Abnormal Location Method and System for Distribution Transformer Area Based on Multimodal Dynamic Feature Analysis

Abstract

The present invention discloses a method and system for abnormal location of a distribution transformer area based on multi-modal dynamic feature analysis, belonging to the technical field of data processing, and comprising the following steps: S1. Obtain multi-source fusion data of the transformer area according to a sliding time window and determine a hidden danger path corresponding to an instantaneous voltage drop event in combination with the line topology; S2. Perform state fluctuation analysis based on the real-time state data and historical state data of the monitoring points to determine a state fluctuation factor, and construct an impedance prediction model according to the path disturbance data to determine an environmental disturbance factor; S3. Perform intensification processing on the hidden danger path according to the state fluctuation factor and the environmental disturbance factor to obtain a deterioration map of the transformer area; S4. Dynamically cut the deterioration map of the transformer area through the hidden danger risk value corresponding to the monitoring point to determine a deteriorated block, and determine corresponding early warning measures according to the deterioration degree corresponding to the deteriorated block. The solution overcomes the limitations of single data dimension analysis and static evaluation, and significantly improves the accuracy of hidden danger identification and the intelligent processing efficiency of monitoring points in complex scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent distribution networks, and specifically, to a method and system for abnormal location of distribution transformer areas based on multi-modal dynamic feature analysis. Background Art

[0002] During the operation of the power system, as the end link of the power system, the real-time monitoring and abnormal location of the operation state of the low-voltage distribution transformer area are crucial for power supply reliability. Traditional methods for abnormal location of distribution transformer areas mainly rely on single-type data and relatively simple analysis means. For the abnormal diagnosis of the transformer area lines, they mainly rely on single monitoring data (such as voltage and current steady-state parameters) and static topology models, and judge faults through threshold comparison or impedance matching algorithms. However, in actual operation, the state of the transformer area is dynamically affected by various factors, such as environmental factors (temperature, humidity, wind speed, etc.) and random changes in user electricity consumption behavior. These dynamic factors will cause the operation parameters of the transformer area to exhibit complex fluctuation characteristics, while traditional methods cannot effectively capture these dynamic features, making it difficult to accurately identify abnormal situations in the early stage and delaying the best time for fault handling.

[0003] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present application, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0004] In view of the problems in the prior art that due to relying on manual inspection or simple analysis of a single data source, the hidden dangers in the distribution transformer area are inaccurately identified and the efficiency is low, the present application proposes a method and system for abnormal location of distribution transformer areas based on multi-modal dynamic feature analysis. By collecting multi-source data of the transformer area for correlation analysis, it overcomes the limitations of single data dimension analysis and static evaluation, realizes accurate location of hidden danger paths from multi-source data fusion and topological association, dynamically evaluates the degradation degree of the transformer area through quantitative analysis of state fluctuations and environmental disturbances combined with the hidden danger risk values corresponding to the monitoring points, and automatically generates differentiated early warning measures, significantly improving the accuracy of hidden danger identification and the intelligent processing efficiency at monitoring points in complex scenarios.

[0005] In a first aspect, a technical solution provided in an embodiment of the present invention is: A method for abnormal location of distribution transformer areas based on multi-modal dynamic feature analysis, including the following steps:

[0006] S1. Obtain multi-source fusion data of the transformer area according to a sliding time window and determine the hidden danger path corresponding to the momentary drop event in combination with the line topology;

[0007] S2. Perform state fluctuation analysis on the real-time state data and historical state data of the monitoring points to determine the state fluctuation factor, and construct an impedance prediction model according to the path disturbance data to determine the environmental disturbance factor;

[0008] S3, according to the state fluctuation factor and the environmental disturbance factor, the potential danger path is intensified to obtain the degradation map of the substation area;

[0009] S4. Dynamically cut the degradation map of the substation area according to the hidden danger risk value corresponding to the monitoring point to determine the degraded block, and determine the corresponding early warning measures according to the degradation degree corresponding to the degraded block.

[0010] Preferably, the method of acquiring multi-source fusion data of the substation area according to the sliding time window and determining the hidden danger path corresponding to the instantaneous drop event in combination with the line topology comprises the following steps:

[0011] S11, dynamically generate a sliding time window based on the response characteristics of the equipment in the substation area and the environmental noise coefficient;

[0012] S12, the distributed acquisition terminal acquires the multi-source fusion data of the substation according to the sliding time window, wherein the multi-source fusion data of the substation includes at least voltage sag waveform data, harmonic distortion rate data, zero-sequence current phase difference data and temperature gradient data;

[0013] S13, using spline interpolation and kernel density estimation to perform spatiotemporal alignment operations on the multi-source fusion data of the station area to obtain a preliminary data set;

[0014] S14, screening the preliminary data set by impedance constraint and instantaneous drop event criterion and determining the candidate path set in combination with line topology coordinate mapping;

[0015] S15. Determine the credibility of each candidate path according to the impedance change rate, node coupling degree, fault density and their corresponding weight coefficients; and take the candidate path with a credibility greater than a set threshold as a potential risk path.

[0016] Preferably, in S2, performing state fluctuation analysis on the real-time state data and historical state data of the monitoring point to determine the state fluctuation factor includes the following steps:

[0017] S211, acquiring voltage data and current data of the monitoring point according to the time series, performing Z-score standardization to obtain a real-time voltage series and a real-time current series;

[0018] S212, comparing the real-time voltage sequence and the real-time current sequence with the historical state data in the same domain in sequence to obtain a voltage fluctuation variance sequence and a current fluctuation variance sequence;

[0019] S213, based on the real-time voltage sequence and the real-time current sequence, calculate the voltage and current anti-correlation in sequence by using the Pearson correlation coefficient method to obtain a voltage and current anti-correlation sequence;

[0020] S214. Calculate the state fluctuation factor corresponding to the monitoring point in the time series according to the voltage fluctuation variance sequence, the current fluctuation variance sequence, the voltage-current anti-correlation sequence, and their corresponding weight coefficients.

[0021] Preferably, in S2, an impedance prediction model is constructed based on the path disturbance data to determine the environmental disturbance factor, including the following steps:

[0022] S221. Map the hidden danger path into a directed graph according to the line topology characteristics, and construct the node set and the directed edge set corresponding to the hidden danger path; divide the hidden danger path into unit grids according to the edge characteristics in the directed edge set, and obtain the path disturbance data of each unit grid; the path disturbance data includes temperature field data, load fluctuation data, and zero-sequence current data.

[0023] S222. Construct the state equation of the time-varying impedance according to the path disturbance data, and construct the observation equation based on Ohm's law; perform iterative calculations on the state equation and the observation equation through the Kalman filter algorithm to obtain the optimal time-varying impedance state equation and use it as the impedance prediction model.

[0024] S223. Obtain the predicted impedance at each moment through the impedance prediction model, and take the difference between the predicted impedance and the measured impedance as the impedance prediction residual; use the principal component analysis method to perform eigen-decomposition on the impedance prediction residual to obtain the temperature disturbance component, the load disturbance component, and the zero-sequence current disturbance component.

[0025] S224. Calculate the environmental disturbance factor of each unit grid according to the temperature characteristic component, the load characteristic component, the zero-sequence current characteristic component, and their corresponding standard deviations.

[0026] Preferably, the state equation is: ;

[0027] The observation equation is: ;

[0028] represents the impedance value of the line at the k-th moment, is the impedance value of the line at the (k - 1)-th moment; is the temperature influence coefficient; is the temperature change amount at the k-th moment relative to the reference moment, is the load influence coefficient; is the load change amount at the k-th moment; is the zero-sequence interference coefficient; is the zero-sequence current phase difference at the k-th moment; is the process noise; is the measured line voltage value at the k-th moment; is the measured line current value at the k-th moment; is the measured noise.

[0029] Preferably, the impedance prediction residual: ;

[0030] Environmental disturbance factor: ;

[0031] wherein, is the impedance prediction residual; is the environmental disturbance factor of the unit grid ; is the temperature change amount at the k-th moment relative to the reference moment; is the load change amount at the k-th moment; is the zero-sequence current phase difference at the k-th moment; , , are the temperature disturbance load coefficient, the load disturbance load coefficient, and the zero-sequence current disturbance load coefficient respectively; are the standard deviations of the temperature disturbance component, the load disturbance component, and the zero-sequence current disturbance component respectively.

[0032] Preferably, the hidden danger path is intensified according to the state fluctuation factor and the environmental disturbance factor to obtain the substation area deterioration map; the method includes the following steps:

[0033] S31. Perform Z-score standardization on the state fluctuation factor and the environmental disturbance factor to obtain a state fluctuation factor sequence and an environmental disturbance factor sequence belonging to the same hidden danger path;

[0034] S32. Configure a corresponding fluctuation trend factor for each state fluctuation factor in the state fluctuation factor sequence and a corresponding disturbance trend factor for each environmental disturbance factor in the environmental disturbance factor sequence;

[0035] S33. Input the standardized state fluctuation factor, the fluctuation trend factor, the standardized environmental disturbance factor, the disturbance trend factor, and the fluctuation disturbance correlation coefficient into the deterioration prediction model to obtain a deterioration degree sequence corresponding to each hidden danger path;

[0036] S34. Combine the deterioration degree sequence with the topological structure of the hidden danger path to obtain the substation area deterioration map.

[0037] Preferably, the degradation prediction model is constructed by a bidirectional LSTM model. The input layer has 5 inputs, and the input parameters of the input layer are the state fluctuation factor, the fluctuation trend factor, the environmental disturbance factor, the disturbance trend factor, and the fluctuation disturbance correlation coefficient. The hidden layer includes a forward LSTM layer and a backward LSTM layer. The output layer has 1 output. The mean square error and the trend penalty mechanism are used as the loss function. The AdamW optimizer is used to optimize the degradation prediction model, and the root mean square error is used as the reliability index for model evaluation.

[0038] Preferably, combining the degradation degree sequence with the topological structure of the hidden danger path to obtain the substation area degradation map includes the following steps:

[0039] S341. Unfold the degradation degree sequence according to the time stamp and map it to a directed graph according to the edge characteristics and node characteristics to form a degradation trajectory.

[0040] S342. Superimpose the degradation values at different times on the degradation trajectory and obtain a spatio-temporal heat map through hierarchical color rendering. Determine the degradation direction and degree in the spatio-temporal heat map through the direction and length of the arrow.

[0041] Preferably, in S4, the substation area degradation map is dynamically cut by the hidden danger risk value corresponding to the monitoring point to determine the degradation block, and the corresponding warning measures are determined according to the degradation degree corresponding to the degradation block. The method includes the following steps:

[0042] S41. Obtain the oxidation characteristics corresponding to the momentary drop events within the time window, and determine the oxidation risk value of each degradation trajectory according to the oxidation characteristics and their corresponding weight factors.

[0043] S42. Obtain the melting characteristics corresponding to the momentary drop events within the time window, and obtain the melting risk value according to the melting characteristic criterion.

[0044] S43. Determine the hidden danger risk value corresponding to the monitoring point according to the oxidation risk value and the melting risk value. Form a multi-dimensional feature vector according to the hidden danger risk value corresponding to the monitoring point in the substation area degradation map and the degradation degree sequence: Dynamically cut the substation area degradation map by combining the multi-dimensional feature vector with the fuzzy C-means clustering method to determine the degradation block.

[0045] S44. Determine the degradation degree of the degradation block according to the degradation block division rule table. Determine the corresponding warning measures according to the degradation degree corresponding to the degradation block.

[0046] In the second aspect, a technical solution provided in the embodiments of the present invention is: a distribution substation abnormal positioning system, applicable to the distribution substation abnormal positioning method based on multi-modal dynamic feature analysis, including:

[0047] Hidden danger preliminary judgment module: Obtain multi-source fusion data of the distribution area according to the sliding time window, and determine the hidden danger path corresponding to the voltage dip event in combination with the line topology;

[0048] First determination module: Analyze the state fluctuation of the real-time state data and historical state data of the monitoring point to determine the state fluctuation factor;

[0049] Second determination module: Construct an impedance prediction model according to the path disturbance data to determine the environmental disturbance factor;

[0050] Construction module: Intensify the hidden danger path according to the state fluctuation factor and the environmental disturbance factor to obtain the degradation map of the distribution area;

[0051] Execution module: Dynamically cut the degradation map of the distribution area through the hidden danger risk value corresponding to the monitoring point to determine the degraded block, and determine the corresponding early warning measures according to the degradation degree corresponding to the degraded block.

[0052] Thirdly, a technical solution provided in an embodiment of the present invention is: An electronic device includes a memory and a processor. A computer program is stored in the memory. When the processor calls the computer program in the memory, the steps of the abnormal positioning method for the distribution area based on multi-modal dynamic feature analysis are implemented.

[0053] Fourthly, a technical solution provided in an embodiment of the present invention is: A storage medium stores computer-executable instructions. When the computer-executable instructions are loaded and executed by a processor, the steps of the abnormal positioning method for the distribution area based on multi-modal dynamic feature analysis are implemented.

[0054] Advantages of the present invention:

[0055] (1) Aiming at the problems of insufficient multi-source data fusion and hidden danger path positioning relying on single-dimensional analysis in the prior art, this application dynamically generates a sliding time window based on the response characteristics of distribution area equipment, distributes and collects multi-source fusion data such as voltage sag waveforms and harmonic distortion rates, and through means such as spatio-temporal alignment, impedance constraint screening, and credibility weighted calculation, overcomes the deficiencies of traditional methods that only rely on a single data source or simple analysis, and significantly improves the accurate positioning ability of hidden danger paths for voltage dip events in complex distribution area environments;

[0056] (2) Aiming at the problems of the traditional method lacking the coupled analysis of the state fluctuations of monitoring points and environmental disturbances, and the static evaluation of deterioration trends, this application standardizes the real-time and historical voltage and current data of monitoring points through Z-score, calculates the variance and anti-correlation of voltage and current fluctuations to obtain the state fluctuation factor. At the same time, an impedance prediction model is constructed based on the Kalman filter, the environmental disturbance components such as temperature, load, and zero-sequence current are decomposed, and the environmental disturbance factor is calculated. Finally, the dual factor and trend characteristics are input into the bidirectional LSTM deterioration prediction model to generate a spatio-temporal thermal deterioration map, which overcomes the defect of the prior art in the comprehensive analysis of the dynamic changes of equipment states and multi-physical quantity disturbances in the external environment, and significantly improves the dynamic quantization evaluation accuracy of the deterioration degree of substation area equipment.

[0057] (3) Aiming at the problems of the existing hidden danger risk level determination being rough, and the lack of real-time and differentiation in early warning measures, this application extracts oxidation characteristics such as voltage difference and low voltage drop, and melting characteristics such as transient drop power and current persistence characteristics, constructs the hidden danger risk value corresponding to the monitoring point including multi-physical quantity parameters, combines the fuzzy C-means clustering algorithm to dynamically cut and divide the deterioration map of the substation area into deterioration blocks, and finally matches the corresponding early warning strategy according to the deterioration degree of the block, which overcomes the deficiencies of the traditional method relying on manual experience and having a single risk assessment dimension, and significantly improves the intelligent level of hidden danger identification at the front end of the monitoring point and the accurate adaptation ability of early warning measures.

[0058] The above invention content is only an overview of the technical solution of the present invention. In order to be able to more clearly understand the technical means of the present invention, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, the following specifically illustrates the specific implementation manners of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] By reading the detailed description of the non-restrictive embodiments with reference to the following drawings, other features, purposes and advantages of the present invention will become more obvious. The drawings are only for the purpose of showing the preferred embodiments, and are not considered as a limitation of the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components.

[0060] Figure 1 It is a flowchart of the abnormal location method for distribution substations based on multi-modal dynamic feature analysis of the present invention.

[0061] Figure 2 It is a flowchart of the hidden danger path generation of the present invention.

[0062] Figure 3 It is a flowchart of the state fluctuation factor generation of the present invention.

[0063] Figure 4 It is a flowchart of the environmental disturbance factor generation of the present invention.

[0064] Figure 5 This is the flowchart for generating the deterioration diagram of the substation area of the present invention.

[0065] Figure 6 This is the structural diagram of the abnormal location system for the distribution substation area of the present invention. Detailed implementation manners

[0066] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific implementation manners described herein are only the best embodiments of the present invention, which are only used to explain the present invention and do not limit the protection scope of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0067] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations (or steps) can be implemented in parallel, concurrently or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but it can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0068] Embodiment 1: At present, traditional methods for abnormal location of distribution substation areas mainly rely on single - type data and relatively simple analysis means. With the access of distributed energy, the intensification of load fluctuations and the complication of environmental factors, traditional methods have exposed significant defects in aspects such as the accuracy of hidden - danger path identification, the response speed of dynamic disturbances, and the location granularity. For example, some methods only judge whether there is an abnormality in the substation area based on electrical quantity data. Since the data source is single, this method cannot comprehensively reflect the actual operation status of the substation area, easily leading to inaccurate abnormal location and omission of some potential hidden dangers. Moreover, traditional methods often focus on the analysis of static data and ignore the dynamic change characteristics during the operation of the substation area. In actual operation, the state of the substation area will be dynamically affected by various factors, such as environmental factors (temperature, humidity, wind speed, etc.) and the random changes in user electricity - consumption behaviors. These dynamic factors will cause the operation parameters of the substation area to show complex fluctuation characteristics, while traditional methods cannot effectively capture these dynamic characteristics and are difficult to accurately identify abnormal situations at an early stage, delaying the best opportunity for fault handling. In addition, existing location methods have great limitations when facing complex low - voltage line topologies. The low - voltage line topology is intricate. When abnormal situations such as momentary voltage dips occur, traditional methods are difficult to quickly and accurately determine the corresponding hidden - danger paths, resulting in time - consuming and laborious troubleshooting of faults, seriously affecting the efficiency of power - supply restoration.

[0069] In view of the deficiencies of the above-mentioned existing technologies, the present application proposes an abnormal location method for a distribution transformer area based on multi-modal dynamic feature analysis. As Figure 1 shown, it includes the following steps:

[0070] S1. Obtain the multi-source fusion data of the transformer area according to the sliding time window and determine the hidden danger path corresponding to the momentary drop event in combination with the line topology.

[0071] As an alternative embodiment, as Figure 2 shown, S1 includes the following steps:

[0072] S11. Dynamically generate a sliding time window based on the response characteristics of the transformer area equipment and the environmental noise coefficient.

[0073] It can be understood that by statistically analyzing the historical data, the action delay of the intelligent switch in the transformer area is 10 minutes, and the minimum step size of the time window is determined = 15 minutes (matching the 15-minute sampling period of the monitoring point); taking the standard deviation of the zero-sequence current phase difference as the noise index, dynamically adjust the length of the time window; ; represents the mean value of the historical zero-sequence current phase difference, which is generally calculated through 30-day historical data. In this embodiment, the length of the time window is dynamically adaptive with the environmental noise, avoiding missed sampling or redundant sampling of high-frequency / low-frequency disturbances caused by a fixed window length; achieving precise matching between the momentary drop event and the equipment response cycle, and improving the pertinence of data acquisition.

[0074] S12. The distributed acquisition terminal obtains the multi-source fusion data of the transformer area according to the sliding time window, and the multi-source fusion data of the transformer area at least includes voltage sag waveform data, harmonic distortion rate data, zero-sequence current phase difference data, and temperature gradient data.

[0075] It can be understood that the voltage sag waveform data includes the voltage sag amplitude and the duration (calculated by counting consecutive monitoring points), and the harmonic distortion rate data generally represents the content of the 3rd harmonic; the zero-sequence current phase difference can be expressed as , where represents the phase angle of the i-th phase current, θ is the average value of the three-phase phase angle. During normal operation, θ < 10°. During abnormal conditions (such as single-phase grounding), θ > 15°. Cable joint temperature: Obtained through distributed optical fiber temperature measurement (DTS). For example, the temperature of the A-phase connection point TA = 75°C, the temperature of the adjacent healthy connection point TB = 50°C, and the gradient difference is 25°C (exceeding the warning threshold of 15°C). By fusing multi-dimensional data such as voltage, current, harmonics, temperature, and phase difference to detect hidden faults (such as joint faults caused by clamp heating), the missed judgment caused by single voltage and current analysis can be avoided, and the recognition rate of hidden faults can be improved. In this embodiment, multi-dimensional data acquisition covers electrical quantities (voltage / current data), harmonic characteristics, three-phase balance degree, and temperature field data, and a three-dimensional hidden danger feature library can be constructed; the one-sidedness of single data dimension analysis is eliminated, providing all-element input for subsequent discrimination of the nature of hidden dangers.

[0076] S13. Perform spatio-temporal alignment operations on the multi-source fusion data of the substation area by using spline interpolation and kernel density estimation to obtain a preliminary selected data set.

[0077] It can be understood that spline interpolation is used to handle missing values in the time series of data. Suppose in the voltage sag waveform data, the data at a certain moment (set as t = 30 minutes) is missing. Using the cubic spline interpolation formula: y(t)=a0+a1(t-tg)+a2(t-tg) 2 +a3(t-tg) 3 (where y(t) is the value at time t after interpolation, tg is the time of the known data point, and a0, a1, a2, a3 are coefficients obtained by solving through the known data points), determine the coefficients through the known adjacent time data points (such as the data points at t = 20 minutes and t = 40 minutes), and thus calculate the voltage value at the missing moment to ensure the continuity of the time series.

[0078] It can be understood that kernel density estimation is used to smooth the data of spatial distribution (such as the measured values of the zero-sequence current phase difference at different positions). Suppose at a certain moment, the zero-sequence current phase differences at different positions in the substation area are measured to obtain a series of data points x1, x2,..., xn. The kernel density estimation formula can be: , where f(x) is the estimated density at position x, n is the number of data points, h is the bandwidth (determined according to the discreteness of the data, set as h = 5° here), and K is the kernel function (such as the Gaussian kernel function). Process the zero-sequence current phase difference data through this formula to make the data smoother and better reflect the true distribution of the data.

[0079] In this embodiment, spline interpolation ensures the integrity of multi-source fusion data in the time dimension, avoiding analysis deviation caused by data loss; kernel density estimation smooths the spatial data, effectively removing the interference of outliers, enabling the data to more accurately reflect the true state of the substation area operation, improving the data quality, and providing a reliable data basis for subsequent precise analysis and screening of potential hidden danger paths.

[0080] S14. Screen the initially selected data set through impedance constraints and momentary dip event criteria, and combine with line topology coordinate mapping to determine the candidate path set.

[0081] It can be understood that impedance constraints screen data by setting a reasonable impedance range. The theoretical range of the known line impedance can be calculated based on parameters such as line material and length. If the impedance value calculated at a certain measurement point exceeds this range, then this data point is regarded as abnormal data and excluded; the momentary dip event criteria include multiple conditions. For example, condition one: the meter voltage difference ΔV > 7V and the measured user voltage V user < 225V; condition two: the current change rate (I high -I low ) / I high > 50%; condition three: the three-phase user voltages do not dip simultaneously. Assume that at a certain moment, the measured data of a certain user shows that the voltage drops from 225V to 210V (ΔV = 15V, V user = 210V), the current rises from 2A to 3.5A (the current change rate (3.5 - 2) / 2 = 75% > 50%), and only this phase among the three phases has a voltage dip, meeting the momentary dip event criteria. Combining with line topology coordinate mapping, associate the data points that meet the impedance constraints and momentary dip event criteria with the line topology diagram of the substation area to determine the line paths that may have hidden dangers, forming a candidate path set.

[0082] In this embodiment, screening the initially selected data set through impedance constraints and momentary dip event criteria can accurately screen out the data related to potential hidden dangers from a large amount of data, remove the interference of invalid data, and narrow the analysis scope; combining with line topology coordinate mapping, achieve accurate positioning from data to actual line paths, providing accurate candidate objects for subsequent determination of the true hidden danger paths, and improving the accuracy and pertinence of hidden danger path positioning.

[0083] S15. Determine the credibility of each candidate path according to the impedance change rate, node coupling degree, fault density, and their corresponding weight coefficients; use the candidate paths with credibility greater than the set threshold as the hidden danger paths.

[0084] It is understandable that the impedance change rate is used to measure the change of line impedance over time, and the node coupling degree reflects the tightness of the connection between line nodes and other nodes or devices. Specifically, the node coupling degree is calculated by the ratio of the sum of the active power and reactive power of two adjacent nodes to the total sum of the active power and reactive power of all nodes in the substation area. The fault density refers to the frequency of faults occurring in a certain line or area within a certain period of time. Weight coefficients are assigned to the impedance change rate, node coupling degree, and fault density of each candidate path, which are set as w1 = 0.4, w2 = 0.3, and w3 = 0.3 respectively. Then the credibility calculation formula for each candidate path is as follows: C = w 1× R rate + w 2× D node + w 3× F density (where C is the credibility, R rate is the impedance change rate, D node is the node coupling degree, F density is the fault density); assuming that the impedance change rate of a certain candidate path is 30%, the node coupling degree is 5, and the fault density is 3 times / year, then its credibility C = 0.4×0.3 + 0.3×5 + 0.3×3 = 2.52. Set a credibility threshold (such as the credibility threshold is set to 2), and determine the candidate paths with credibility greater than this threshold as potential hazard paths.

[0085] In this embodiment, by comprehensively considering the impedance change rate, node coupling degree, fault density, and their weight coefficients to determine the credibility of candidate paths, the potential hazard degree of each candidate path can be quantitatively evaluated. This quantitative evaluation method makes the judgment of potential hazard paths more scientific and accurate, avoiding the errors of subjective judgment; taking the paths with credibility greater than the set threshold as potential hazard paths further focuses on the lines with real potential hazards, improves the accuracy of potential hazard identification, provides a reliable basis for subsequent targeted maintenance measures, reasonably allocates maintenance resources, and improves maintenance efficiency.

[0086] S2. Perform state fluctuation analysis on the real-time state data and historical state data of the monitoring points to determine the state fluctuation factor, and construct an impedance prediction model based on the path disturbance data to determine the environmental disturbance factor.

[0087] As an alternative embodiment, in S2, perform state fluctuation analysis on the real-time state data and historical state data of the monitoring points to determine the state fluctuation factor, as Figure 3 shown, including the following steps:

[0088] S211 , obtaining voltage data and current data of the monitoring point according to the time series, performing Z-score standardization, and obtaining a real-time voltage series and a real-time current series.

[0089] In this embodiment, for example, in a specific distribution station area, a voltage data sequence {V1, V2, ..., Vn} and a current data sequence {I1, I2, ..., In} are obtained from a monitoring point for a period of time (e.g., once every 15 minutes in a day), and the voltage data and current data are standardized using the Z-score standardization method to obtain a real-time voltage sequence and real-time current sequence ; Z-score standardization eliminates the dimensionality of voltage and current data, making data of different magnitudes comparable. This lays the foundation for the subsequent accurate analysis of voltage and current fluctuation characteristics, ensuring that the weights of each data point are consistent when calculating fluctuation variance and correlation, and improving the reliability and accuracy of the analysis results.

[0090] S212, comparing the real-time voltage sequence and the real-time current sequence with the historical state data in the same domain in sequence to obtain the voltage fluctuation variance sequence and the current fluctuation variance sequence.

[0091] In this embodiment, the voltage fluctuation variance sequence and the current fluctuation variance sequence are calculated to quantify the fluctuation degree of the voltage and current at different times relative to the historical data. The variance sequence can reflect the stability change of the voltage and current. The larger the fluctuation variance, the more drastic the change of the voltage or current at that moment compared with the historical data, which provides an important quantitative indicator for judging whether the state (deterioration trend) of the monitoring point is abnormal.

[0092] S213. Based on the real-time voltage sequence and the real-time current sequence, the voltage and current anti-correlation is calculated in sequence by using the Pearson correlation coefficient method to obtain a voltage and current anti-correlation sequence.

[0093] In this embodiment, based on the standardized real-time voltage sequence and real-time current sequence, the Pearson correlation coefficient method is used to calculate the voltage and current anti-correlation. The voltage and current anti-correlation sequence reflects the degree of correlation between the voltage and current change trends. Under normal circumstances, the changes in voltage and current have certain rules, and when abnormalities occur (such as poor line contact, load mutation, etc.), this correlation will change. By analyzing the anti-correlation sequence, potential abnormal situations can be discovered, which supplements the deficiencies of variance analysis based only on voltage or current fluctuations, and improves the ability to detect abnormalities in the state of the monitoring point. It should be noted that the Pearson correlation coefficient method is a data analysis technique that can be mastered by technicians in this field, and will not be described here.

[0094] S214. Calculate the state fluctuation factor corresponding to the monitoring point in the time series according to the voltage fluctuation variance sequence, the current fluctuation variance sequence, the voltage-current anti-correlation sequence, and their corresponding weight coefficients.

[0095] In this embodiment, according to the voltage fluctuation variance sequence , the current fluctuation variance sequence , the voltage-current anti-correlation sequence {r1, r2,..., rn} and their corresponding weight coefficients , , (assuming = 0.4, , = 0.3), calculate the state fluctuation factor fluctuation corresponding to the monitoring point in the time series: ; for example, at a certain moment i, the voltage fluctuation variance = 0.5, the current fluctuation variance = 0.3, and the voltage-current correlation coefficient ri = -0.8, then the state fluctuation factor fluctuation at this moment .

[0096] In this embodiment, the state fluctuation factor comprehensively considers voltage fluctuation, current fluctuation, and their correlation, and comprehensively quantifies the state fluctuation of the monitoring point. By adjusting the weight coefficients, the influence of different factors on the state fluctuation can be highlighted according to the actual situation. This factor can more accurately reflect the operating state of the monitoring point, provide a comprehensive quantitative index for judging whether there are potential hidden dangers at the monitoring point, and improve the accuracy and comprehensiveness of the state assessment of the monitoring point.

[0097] As an alternative embodiment, in S2, construct an impedance prediction model based on the path disturbance data to determine the environmental disturbance factor; as Figure 4 shown, it includes the following steps:

[0098] S221. Map the hidden danger path into a directed graph according to the line topology characteristics, and construct the node set and the directed edge set corresponding to the hidden danger path; divide the hidden danger path into unit grids according to the edge characteristics in the directed edge set, and obtain the path disturbance data of each unit grid; the path disturbance data includes temperature field data, load fluctuation data, and zero-sequence current data.

[0099] In this embodiment, taking a certain determined hidden danger path as an example, a directed graph mapping is performed on it according to the line topology characteristics. Each connection point on this path is defined as a node, and a node set N = {n1, n2,..., ns} is constructed; the line segments connecting the nodes are defined as directed edges, and a directed edge set E = {e1, e2,..., et} is constructed. According to the characteristics of the directed edges such as length and material, the hidden danger path is divided into multiple unit grids, and multi-dimensional path perturbation data of each unit grid is obtained, realizing the refined analysis of the hidden danger path. These perturbation data reflect the real-time operating conditions on the path from different physical quantity perspectives, providing rich input information for constructing the impedance prediction model subsequently, helping to more accurately analyze the influence of environmental factors on the line impedance, and further enhancing the ability to identify potential hidden dangers.

[0100] S222. Construct a state equation of time-varying impedance based on the path perturbation data, and construct an observation equation based on Ohm's law; perform iterative calculations on the state equation and the observation equation through the Kalman filtering algorithm to obtain the optimal time-varying impedance state equation and use it as the impedance prediction model.

[0101] Further, the state equation is: ; the observation equation is: ; represents the impedance value of the line at the k-th moment, is the impedance value of the line at the (k - 1)-th moment; is the temperature influence coefficient; is the temperature change amount at the k-th moment relative to the reference moment, is the load influence coefficient; is the load change amount at the k-th moment; is the zero-sequence interference coefficient; is the zero-sequence current phase difference at the k-th moment; is the process noise; is the measured line voltage value at the k-th moment; is the measured line current value at the k-th moment; is the measurement noise.

[0102] In this embodiment, the constructed impedance prediction model comprehensively considers the influence of various factors such as temperature, load, and zero-sequence current on the line impedance, and optimizes the model through the Kalman filtering algorithm, enabling it to more accurately predict the change of the line impedance. This model can dynamically track the real-time change of the line impedance, providing a reliable basis for subsequent analysis of the influence of environmental disturbances on the line, improving the prediction ability of the line operating state, and helping to detect potential impedance abnormal problems in advance.

[0103] S223. Obtain the predicted impedance at each moment through the impedance prediction model, and use the difference between the predicted impedance and the measured impedance as the impedance prediction residual; adopt the principal component analysis method to perform eigen-decomposition on the impedance prediction residual to obtain the temperature disturbance component, the load disturbance component, and the zero-sequence current disturbance component.

[0104] Further, use the established impedance prediction model to obtain the predicted impedance at each moment, and use the difference between the predicted impedance and the measured impedance as the impedance prediction residual; impedance prediction residual: ; where is the impedance prediction residual; is the environmental disturbance factor of the unit grid ; is the temperature change amount at the k-th moment relative to the reference moment; is the load change amount at the k-th moment; is the zero-sequence current phase difference at the k-th moment; , , are the temperature disturbance load coefficient, the load disturbance load coefficient, and the zero-sequence current disturbance load coefficient respectively;

[0105] In this embodiment, by calculating the impedance prediction residual and performing eigen-decomposition on the impedance prediction residual using the principal component analysis method, the influence degrees of different factors on the impedance change can be separated. It helps to clarify whether factors such as temperature, load, or zero-sequence current cause the abnormal change of the impedance, provides detailed information for in-depth analysis of the influence of environmental disturbance on the line impedance, improves the analysis ability of the cause of the line impedance change, and thus more accurately determines the source of potential hidden dangers.

[0106] S224. Calculate the environmental disturbance factor of each unit grid according to the temperature characteristic component, the load characteristic component, the zero-sequence current characteristic component, and their corresponding standard deviations.

[0107] In this embodiment, calculate the environmental disturbance factor of each grid unit according to the temperature characteristic component, the load characteristic component, the zero-sequence current characteristic component, and their corresponding standard deviations: ; is the environmental disturbance factor of the unit grid ; are the standard deviations of the temperature disturbance component, the load disturbance component, and the zero-sequence current disturbance component respectively. For example, in a certain unit grid, the temperature disturbance load coefficient , the standard deviation of the temperature disturbance component ; the load disturbance load coefficient , the standard deviation of the load disturbance component ; the zero-sequence current disturbance load coefficient , the standard deviation of the zero-sequence current disturbance component , then the environmental disturbance factor of the unit grid is .

[0108] In this embodiment, the environmental disturbance factor comprehensively considers the disturbance degree of factors such as temperature, load, and zero-sequence current. By calculating the environmental disturbance factor, the comprehensive impact of environmental disturbance on each unit grid can be quantified. This provides a unified quantitative index for evaluating the degree of interference of environmental factors on the line at different locations, helps to determine the areas that are most affected by the environment, provides a clear direction for key monitoring and hidden danger investigation, and improves the ability to locate and evaluate potential hidden dangers.

[0109] S3. According to the state fluctuation factor and the environmental disturbance factor, the potential danger path is intensified to obtain the substation degradation map.

[0110] As an optional embodiment, Figure 5 As shown, S3 includes the following steps:

[0111] S31. Perform Z-score standardization on the state fluctuation factor and the environmental disturbance factor to obtain the state fluctuation factor sequence and the environmental disturbance factor sequence belonging to the same hidden danger path.

[0112] In this embodiment, the state fluctuation factor sequence is assumed to be , environmental disturbance factor sequence The state fluctuation factor is standardized by Z-score to obtain , the environmental disturbance factor is standardized by Z-score to obtain Through Z-score standardization, the dimensional differences of state fluctuation factors and environmental disturbance factors are eliminated, making data of different magnitudes comparable. This ensures that in subsequent analysis, each factor can be comprehensively considered on a unified scale, avoiding the excessive or neglected impact of certain factors due to dimensional issues, and provides a basis for accurately analyzing the degradation of hidden danger paths.

[0113] S32, configuring a corresponding fluctuation trend factor for each state fluctuation factor in the state fluctuation factor sequence and configuring a corresponding disturbance trend factor for each environmental disturbance factor in the environmental disturbance factor sequence.

[0114] In this embodiment, the change trend of the state fluctuation factor is determined by calculating the change rate of the state fluctuation factor within adjacent unit time periods. Similarly, the change trend of the environmental disturbance factor is determined by calculating the change rate of the environmental disturbance factor within adjacent unit time periods. By configuring the fluctuation trend factor and the disturbance trend factor, a quantitative basis is provided for analyzing the dynamic changes of state fluctuations and environmental disturbances. These trend factors can capture the change rate of the factors over time, helping to discover potential change trends, such as determining whether the state fluctuation or environmental disturbance is gradually intensifying or tending to be stable. This is of great significance for predicting the future deterioration trend of the hidden danger path, making the assessment of the state of the substation area equipment more forward-looking.

[0115] S33. Input the standardized state fluctuation factor, fluctuation trend factor, standardized environmental disturbance factor, disturbance trend factor, and fluctuation disturbance correlation coefficient into the deterioration prediction model to obtain the deterioration degree sequence corresponding to each hidden danger path.

[0116] As an alternative embodiment, the deterioration prediction model is constructed by a bidirectional LSTM model. There are 5 input layers, and the state fluctuation factor, fluctuation trend factor, environmental disturbance factor, disturbance trend factor, and fluctuation disturbance correlation coefficient are used as the input parameters of the input layer. The hidden layer includes a forward LSTM layer and a backward LSTM layer. There is 1 output layer, which outputs the deterioration degree sequence {D1, D2,..., Dm} corresponding to each hidden danger path. The mean square error and the trend penalty mechanism are used as the loss function. The AdamW optimizer is used to optimize the deterioration prediction model, and the root mean square error is used as the reliability index for model evaluation.

[0117] In this embodiment, the fluctuation disturbance correlation coefficient Corr is obtained by calculating the correlation between the standardized state fluctuation factor sequence and the environmental disturbance factor sequence. The calculation method of the correlation degree belongs to the conventional technology mastered by those skilled in the art and will not be elaborated here. During model training, the mean square error MSE (the mean square error between the actual deterioration degree and the predicted deterioration degree) and the trend penalty mechanism are used as the loss function. The AdamW optimizer is used to optimize the model, and the root mean square error RMSE is used to evaluate the reliability of the model. This embodiment uses a bidirectional LSTM model with multi-feature input to predict the deterioration degree sequence, fully considering various factors such as state fluctuations, environmental disturbances, their change trends, and correlations. Through the learning and prediction of the model, the change of the deterioration degree of each hidden danger path over time can be more accurately evaluated, providing strong support for formulating targeted maintenance strategies in the future and improving the accuracy and reliability of predicting the deterioration state of the substation area equipment.

[0118] S34. Combine the deterioration degree sequence with the topological structure of the hidden danger path to obtain the substation area deterioration map.

[0119] As an alternative embodiment, combining the deterioration degree sequence with the topological structure of the potential hazard path to obtain the substation area deterioration map includes the following steps:

[0120] S341. Expand the deterioration degree sequence according to the time stamp and map it onto a directed graph according to the edge characteristics and node features to form a deterioration trajectory;

[0121] S342. Superimpose the deterioration values at different times on the deterioration trajectory and obtain a spatio-temporal heat map through hierarchical color rendering; determine the deterioration direction and degree in the spatio-temporal heat map through the direction and length of the arrow.

[0122] It can be understood that the deterioration degree sequence {D1, D2,..., Dm} is expanded according to the time stamp, and the time stamps are assumed to be t1, t2,..., tm. The deterioration degree is mapped onto a directed graph according to the edge characteristics (such as line length, material, etc.) and node features (such as the number of devices connected to the node, importance, etc.). For example, if an edge connects two important nodes and the deterioration degree corresponding to this edge is relatively high at a certain moment, mark the corresponding deterioration information at the position of this edge on the directed graph to form a deterioration trajectory. In S342, the deterioration values at different times on the deterioration trajectory are superimposed. For example, the deterioration value of a certain node at time t1 is D1, and at time t2 is D2. After superimposition, the comprehensive deterioration value D = D1 + D2 of this node at these two times is obtained. Through hierarchical color rendering, for example, the deterioration values are divided into three levels: low, medium, and high, which can be represented by green, yellow, and red respectively. At the same time, determine the arrow direction and length according to the change direction and amplitude of the deterioration value. The arrow direction represents the deterioration trend (such as the direction from low deterioration to high deterioration), and the length represents the change amplitude of the deterioration degree.

[0123] In this embodiment, combining the deterioration degree sequence with the topological structure of the potential hazard path to construct the substation area deterioration map visually displays the deterioration situation of the substation area equipment in a visual way. Through the deterioration trajectory, spatio-temporal heat map and arrow indication, the maintenance personnel can quickly understand the distribution, change trend and development direction of the deterioration degree at different positions in the substation area, which is convenient for visually identifying high-risk areas, provides a clear basis for reasonably arranging maintenance plans and optimizing resource allocation, and improves the efficiency and pertinence of operation and maintenance management.

[0124] S4. Dynamically cut the substation area deterioration map through the potential hazard risk value corresponding to the monitoring point to determine the deterioration block, and determine the corresponding warning measures according to the deterioration degree corresponding to the deterioration block.

[0125] As an alternative embodiment, S4 includes the following steps:

[0126] S41. Obtain the oxidation characteristics corresponding to the instantaneous drop events within the time window, and determine the oxidation risk value of each deterioration trajectory according to the oxidation characteristics and their corresponding weight factors; wherein, the oxidation characteristics include voltage difference, low voltage drop, theoretical resistance, and resistance change rate.

[0127] In this embodiment, within a specific time window (such as the past 24 hours), for the instantaneous drop events occurring in the power distribution area, the oxidation characteristics are extracted. The voltage difference ΔV is the difference in voltage before and after the instantaneous drop, and the low voltage drop refers to the drop amplitude below the normal voltage (such as 220V). The theoretical resistance is the ratio of the voltage difference to the current difference before and after the instantaneous drop event. The resistance change rate is the ratio of the difference between the current resistance and the resistance in the same historical time window to the historical resistance. Let the weight factors corresponding to the voltage difference, low voltage drop, theoretical resistance, and resistance change rate be g1, g2, g3, g4 respectively (for example, g1 = 0.3, g2 = 0.2, g3 = 0.3, g4 = 0.2). . Suppose in a certain instantaneous drop event, the voltage drops from 220V to 200V, then the voltage difference ΔV = 20V, and the low voltage drop is 20V. If the current changes from 5A to 6A, then the theoretical resistance .

[0128] By comprehensively considering the oxidation characteristics and their weight factors to calculate the oxidation risk value in this embodiment, the potential hidden danger degree of the line due to oxidation can be quantified. This quantification method integrates multiple electrical parameters related to oxidation, comprehensively reflects the impact of oxidation on the line state, provides a comprehensive index for evaluating the potential oxidation risk of the power distribution area line, and makes the evaluation of oxidation hidden dangers more scientific and accurate.

[0129] S42. Obtain the melting characteristics corresponding to the instantaneous drop events within the time window, and obtain the melting risk value according to the melting characteristic criterion; the melting characteristics include instantaneous drop power, voltage-current anti-correlation, current persistence characteristic, and recovery period electrical parameter characteristic.

[0130] In this embodiment, within the same time window, the melting characteristics are obtained for the instantaneous drop events. The instantaneous drop power , where is the lowest current value during the voltage instantaneous drop process. The voltage-current anti-correlation is measured by calculating the Pearson correlation coefficient CorrVI of voltage and current. When CorrVI is less than a certain threshold (such as -0.5), it is considered that there is voltage-current anti-correlation. The current persistence characteristic refers to whether the current continuously remains at a certain level (such as I and the duration exceeds a certain length, assumed to be 10 minutes) during the voltage recovery process. The recovery period electrical parameter characteristics include whether the voltage V and current I at multiple measurement points satisfy specific conditions (such as I ). Assume a certain voltage dip event where ΔV = 20V, , then the voltage dip power , and the calculated CorrVI = -0.6, which meets the condition of voltage - current inverse correlation. During the voltage recovery process, the current remains at 3.5A for 10 minutes ( ), which meets the current persistence characteristic. Within 24 hours after the voltage dip ends, the average current of multiple measurement points I = 5A (5A > 1.2×4A), and the average voltage V = 210V ( , assume ), which meets the electrical parameter characteristics during the recovery period. The melting characteristic criterion is set such that when all the above - mentioned melting characteristics are met, the melting risk value melts ; otherwise .

[0131] In this embodiment, obtaining the melting risk value based on the melting characteristic criterion can specifically identify potential hidden dangers of melting caused by overheating in the circuit. Through the combined judgment of multiple key characteristics, the detection accuracy of melting hidden dangers is improved, avoiding misjudgment or missed judgment caused by single - characteristic judgment, and providing an effective means for timely discovering and handling melting hidden dangers that may cause serious damage to the circuit.

[0132] S43. Determine the hidden - danger risk value corresponding to the monitoring point according to the oxidation risk value and the melting risk value; form a multi - dimensional feature vector based on the hidden - danger risk value corresponding to the monitoring point in the sub - station degradation map and the degradation degree sequence: Dynamically cut the sub - station degradation map by using the fuzzy C - means clustering method combined with the multi - dimensional feature vector to determine the degraded blocks.

[0133] In this embodiment, the hidden - danger risk value corresponding to the monitoring point . Combining the hidden - danger risk value corresponding to the monitoring point in the sub - station degradation map with the previously obtained degradation degree sequence {D1, D2,..., Dm}, a multi - dimensional feature vector {D total, D1, D2,..., Dm} is formed. The fuzzy C - means clustering method is a partitioning - based clustering algorithm. By continuously iteratively optimizing the objective function, the membership degree and the cluster centers are continuously updated, and finally the sub - station degradation map is dynamically cut into different degraded blocks. The fuzzy C - means clustering method is a conventional technique mastered by those skilled in the art and will not be elaborated here.

[0134] In this embodiment, by combining the oxidation risk value and the melting risk value to obtain the hidden - danger risk value corresponding to the monitoring point, constructing a multi - dimensional feature vector with the degradation degree sequence, and using the fuzzy C - means clustering method for dynamic cutting, the sub - station can be automatically divided into degraded blocks with similar characteristics according to the comprehensive risk and degradation degree of different regions. This division method considers multiple factors, can more reasonably reflect the actual risk situation of different regions in the sub - station, and provides a clear regional division basis for subsequent targeted management and maintenance.

[0135] S44. Determine the deterioration degree of the deteriorated block according to the deterioration block division rule table; determine the corresponding early warning measures according to the deterioration degree corresponding to the deteriorated block.

[0136] In this embodiment, according to the pre-established deterioration block division rule table, determine the deterioration degree of each deteriorated block. For example, when the hidden danger risk value corresponding to the monitoring point of the deteriorated block is greater than a certain high-risk threshold (such as ), and the average deterioration value in the deterioration degree sequence exceeds a certain standard, it is determined as a high deterioration degree; when is in a certain medium-risk interval (such as ), and the deterioration degree is moderate, it is determined as a medium deterioration degree; when is less than a certain low-risk threshold (such as ), and the deterioration degree is low, it is determined as a low deterioration degree.

[0137] Furthermore, corresponding early warning measures are formulated for different deterioration degrees. For example, for the block with a high deterioration degree, take the measure of immediate power outage for maintenance; for the block with a medium deterioration degree, arrange key monitoring and pre-maintenance in the near future; for the block with a low deterioration degree, conduct regular inspections and status tracking.

[0138] In this embodiment, corresponding early warning measures are formulated according to the deterioration degree of the deteriorated block, realizing the hierarchical management of potential hazards in the substation area. This differentiated early warning strategy can reasonably allocate resources according to the severity of the risk, give priority to dealing with high-risk areas, and effectively avoid the occurrence or expansion of faults. At the same time, different management methods are adopted for areas with different risk levels, improving the pertinence and efficiency of operation and maintenance management, and ensuring the stable operation of the substation area.

[0139] Embodiment 2: Another technical solution provided in the embodiments of the present invention is: a distribution substation area anomaly positioning system, applicable to the distribution substation area anomaly positioning method based on multi-modal dynamic feature analysis, as Figure 6 shown, including:

[0140] Hidden danger preliminary judgment module 101: Obtain the multi-source fusion data of the substation area according to the sliding time window and determine the hidden danger path corresponding to the momentary voltage drop event in combination with the line topology;

[0141] First determination module 102: Perform state fluctuation analysis on the real-time state data and historical state data of the monitoring point to determine the state fluctuation factor;

[0142] Second determination module 103: Construct an impedance prediction model according to the path disturbance data to determine the environmental disturbance factor;

[0143] Construction module 104: performing intensification processing on the hidden danger path according to the state fluctuation factor and the environmental disturbance factor to obtain a substation degradation map;

[0144] Execution module 105: dynamically cut the degradation map of the substation area according to the hidden danger risk value corresponding to the monitoring point to determine the degraded block, and determine the corresponding early warning measures according to the degradation degree corresponding to the degraded block.

[0145] Embodiment 3: A technical solution also provided by the embodiment of the present invention is: an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of the method for locating abnormalities in distribution station areas based on multimodal dynamic feature analysis are implemented.

[0146] Embodiment 4: A technical solution provided in the embodiment of the present invention is: a storage medium, in which computer executable instructions are stored. When the computer executable instructions are loaded and executed by a processor, the steps of the distribution station area abnormality positioning method based on multimodal dynamic feature analysis are implemented.

[0147] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the specific device can be divided into different functional modules to complete all or part of the functions described above.

[0148] In the embodiments provided in the present application, it should be understood that the disclosed structures and methods can be implemented in other ways. For example, the embodiments of the structure described above are only schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another structure, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, structures or units, which can be electrical, mechanical or other forms.

[0149] The units described as separate components may or may not be physically separated, and the components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple different places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0150] In addition, each functional unit in the embodiments of the present application may be integrated into a processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0151] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The software product is stored in a storage medium and includes several instructions for causing a device (which may be a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods in the various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0152] The above-described specific implementation manners are the preferred implementation manners of the abnormal location method and system for a distribution transformer area based on multi-modal dynamic feature analysis of the present invention, and do not limit the specific scope of the present invention thereby. The scope of the present invention includes but is not limited to this specific implementation manner. All equivalent changes made according to the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A method for abnormal location of a distribution transformer area based on multi-modal dynamic feature analysis, characterized in that It includes the following steps: S1. Obtain the multi-source fusion data of the distribution area according to the sliding time window, and combine the line topology to determine the hidden danger path corresponding to the momentary voltage dip event; S2. Conduct state fluctuation analysis on the real-time state data and historical state data of the monitoring points to determine the state fluctuation factor, divide the hidden danger path into unit grids, obtain the path disturbance data of each unit grid, and construct an impedance prediction model according to the path disturbance data to determine the environmental disturbance factor; S3. Carry out intensification processing on the hidden danger path according to the state fluctuation factor and the environmental disturbance factor to obtain the deterioration map of the distribution area; S4. Dynamically cut the deterioration map of the distribution area through the hidden danger risk value corresponding to the monitoring point to determine the deteriorated block, and determine the corresponding warning measures according to the deterioration degree corresponding to the deteriorated block; In S2, conducting state fluctuation analysis on the real-time state data and historical state data of the monitoring points to determine the state fluctuation factor includes the following steps: S211. Obtain the voltage data and current data of the monitoring points according to the time series, and perform Z-score standardization to obtain the real-time voltage sequence and the real-time current sequence; S212. Compare and calculate the real-time voltage sequence and the real-time current sequence with the historical state data in the same time domain in turn to obtain the voltage fluctuation variance sequence and the current fluctuation variance sequence; S213. Based on the real-time voltage sequence and the real-time current sequence, calculate the voltage-current anti-correlation in turn through the Pearson correlation coefficient method to obtain the voltage-current anti-correlation sequence; S214. Calculate the state fluctuation factor corresponding to the monitoring point in the time series according to the voltage fluctuation variance sequence, the current fluctuation variance sequence, the voltage-current anti-correlation sequence and their corresponding weight coefficients; The path disturbance data includes: temperature field data, load fluctuation data and zero-sequence current data.

2. The abnormal location method for the distribution area based on multi-modal dynamic feature analysis according to claim 1, characterized in that The step of obtaining the multi-source fusion data of the distribution area according to the sliding time window and combining the line topology to determine the hidden danger path corresponding to the momentary voltage dip event; includes the following steps: S11. Dynamically generate a sliding time window based on the response characteristics of the distribution area equipment and the environmental noise coefficient; S12. The distributed acquisition terminal obtains the multi-source fusion data of the distribution area according to the sliding time window, and the multi-source fusion data of the distribution area at least includes voltage sag waveform data, harmonic distortion rate data, zero-sequence current phase difference data and temperature gradient data; S13. Perform spatio-temporal alignment operation on the multi-source fusion data of the distribution area by using spline interpolation and kernel density estimation to obtain the primary selection data set; S14. Screen the primary selection data set through impedance constraints and momentary voltage dip event criteria, and combine the line topology coordinate mapping to determine the candidate path set; S15. Determine the credibility of each candidate path according to the impedance change rate, node coupling degree, fault density and their corresponding weight coefficients; and use the candidate path with a credibility greater than the set threshold as the hidden danger path.

3. The abnormal location method for the distribution area based on multi-modal dynamic feature analysis according to claim 1, characterized in that In S2, constructing an impedance prediction model according to the path disturbance data to determine the environmental disturbance factor; includes the following steps: S221. Map the hidden danger path into a directed graph according to the line topology characteristics, and construct the node set and directed edge set corresponding to the hidden danger path; divide the hidden danger path into unit grids according to the edge characteristics in the directed edge set, and obtain the path perturbation data of each unit grid. S222. Construct the state equation of time-varying impedance according to the path perturbation data, and construct the observation equation based on Ohm's law; perform iterative calculations on the state equation and the observation equation through the Kalman filtering algorithm to obtain the optimal time-varying impedance state equation and use it as the impedance prediction model. S223. Obtain the predicted impedance at each moment through the impedance prediction model, and take the difference between the predicted impedance and the measured impedance as the impedance prediction residual; use the principal component analysis method to perform eigen-decomposition on the impedance prediction residual to obtain the temperature perturbation component, load perturbation component, and zero-sequence current perturbation component. S224. Calculate the environmental perturbation factor of each unit grid according to the temperature characteristic component, load characteristic component, zero-sequence current characteristic component, and their corresponding standard deviations.

4. The abnormal location method for a distribution transformer substation based on multi-modal dynamic feature analysis according to claim 3, wherein The state equation is as follows: ; The observation equation is as follows: ; represents the impedance value of the line at the k-th moment, is the impedance value of the line at the (k - 1)-th moment; is the temperature influence coefficient; is the temperature change amount at the k-th moment relative to the reference moment, is the load influence coefficient; is the load change amount at the k-th moment; is the zero-sequence interference coefficient; is the zero-sequence current phase difference at the k-th moment; is the process noise; is the line voltage value measured at the k-th moment; is the line current value measured at the k-th moment; is the measurement noise.

5. The abnormal location method for a distribution transformer substation based on multi-modal dynamic feature analysis according to claim 3, wherein Impedance prediction residual: ; Environmental disturbance factor: ; Among them, is the impedance prediction residual; is the environmental disturbance factor of the unit grid ; is the temperature change amount at the k-th moment relative to the reference moment; is the load change amount at the k-th moment; is the zero-sequence current phase difference at the k-th moment; , , are the temperature disturbance load coefficient, the load disturbance load coefficient, and the zero-sequence current disturbance load coefficient, respectively; are the standard deviations of the temperature disturbance component, the load disturbance component, and the zero-sequence current disturbance component, respectively.

6. The abnormal location method for a distribution transformer area based on multimodal dynamic feature analysis according to claim 1, wherein The step of intensifying the hidden danger path according to the state fluctuation factor and the environmental perturbation factor to obtain the substation deterioration map; includes the following steps: S31. Perform Z-score standardization on the state fluctuation factor and the environmental perturbation factor to obtain the state fluctuation factor sequence and the environmental perturbation factor sequence belonging to the same hidden danger path. S32. Configure the corresponding fluctuation trend factor for each state fluctuation factor in the state fluctuation factor sequence and the corresponding perturbation trend factor for each environmental perturbation factor in the environmental perturbation factor sequence. S33. Input the standardized state fluctuation factor, fluctuation trend factor, standardized environmental perturbation factor, perturbation trend factor, and fluctuation perturbation correlation coefficient into the deterioration prediction model to obtain the deterioration degree sequence corresponding to each hidden danger path. S34. Combine the deterioration degree sequence with the topological structure of the hidden danger path to obtain the substation deterioration map.

7. The abnormal location method for a distribution transformer substation based on multi-modal dynamic feature analysis according to claim 6, wherein The deterioration prediction model is constructed by a bidirectional LSTM model, and there are 5 input layers, and the state fluctuation factor, fluctuation trend factor, environmental perturbation factor, perturbation trend factor, and fluctuation perturbation correlation coefficient are used as the input parameters of the input layer. The hidden layer includes a forward LSTM layer and a backward LSTM layer; there is 1 output layer. Use the mean square error and the trend penalty mechanism as the loss function; use the AdamW optimizer to optimize the deterioration prediction model and use the root mean square error as the model evaluation reliability index.

8. The abnormal location method for a distribution transformer substation based on multi-modal dynamic feature analysis according to claim 1, wherein Combining the deterioration degree sequence with the topological structure of the hidden danger path to obtain the substation deterioration map, includes the following steps: S341, expanding the degradation degree sequence according to the timestamp and mapping it onto a directed graph according to the edge characteristics and node characteristics to form a degradation trajectory; S342, superimposing the degradation values at different times on the degradation trajectory and obtaining a spatiotemporal heat map through graded color rendering; determining the degradation direction and degree in the spatiotemporal heat map through the direction and length of the arrow.

9. The method for locating abnormalities in distribution station areas based on multi-modal dynamic feature analysis according to claim 1, characterized in that: S4. Dynamically cut the degradation map of the substation area according to the risk value of hidden dangers corresponding to the monitoring point to determine the degradation block, and determine the corresponding early warning measures according to the degree of degradation corresponding to the degradation block; including the following steps: S41, obtaining oxidation characteristics corresponding to the instantaneous drop event within the time window, and determining the oxidation risk value of each degradation trajectory by weighted summation calculation based on the oxidation characteristics and their corresponding weight factors; the oxidation characteristics include: voltage difference, low voltage drop amplitude, theoretical electric group, and electric group change rate; S42, obtaining melting characteristics corresponding to the instantaneous drop event within the time window, and obtaining a melting risk value according to a melting characteristic criterion; the melting characteristics include instantaneous drop power, voltage-current anti-correlation, current continuity characteristics, and recovery period electrical parameter characteristics; the melting characteristic criterion is set such that when the threshold conditions of all melting characteristics are met, the melting risk value is 1; otherwise, the melting risk value is 0; S43, determining the hidden danger risk value corresponding to the monitoring point according to the oxidation risk value and the melting risk value; forming a multidimensional feature vector according to the hidden danger risk value corresponding to the monitoring point in the substation degradation map and the degradation degree sequence: dynamically cutting the substation degradation map to determine the degradation block through the fuzzy C-means clustering method combined with the multidimensional feature vector; S44. Determine the degradation degree of the degraded block according to the degraded block division rule table; and determine corresponding early warning measures according to the degradation degree corresponding to the degraded block.

10. A distribution transformer area anomaly location system, applicable to the distribution transformer area anomaly location method based on multi-modal dynamic feature analysis as described in any one of claims 1 to 9, characterized in that, include: Hidden danger preliminary judgment module: obtain multi-source fusion data of the substation area according to the sliding time window and determine the hidden danger path corresponding to the instantaneous drop event in combination with the line topology; The first determination module: performs state fluctuation analysis based on the real-time state data and historical state data of the monitoring point to determine the state fluctuation factor; The second determination module: constructs an impedance prediction model based on the path disturbance data to determine the environmental disturbance factor; Construction module: According to the state fluctuation factor and environmental disturbance factor, the potential danger path is intensified to obtain the degradation map of the substation area; Execution module: Dynamically cut the degradation map of the substation to determine the degraded blocks through the hidden danger risk values corresponding to the monitoring points, and determine the corresponding early warning measures according to the degree of degradation corresponding to the degraded blocks.

11. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, it implements the steps of the distribution station area anomaly locating method based on multimodal dynamic feature analysis as described in any one of claims 1 to 9.

12. A storage medium, characterized in that: The storage medium stores computer executable instructions, which, when loaded and executed by the processor, implement the steps of the method for locating abnormalities in distribution station areas based on multimodal dynamic feature analysis as described in any one of claims 1 to 9.

Images