, and the analytic signal of is constructed using convolution operation (denoted by the symbol *) to make it have positive spectrum information; the exponential factor is used to shift the spectrum of the mode signal so that the central frequency of this mode moves to zero frequency, facilitating bandwidth measurement; represents the norm on the given interval, calculating the sum of the squares of the bandwidth energies of the signal; the objective term of this optimization problem is to accumulate the bandwidth energies of all modes.

[0047] The fact that the objective term of the optimization problem is to accumulate the bandwidth energies of all modes means that minimizing this objective can make the high-frequency energy of each mode as low as possible after spectrum shifting, thus ensuring its spectrum is compact, that is, having a narrow bandwidth;

[0048] This optimization aims to ensure that each mode has a high concentration in the spectrum and can reconstruct the original signal without distortion;

[0049] The obtained output is a set, including the mode and the corresponding central frequency , where the modes with lower center frequencies usually carry the steady-state components of the signal, while the high-frequency modes may contain abnormal information;

[0050] Output modes will be used as the original input for subsequent baseline reconstruction;

[0051] It should be noted that the parameter K represents the preset decomposition level, and its value must be determined according to the spectral complexity of the actual signal. The center frequency is used to distinguish between low-frequency and high-frequency modes, that is, the low-frequency modes are classified as part of the baseline according to the spectral threshold.

[0052] Perform baseline reconstruction and multi-scale baseline construction. That is, after obtaining the decomposition results of each mode, reconstructing the baseline signal reflecting the stable operation trend of the distribution network becomes the reference for anomaly detection. This process ensures that the constructed baseline can smoothly reflect the global dynamic trend and eliminate short-term disturbance interference through the adaptive screening and reconstruction of low-frequency modes. The specific steps are as follows:

[0053] Determine the set of low-frequency mode indices according to the magnitudes of the center frequencies of each mode . The low-frequency modes need to satisfy , where is the preset frequency threshold;

[0054] Define the baseline reconstruction operation R: , where R(⋅) represents the reconstruction operation based on the spectral selection criterion, which requires maintaining smoothness while retaining the dynamic trend of the system. L is the set of low-frequency mode indices, and its determination depends on the threshold setting of the center frequencies of each mode;

[0055] Pass the low-frequency modes obtained by VMD decomposition through the baseline reconstruction operation to output the baseline signal , and this signal is used as the reference input for subsequent residual calculation.

[0056] According to the deviation between the baseline signal and the original signal , define the residual signal to reflect transient anomaly characteristics, and introduce a robust metric in the residual calculation to reduce the interference of extreme deviations on the judgment. Directly define the robust residual as: ;

[0057] At the same time, to avoid the excessive influence of extreme values on subsequent judgments, use the Huber loss function to construct a robust residual measure. Specifically, define the Huber loss function as: , where is the switching parameter, which is used to determine the degree of using quadratic penalty and linear penalty. This switching parameter is set according to the actual system noise level and anomaly amplitude characteristics, represents the residual signal between the original signal and the baseline signal;

[0058] Meanwhile, construct the anomaly index function , which is defined as the derivative of the Huber loss function: , where sgn(⋅) is the sign function used to preserve the anomaly direction information;

[0059] Based on the calculation, obtain the robust residual as the real-time anomaly criterion.

[0060] Conduct anomaly measurement criterion construction and data transfer as follows:

[0061] To enable the panoramic perception of the distribution network to respond promptly to the transient anomalies of the distribution network, it is necessary to convert the robust residual at time t into the anomaly energy index A(t), so as to achieve the effect of comprehensively accumulating the anomaly effects within the local time window;

[0062] Based on the output of the anomaly index function , construct the anomaly measurement index to determine the occurrence of anomaly events in the panoramic perception. This index is used to guide the activation of the ultra-high frequency sampling mode;

[0063] Use the local time window integration method to construct the anomaly energy index: , where is the time window width;

[0064] The calculated anomaly energy index is output as an anomaly trigger signal in the real-time monitoring link. Once the anomaly energy index exceeds the predetermined threshold, the subsequent ultra-high frequency sampling and anomaly capture processes are activated.

[0065] In summary, through the above steps, after the original distribution network signal is decomposed by VMD, multiple modes are obtained. The baseline signal reflecting the steady-state operation trend is adaptively constructed from the low-frequency mode set. Then, the residual is constructed based on the difference between the baseline and the original signal, and the Huber loss function is used to design the robust anomaly index. Finally, the anomaly energy index is constructed through local time integration, and this index can be used to indicate anomaly events.

[0066] After constructing the anomaly energy index, trigger the anomaly event in a timely manner and obtain higher-resolution anomaly data by dynamically adjusting the sampling period.

[0067] Step 2: Perform event triggering and ultra-high frequency sampling switching. The abnormal energy index reflects the cumulative effect of signal deviation from the baseline within a local time window. However, since the conventional sampling period may not meet the requirement of capturing instantaneous abnormal details in terms of sampling density, a dynamic sampling switching mechanism needs to be designed: when it is detected that the abnormal energy exceeds the preset threshold, immediately switch from the conventional mode to the ultra-high frequency mode, and set a certain sampling time window before and after the occurrence of the abnormality, so as to completely record the abnormal characteristics and their evolution process. The specific steps are as follows:

[0068] After triggering the abnormal event, compare the real-time calculated abnormal energy index with the preset abnormal energy threshold , when , that is, calibrate the abnormal event and record the abnormal trigger moment of the abnormal event as the trigger benchmark for subsequent high-frequency sampling;

[0069] Meanwhile, determine the time window and calibrate the event moment. After determining the abnormal trigger moment , set the capture window as: , where is the reserved sampling time before the occurrence of the abnormality, used to capture the precursors or change trends before the start of the event; is the continuous sampling time after the occurrence of the abnormality, used to record the recovery process after the abnormality;

[0070] After determining the event trigger moment and the sampling time window, use the ultra-high frequency sampling mode within this time interval, and within the time interval, the sampling interval can be adjusted to: , in the formula, is the total number of samples expected to be collected within the abnormal acquisition window, is the ultra-high frequency sampling period, which is required to be much smaller than the original sampling period to meet the capture of transient abnormalities.

[0071] During the ultra-high frequency sampling period, due to the sharp increase in data volume, it is necessary to store the collected data in real time through a cache mechanism. After the abnormal event ends, restore to the conventional sampling mode to avoid overload, so as to realize the dynamic switching and management of the sampling state;

[0072] Within the sampling window, define that the ultra-high frequency sampling data sequence satisfies: ;

[0073] Store the collected data in the cache area. After the abnormal acquisition interval ends, that is, at the moment , automatically issue an instruction to switch back to the conventional sampling mode.

[0074] It should be noted that the preset abnormal energy threshold is determined based on historical data, equipment characteristics and engineering requirements, and is used to distinguish normal fluctuations from real abnormal events.

[0075] Step 3: Perform high-frequency data preprocessing and filtering. The signal sequence obtained by ultra-high-frequency sampling can completely record the transient details during abnormal events. However, it will also introduce high-frequency noise and random interference. To enable the abnormal features contained in the signal to be used in subsequent processing, preprocess and filter the data sampled at ultra-high frequencies. The specific steps are as follows:

[0076] Use differential operation to perform local change analysis on continuous sampling data. Construct a local perturbation measure based on the difference between adjacent sampling points. Define the time-domain differential operation as: , where represents the ultra-high-frequency sampling data at time , represents the ultra-high-frequency sampling period; represents the increment between adjacent samplings, reflecting the local signal mutation situation;

[0077] Perform robust filtering on the operation result and the original signal data to solve the filtered signal z(t), and construct the following optimization problem: , where z(t) represents the signal obtained after filtering; is the first derivative of z(t), and its absolute integral is the total variation of the signal, which is used to reflect the smoothness of the signal; is the regularization parameter, which balances the data fidelity term and the smoothing regularization term; is the robust loss function. Select, for example, the Huber loss or the truncated quadratic function to reduce the influence of extreme errors on the filtering result; represents the time interval where the ultra-high-frequency sampling data is located, corresponding to the abnormal acquisition window determined in Step 2;

[0078] After robust filtering, the filtered signal z(t) can retain the edge information of abnormal events. Therefore, introduce an edge-preserving correction step to prevent excessive smoothing of abnormal details during the filtering process through comparison and verification of local high-frequency components;

[0079] Define the local edge index E(t) as the ratio of the differential amplitude between the original signal data and the filtered signal z(t) in the local area: , where is the ultra-high-frequency sampling period, is the ultra-high-frequency sampling data at time t, is a constant term greater than 1;

[0080] When the differential amplitude ratio is near 1, it indicates that the filtering does not lose the abnormal edge features; if it is significantly lower than 1, it indicates that the filtering process may be over-smoothed and the filtering parameters need to be adjusted.

[0081] The differential amplitude ratio is used as feedback information to provide a basis for adaptive adjustment of the filtering algorithm and is also used for subsequent anomaly detection, fault diagnosis, etc., so as to ensure that the panoramic intelligent perception of the entire distribution network state can capture accurate, continuous, and high-quality anomaly features during real-time monitoring. The adjusted signal is used for the panoramic perception analysis of the distribution network state to realize the identification and judgment of abnormal events.

[0082] In summary, by constructing local perturbation metrics through time-domain differentiation to provide a basis for noise modeling, then constructing a robust total variation regularization filtering model and solving the optimization problem to obtain the filtered signal, so as to retain the anomaly edges while suppressing high-frequency noise. Then, the filtering result is verified for consistency and corrected for edge preservation by calculating the local edge index, and the feedback information is used to dynamically adjust the filtering parameters to ensure that abnormal transient information (such as mutation edges, rapid fluctuations) is accurately retained, providing a reliable basis for subsequent fault diagnosis and making the monitoring and early warning functions of the entire distribution network more accurate and efficient.

[0083] It should be noted that the threshold information related in this embodiment is pre-set by professionals and will not be explained in detail here. In the embodiment, some parameter English letters are the same, but different meanings are explained when used, and they will not be explained one by one here.

[0084] Through the precise acquisition and preprocessing of the distribution network node signals, the present invention realizes the multi-scale decomposition and baseline construction of the signals, effectively extracts the steady-state and abnormal dynamic characteristics of the distribution network, uses the variational mode decomposition technology to construct the baseline by screening the low-frequency modes of the central frequency, and then combines the difference between the baseline and the original signal to construct the robust residual and abnormal energy indicators to accurately determine abnormal events. After the anomaly is triggered, the distribution network node switches to the ultra-high-frequency sampling mode, uses time-domain differentiation to extract local perturbation characteristics, and reduces noise through robust filtering, and adaptively adjusts the filtering parameters according to the local edge ratio, so as to fully retain the abnormal edge information, thus realizing the real-time and high-precision perception of the abnormal state of the distribution network, providing reliable data support for fault early warning and intelligent scheduling, and significantly improving the response speed and accuracy of the panoramic perception of the distribution network.

[0085] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by software simulation of a large amount of collected data to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0086] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0087] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0088] In addition, in each embodiment of this application, the functional modules can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0089] As mentioned above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claimed rights.

[0090] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for panoramic intelligent perception of distribution network status, characterized by: The steps include: The distribution network node signals are collected and processed to obtain the original signals, and the original signals are decomposed into multiple scales using variational mode decomposition. The low-frequency modes are screened out through the center frequency and used to reconstruct the baseline. According to the difference between the reconstructed baseline signal and the original signal, a robust residual and anomaly energy index are constructed to determine the abnormal event; After an abnormal event is triggered, the sampling mode is switched to an ultra-high frequency sampling mode to collect data on the abnormal event; Perform time-domain differentiation on the ultra-high frequency sampled data to extract local disturbance characteristics, and perform noise reduction through robust filtering. Adjust the parameters of the robust filtering according to the local edge ratio, and use the adjusted collected signals for distribution network perception analysis; After an abnormal event is triggered, the sampling mode is switched to the ultra-high frequency sampling mode to collect data for the abnormal event. The specific steps are as follows: After an abnormal event is triggered, the abnormal energy index calculated in real time is compared with the preset abnormal energy threshold: When the abnormal energy index is greater than the preset abnormal energy threshold, it is marked as an abnormal event and the abnormal triggering time is recorded and the sampling time window is determined; Adjusting the sampling mode to perform continuous data sampling within the sampling time window where the abnormal event is determined; The ultra-high frequency sampled data is differentiated in the time domain to extract the local disturbance characteristics, and the noise is reduced through robust filtering. The parameters of the robust filtering are adjusted according to the local edge ratio, and the adjusted collected signals are used for perception analysis of the distribution network, including the following steps: Use time-domain difference operation to analyze the local changes of the continuously sampled data, and determine the local disturbance measure based on the difference between adjacent sampling points; Performing robust filtering on the local disturbance measure determined by sampling and the original signal data to determine a defiltered signal; Define local edge indicators, perform comparative verification of local high-frequency components, determine the differential amplitude ratio between the original signal data and the filtered signal in the local area, and adjust the robust filtering parameters according to the differential amplitude ratio; The adjusted collected signals are used for perception and analysis of the distribution network status.

2. A method for panoramic intelligent perception of distribution network status according to claim 1, characterized in that: The distribution network node signals are collected and processed to obtain the original signals, and the original signals are decomposed into multiple scales using variational mode decomposition. The specific steps are as follows: Collect distribution network node signals, including voltage signals, current signals, power signals and auxiliary signals; All collected signals are corrected and normalized according to a unified timestamp to obtain the original signal; The normalized original signal is subjected to multi-scale signal decomposition to obtain the mode and the corresponding center frequency set.

3. A method for panoramic intelligent perception of distribution network status according to claim 2, characterized in that: The low-frequency modes are filtered out through the center frequency and used to reconstruct the baseline. The specific steps include: Compare the center frequency with the spectrum threshold, filter out the low-frequency mode, and determine the low-frequency mode index set; The low-frequency modes obtained by decomposing the multi-scale signal are subjected to baseline reconstruction operation and the baseline signal is output.

4. A method for panoramic intelligent perception of distribution network status according to claim 3, characterized in that: According to the difference between the reconstructed baseline signal and the original signal, the robust residual and abnormal energy index are constructed to determine the abnormal event. The specific steps are as follows: Based on the deviation between the baseline and the original signal, a residual signal is defined; Robust residual and anomaly index functions are constructed according to the Huber loss function, and the anomaly energy index is constructed using the local time window integration method. The calculated anomaly energy index is used as the basis for real-time anomaly judgment in the real-time monitoring link to determine abnormal events.

Images