A method for intelligent fault response in power distribution network

Abstract

The present invention relates to the field of fault diagnosis technology, specifically to an intelligent fault response method for a power distribution network, comprising the following steps: collecting operating data of distribution lines and transformers, recording power flow and load changes, integrating data using timestamps to generate monitoring records, setting a multi-scale analysis framework, dynamically selecting a time window, segmenting data and extracting key features, performing waveform comparison, identifying abnormal waveforms through shape basis fitting analysis, deploying adaptive classification processing, analyzing the cause of the fault, and performing classified response. In the present invention, by real-time monitoring of the power flow and load changes of distribution lines and transformers, data updates are made more timely, a multi-scale analysis framework is introduced, the flexibility and comprehensiveness of data processing are enhanced, waveform comparison is performed through shape basis fitting analysis, the accuracy of identifying abnormal patterns is improved, and adaptive classification processing provides targeted response decisions according to abnormal patterns, which significantly improves the efficiency of power grid fault response.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to an intelligent fault response method for a power distribution network. Background Art

[0002] The field of fault diagnosis technology covers a range of methods and tools designed to quickly and accurately detect, identify and analyze abnormal operating conditions of systems or equipment. These technologies are widely used in various industrial and power systems, especially in power distribution systems, where fault diagnosis technology is essential to ensure the stable operation and safety of the power grid. This technology usually includes methods such as signal processing, pattern recognition, machine learning and data mining to predict and diagnose potential faults by analyzing data obtained from sensors and system operations. This can not only prevent damage caused by faults in advance, but also optimize maintenance plans and reduce unnecessary maintenance costs.

[0003] Among them, the intelligent fault response method of power distribution network refers to the technology that uses advanced information technology and artificial intelligence algorithms to monitor, quickly locate and effectively respond to faults in the power system in real time. This method integrates a large amount of power grid operation data and environmental information, and uses intelligent algorithms such as machine learning to analyze fault signals to accurately determine the type and location of the fault. Its main uses include improving the reliability and efficiency of the power distribution network, reducing the power outage time caused by faults, and optimizing the allocation and use of power resources.

[0004] Existing technologies have shown limitations in real-time data processing and accurate fault location. Fixed time windows and limited data processing frameworks result in insensitive or inaccurate identification and classification of fault signals, especially in complex and dynamically changing power systems. Static processing methods fail to effectively adapt to the rapid changes in power demand and the variability of environmental conditions. Especially in extreme weather conditions, fault diagnosis and response may be delayed due to insufficiently timely or comprehensive data processing. Fault analysis without adaptive mechanisms cannot provide the most optimized maintenance and response strategies based on different fault types, increasing the risk and maintenance costs of power grid operation. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent fault response method for a power distribution network.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a method for intelligent fault response of power distribution network, comprising the following steps:

[0007] S1: Collects real-time operation data from distribution lines and transformers, records changes in power flow and power load, integrates data through timestamps, and generates power grid data monitoring records;

[0008] S2: Setting a multi-scale analysis framework, dynamically selecting a time window according to the power grid data monitoring record, segmenting the data of cables and distribution lines, extracting key data features within multiple time windows, and generating multi-scale data segments;

[0009] S3: performing waveform comparison, performing shape basis fitting analysis on each segment in the multi-scale data segment, identifying and clustering abnormal waveforms, extracting fault features from the signal received by the fault indicator, and generating abnormal pattern recognition results;

[0010] S4: deploy adaptive classification processing, analyze and distinguish the causes of faults according to the abnormal pattern recognition results, classify equipment failures, operating errors or external influences, match the current fault mode to respond, and generate fault classification and response decisions.

[0011] As a further solution of the present invention, the steps of acquiring the power grid data monitoring record are specifically as follows:

[0012] S111: Use power sensors to measure the real-time voltage and current of distribution lines and transformers, and obtain the power factor angle from the system using the formula:

[0013]

[0014] Generate electricity flow;

[0015] Among them, V represents voltage, I represents current, φ represents the square root of the power factor, and F represents the power flow;

[0016] S112: Add a timestamp to the power flow, and record the result as a corresponding time-power flow pair (t, F t ), using the formula:

[0017]

[0018] Get the marked power flow;

[0019] Among them, t r Represents the timestamp, F t represents the power flow after the mark;

[0020] S113: For the marked power flow, a time series analysis method is used to perform data integration to identify the change trend of the power load, using the formula:

[0021]

[0022] Calculate and generate weighted average power flow;

[0023] in, represents the timestamped power flow, i r represents the index in the time series, F avg represents the weighted average power flow;

[0024] S114: According to the weighted average power flow, the recorded data format is adjusted, using the formula:

[0025]

[0026] Convert data units to kilowatt-hours and generate power grid data monitoring records;

[0027] Wherein, M represents the power grid data monitoring record.

[0028] As a further solution of the present invention, the step of acquiring the multi-scale data segment is specifically as follows:

[0029] S211: According to the power grid data monitoring record, a multi-scale analysis framework is set up, using the formula:

[0030]

[0031] Calculate and establish time window settings;

[0032] Among them, δ u represents the time scale adjustment coefficient, M u Represents a subscript adjustment value of the power grid data monitoring record, T win represents the time window;

[0033] S212: Use the time window setting to dynamically select a time window according to the volatility of the power grid data monitoring record, using the formula:

[0034]

[0035] Calculate the data fluctuation within the window and obtain the selected time window;

[0036] Where W represents the fluctuation difference of the data in the window;

[0037] S213: Segment the data of the cable and the distribution line according to the selected time window, using the formula:

[0038]

[0039] Calculate the segmented data segments and generate data segmentation;

[0040] in, Represents the time window W u A single data point in the segment, S represents the segmented data segment, and W uRepresents the fluctuation difference of the data in the window, and n represents the number of data points;

[0041] S214: Extract key data features from the data segmentation using the formula:

[0042]

[0043] Calculate the feature set to obtain multi-scale data fragments;

[0044] in, represents a single data point in a data segment, represents the average value of data segment S, F ms Represents the extracted key data features.

[0045] As a further solution of the present invention, the step of acquiring the abnormal waveform is specifically as follows:

[0046] S311: Perform waveform comparison on the multi-scale data segments using the formula:

[0047]

[0048] Calculate the Euclidean distance between each data segment and the reference waveform to generate a difference measure;

[0049] in, represents the jth data point in the multi-scale data segment, represents the jth data point in the reference waveform, and D represents the difference measure;

[0050] S312: Based on the difference metric, using a shape matching algorithm, using the formula:

[0051]

[0052] Calculate the shape basis fit of each data segment and generate the fit;

[0053] Among them, Fit represents the degree of fit;

[0054] S313: According to the degree of fit, the formula is used:

[0055]

[0056] Generate abnormal waveforms;

[0057] Among them, α v , β v Represents the slope and position of the adjustment logic curve, and A represents the abnormal waveform.

[0058] As a further solution of the present invention, the step of obtaining the abnormal pattern recognition result is specifically:

[0059] S321: According to the signal received from the fault indicator and the abnormal waveform, the formula is used:

[0060]

[0061] Calculate and generate fault signature scores;

[0062] in, represents the weight of the fault signal, Represents the square root of the absolute value of the abnormal waveform, n t Represents the number of fault signals, F feat represents the fault feature score;

[0063] S322: Using the fault feature scores, cluster analysis is performed to identify abnormal patterns using the formula:

[0064]

[0065] generating abnormal pattern recognition results;

[0066] Among them, F feat represents the fault feature score, θ t represents the clustering threshold, M result Represents the abnormal pattern recognition result.

[0067] As a further solution of the present invention, the steps of obtaining the fault classification and response decision are specifically as follows:

[0068] S411: Calculate the probability distribution of differentiated fault causes using the abnormal pattern recognition result, using the formula:

[0069]

[0070] Generate failure cause probabilities;

[0071] Among them, a h represents the model coefficient, P faulth represents the probability of the cause of the failure;

[0072] S412: Calculate the matching degree of multiple fault modes according to the fault cause probability, using the formula:

[0073]

[0074] Generate fault matching degree D h ;

[0075] Among them, b h represents the adjustment parameter, c h represents the adjustment parameter, D h represents the fault matching degree;

[0076] S413: Formulate a response strategy based on the fault matching degree, using the formula:

[0077]

[0078] Convert the matching degree into response level to generate fault classification and response decision R h ;

[0079] Among them, d h represents the decision weight parameter, R h Represents fault classification and response decision.

[0080] Compared with the prior art, the advantages and positive effects of the present invention are:

[0081] In the present invention, by real-time monitoring of the power flow and load changes of the distribution lines and transformers, data updates are made more timely, ensuring the effectiveness of fault handling. A multi-scale analysis framework is introduced, and different time windows are selected to conduct detailed analysis of the data, which enhances the flexibility and comprehensiveness of data processing. Waveform comparison is performed through shape basis fitting analysis, which improves the accuracy of abnormal pattern recognition. Adaptive classification processing provides targeted response decisions based on abnormal patterns, significantly improving the efficiency of power grid fault response. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 It is a schematic diagram of the main steps of the present invention;

[0083] Figure 2 A flowchart of the steps for obtaining the power grid data monitoring record of the present invention;

[0084] Figure 3 A flowchart of the steps for obtaining multi-scale data segments of the present invention;

[0085] Figure 4 is a flow chart of the steps of obtaining abnormal waveforms of the present invention;

[0086] Figure 5 A flowchart of the steps for obtaining abnormal pattern recognition results of the present invention;

[0087] Figure 6 This is a flow chart of the steps for obtaining fault classification and response decision of the present invention. DETAILED DESCRIPTION

[0088] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0089] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0090] Embodiment 1

[0091] See also Figure 1 The present invention provides a technical solution: a method for intelligent fault response of a power distribution network, comprising the following steps:

[0092] S1: Collects real-time operation data from distribution lines and transformers, records changes in power flow and power load, integrates data through timestamps, and generates power grid data monitoring records;

[0093] S2: Set up a multi-scale analysis framework, dynamically select time windows based on power grid data monitoring records, segment the data of cables and distribution lines, extract key data features within multiple time windows, and generate multi-scale data segments;

[0094] S3: perform waveform comparison, perform shape basis fit analysis on each segment in the multi-scale data segment, identify and cluster abnormal waveforms, extract fault features from the signal received by the fault indicator, and generate abnormal pattern recognition results;

[0095] S4: Deploy adaptive classification processing to analyze and distinguish the causes of faults based on the abnormal pattern recognition results, classify equipment failures, operational errors or external influences, match the current fault mode to respond, and generate fault classification and response decisions.

[0096] The power grid data monitoring records include timestamps, power flow, and power load. The multi-scale data segments include time windows, key data features, and data segmentation. The abnormal pattern recognition results are specifically abnormal waveforms, fault characteristics, and waveform comparison results. Fault classification and response decisions include fault cause analysis, fault pattern matching, and response strategies.

[0097] See also Figure 2 ,The specific steps for obtaining power grid data monitoring records are:

[0098] S111: Use power sensors to measure the real-time voltage and current of distribution lines and transformers, and obtain the power factor angle from the system using the formula:

[0099]

[0100] Generate electricity flow;

[0101] Among them, V represents voltage, I represents current, φ represents the square root of the power factor, and F represents the power flow;

[0102] S112: Add a timestamp to the power flow and record the result as a corresponding time-power flow pair (t, F t ), using the formula:

[0103]

[0104] Get the marked power flow;

[0105] Among them, t r Represents the timestamp, F t represents the power flow after the mark;

[0106] S113: For the marked power flow, a time series analysis method is used to perform data integration and identify the change trend of power load, using the formula:

[0107]

[0108] Calculate and generate weighted average power flow;

[0109] Among them, F ti represents the timestamped power flow, i r represents the index in the time series, F avg represents the weighted average power flow;

[0110] S114: According to the weighted average power flow, the recorded data format is adjusted using the formula:

[0111]

[0112] Convert data units to kilowatt-hours and generate power grid data monitoring records;

[0113] Wherein, M represents the power grid data monitoring record.

[0114] formula:

[0115]

[0116] Operation process:

[0117] Voltage (V): Set the voltage measurement value to 230V.

[0118] Current (I): Set the current measurement value to 5A.

[0119] Power factor angle (φ): The general power factor angle is around the cosine value of 0.9, which corresponds to an angle of about 25.8°. However, the square root of the angle is used, so the cosine value of 0.9 is first converted to an angle and then the square root is taken.

[0120]

[0121] cos(5.08)≈0.996

[0122] Calculate F:

[0123] F=230×5×0.996=1148.4watts

[0124] Example results:

[0125] The power flow F is 1148.4 watts, which represents the instantaneous power of the grid at the moment of measurement.

[0126] formula:

[0127]

[0128] Operation process:

[0129] Timestamp (t): Sets the measurement to be taken at the 3600th second of the day.

[0130] Known power flow (F): From the above results, F = 1148.4 watts.

[0131] Calculate F t :

[0132]

[0133] Example results:

[0134] The marked power flow F t It is 19.13 watts, which represents the power value after time adjustment.

[0135] formula:

[0136]

[0137] Operation process:

[0138] There are three data points in the setup, Calculate F avg :

[0139]

[0140] Example results:

[0141] Weighted average power flow F avg is 14.23 watts, which represents the average power after adjusting for time weighting. Formula:

[0142]

[0143] Operation process:

[0144] Known weighted average power flow (F avg ): From the above results, F avg = 14.23 watts. Calculate M:

[0145]

[0146] Example results:

[0147] The grid data monitoring record M is 3677 kWh, which represents the adjusted total energy consumption.

[0148] See also Figure 3 , the steps for obtaining multi-scale data fragments are as follows:

[0149] S211: According to the power grid data monitoring records, a multi-scale analysis framework is set up, using the formula:

[0150]

[0151] Calculate and establish time window settings;

[0152] Among them, δ u represents the time scale adjustment coefficient, M u Represents a subscript adjustment value of the power grid data monitoring record, T win represents the time window;

[0153] S212: Use the time window setting to dynamically select the time window according to the volatility of the power grid data monitoring record, using the formula:

[0154]

[0155] Calculate the data fluctuation within the window and obtain the selected time window;

[0156] Where W represents the fluctuation difference of the data in the window;

[0157] S213: Split the data of cables and distribution lines according to the selected time window, using the formula:

[0158]

[0159] Calculate the segmented data segments and generate data segmentation;

[0160] in, Represents the time window W u A single data point in the segment, S represents the segmented data segment, and W u Represents the fluctuation difference of the data in the window, and n represents the number of data points;

[0161] S214: Extract key data features from data segmentation using the formula:

[0162]

[0163] Calculate the feature set to obtain multi-scale data fragments;

[0164] in, represents a single data point in a data segment, represents the average value of data segment S, F ms Represents the extracted key data features.

[0165] formula:

[0166]

[0167] in:

[0168] M: The average current value recorded by the power grid data monitoring is set to 500A.

[0169] δ u : Time scale adjustment coefficient, set to 0.05.

[0170] M u : The coefficient of variation of the power grid data monitoring record is set to 100A.

[0171] Calculation process:

[0172] First, calculate the logarithmic part of the adjustment factor:

[0173] log(δ u +1)=log(0.05+1)=log(1.05)≈0.021

[0174] Compute the square root in the denominator:

[0175]

[0176] Calculate the entire time window T win :

[0177]

[0178] This time window represents the dynamically selected window size for power grid data monitoring.

[0179] formula:

[0180]

[0181] in:

[0182] max(M u )、min(M u ): Set the maximum value of the power grid data to 600A and the minimum value to 400A.

[0183] T win : 1.05 seconds as calculated above.

[0184] Calculation process:

[0185] Calculate the difference between the maximum and minimum values:

[0186] max(M u )-min(M u )=600-400=200

[0187] Calculate the data fluctuation W within the time window:

[0188]

[0189] This value represents the rate of change of current within a given time window. Formula:

[0190]

[0191] in:

[0192] Data points, set to average 500A.

[0193] W u : The current change rate calculated above, 190.48A / s.

[0194] n: number of data points, set to 10.

[0195] Calculation process:

[0196] Calculate the denominator:

[0197] W u +1=190.48+1=191.48

[0198] Compute the square root in the numerator:

[0199]

[0200] Calculate S:

[0201]

[0202] This value represents the weighted sum of the segmented data.

[0203] formula:

[0204]

[0205] in:

[0206] A single data point in a data segment, set to 360.2A.

[0207] The average value of the data segment is also set to 360.2A.

[0208] n u : The number of data points is 10.

[0209] Calculation process:

[0210] Calculate the absolute difference:

[0211]

[0212] Calculate the feature set F ms :

[0213]

[0214] Since the data points and the mean are set to be the same, the variance is zero. In real-world applications, this would indicate stability or uniformity in the data.

[0215] See also Figure 4 , the specific steps for obtaining abnormal waveforms are:

[0216] S311: Perform waveform comparison on the multi-scale data segments using the formula:

[0217]

[0218] Calculate the Euclidean distance between each data segment and the reference waveform to generate a difference measure;

[0219] in, represents the jth data point in the multi-scale data segment, represents the jth data point in the reference waveform, and D represents the difference measure;

[0220] S312: Based on the difference measurement, using the shape matching algorithm, the formula is:

[0221]

[0222] Calculate the shape basis fit of each data segment and generate the fit;

[0223] Among them, Fit represents the degree of fit;

[0224] S313: According to the degree of fit, the formula is used:

[0225]

[0226] Generate abnormal waveforms;

[0227] Among them, α v , β v Represents the slope and position of the adjustment logic curve, and A represents the abnormal waveform.

[0228] formula:

[0229]

[0230] Parameter explanation and derivation:

[0231] The jth data point in the multi-scale data segment.

[0232] The jth data point in the reference waveform.

[0233] D: Euclidean distance, used to measure F ms and F ref The difference between.

[0234] Example calculation:

[0235] A data segment and reference waveform are provided, each including 3 data points.

[0236] F ms =[1.0,2.0,3.0]

[0237] F ref =[1.1,1.9,3.1]

[0238] Calculate the difference D:

[0239]

[0240] The difference measure D represents the distance between two waveforms, and a smaller value means that the two are very close. Formula:

[0241]

[0242] Example calculation:

[0243] Using the previously calculated D = 0.173:

[0244]

[0245] The fit degree Fit indicates the similarity between the data segment and the reference waveform. A value close to 1 indicates that the waveform is very similar. Formula:

[0246]

[0247] Set the parameter to α v =10 and β v =0.95:

[0248] Example calculation:

[0249] Using Fit = 0.985:

[0250]

[0251] The probability score A represents the likelihood of a data segment being identified as an anomaly, with higher values ​​indicating a greater likelihood of an anomaly.

[0252] See also Figure 5 ,The specific steps for obtaining the abnormal pattern recognition results are:

[0253] S321: Based on the signal received from the fault indicator and the abnormal waveform, the formula is used:

[0254]

[0255] Calculate and generate fault signature scores;

[0256] in, represents the weight of the fault signal, Represents the square root of the absolute value of the abnormal waveform, n t Represents the number of fault signals, F feat represents the fault feature score;

[0257] S322: Use the fault feature scores to perform cluster analysis to identify abnormal patterns using the formula:

[0258]

[0259] generating abnormal pattern recognition results;

[0260] Among them, F feat represents the fault feature score, θ t represents the clustering threshold, M result Represents the abnormal pattern recognition result.

[0261] formula:

[0262]

[0263] Parameter details and derivation process:

[0264] Abnormal waveform, which represents the abnormal indicators of the signal waveform detected from the fault indicator. These values ​​are calculated from the raw data by signal processing technology. For example, if the detected voltage or current suddenly deviates from the normal range, the value will be high.

[0265] A weighting factor is given based on the importance or reliability of the signal. For example, some signals are more critical or accurate than others and therefore have a higher weight.

[0266] n t : The total number of fault signals, this value is based on the number of system monitoring points.

[0267] Set up the study:

[0268] Three fault signals are set, and the abnormal waveform parameter A of each signal is 16, 9, and 4 respectively (it is the square of the amplitude measured from a certain sensor).

[0269] The corresponding weights w are set to 0.5, 0.3, and 0.2 respectively.

[0270] The calculation process is as follows:

[0271]

[0272] Result analysis:

[0273] F feat =3.3 is a weighted sum, which represents the fault feature score with comprehensive reference to the importance of the differentiated signal.

[0274] Abnormal pattern cluster analysis

[0275] formula:

[0276]

[0277] Parameter details and derivation process:

[0278] F feat : The fault feature score calculated from the previous step.

[0279] θ t : The threshold for cluster analysis, which is determined based on historical data analysis and is set to 2.

[0280] Calculation process of the example:

[0281] Using F feat =3.3 and θ t =2:

[0282]

[0283] Result analysis:

[0284] M result ≈1.435 represents the value of the abnormal pattern recognition result calculated according to the set threshold θ, which is used to evaluate whether there is a significant abnormal pattern.

[0285] See also Figure 6 ,The specific steps for obtaining fault classification and response decision are:

[0286] S411: Calculate the probability distribution of differentiated fault causes using the abnormal pattern recognition results, using the formula:

[0287]

[0288] Generate failure cause probabilities;

[0289] Among them, a h represents the model coefficient, P faulth represents the probability of the cause of the failure;

[0290] S412: Calculate the matching degree of multiple fault modes according to the probability of fault causes, using the formula:

[0291]

[0292] Generate fault matching degree D h ;

[0293] Among them, b h represents the adjustment parameter, c h represents the adjustment parameter, D h represents the fault matching degree;

[0294] S413: According to the fault matching degree, formulate a response strategy using the formula:

[0295]

[0296] Convert the matching degree into response level to generate fault classification and response decision R h ;

[0297] Among them, d h represents the decision weight parameter, R h Represents fault classification and response decision.

[0298] formula:

[0299]

[0300] Parameter explanation and acquisition method:

[0301] a h : Model coefficients, determined based on historical data through the least squares method or other statistical methods.

[0302] Derivation and examples:

[0303] There are three types of faults, and the abnormal pattern recognition results M resulth are 1.2, 0.8 and 0.5 respectively. Model coefficient a h Uniformly assumed to be 0.5.

[0304] Calculate each type of exp(a h ×M resulthj ):

[0305] exp(0.5×1.2)=exp(0.6)≈1.822

[0306] exp(0.5×0.8)=exp(0.4)≈1.491

[0307] exp(0.5×0.5)=exp(0.25)≈1.284

[0308] Calculate the sum:

[0309] Calculate P faulth :

[0310] right

[0311] This value of 0.396 represents the probability of type failure occurring.

[0312] formula:

[0313]

[0314] Parameter explanation and acquisition method:

[0315] b h 、c h : Adjustment parameters are set based on historical data analysis and expert input.

[0316] Derivation and examples:

[0317] Setting b h =2 and c h =0.4.

[0318] Calculate D h :

[0319]

[0320] The value 0.499 represents the degree of current fault matching.

[0321] formula:

[0322]

[0323] Parameter explanation and acquisition method:

[0324] d h : Decision weighting parameter, determined based on optimization model or experiment.

[0325] Derivation and examples:

[0326] Setting h =0.3.

[0327] Calculate R h :

[0328] R h =0.499×e 0.3×0.499 ≈0.499×e 0.1497 ≈0.499×1.1615≈0.579

[0329] The value of 0.579 indicates the effectiveness of the fault response decision.

[0330] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for intelligent fault response in a power distribution network, characterized in that: The following steps are involved: Collect real-time operation data from distribution lines and transformers, record changes in power flow and power load, integrate data through timestamps, and generate power grid data monitoring records; Setting a multi-scale analysis framework, dynamically selecting a time window based on the power grid data monitoring record, segmenting the data of cables and distribution lines, extracting key data features within multiple time windows, and generating multi-scale data segments; Performing waveform comparison, performing shape basis fit analysis on each segment in the multi-scale data segment, identifying and clustering abnormal waveforms, extracting fault features from signals received by the fault indicator, and generating abnormal pattern recognition results; The steps for obtaining the abnormal pattern recognition result are specifically as follows: According to the signal received from the fault indicator and the abnormal waveform, the formula is used: Calculate and generate fault signature scores; in, represents the weight of the fault signal, Represents the square root of the absolute value of the abnormal waveform, n t Represents the number of fault signals, F feat represents the fault feature score; Using the fault feature scores, cluster analysis is performed to identify abnormal patterns using the formula: generating abnormal pattern recognition results; Among them, F feat represents the fault feature score, θ t represents the clustering threshold, M result represents the abnormal pattern recognition result; Deploy adaptive classification processing, analyze and distinguish the causes of faults according to the abnormal pattern recognition results, classify equipment failures, operating errors or external influences, match the current fault mode to respond, and generate fault classification and response decisions; The steps for obtaining the fault classification and response decision are specifically as follows: Using the abnormal pattern recognition results, the probability distribution of differentiated fault causes is calculated using the formula: Generate failure cause probabilities; Among them, a h represents the model coefficient, P faulth represents the probability of the cause of the failure; According to the failure cause probability, the matching degree of multiple failure modes is calculated using the formula: Generate fault matching degree D h ; Among them, b h represents the adjustment parameter, c h represents the adjustment parameter, D h represents the fault matching degree; According to the fault matching degree, a response strategy is formulated using the formula: Convert the matching degree into response level to generate fault classification and response decision R h ; Among them, d h represents the decision weight parameter, R h Represents fault classification and response decision.

2. The power distribution network intelligent fault response method according to claim 1, characterized in that: The steps for obtaining the power grid data monitoring record are specifically as follows: The real-time voltage and current of the distribution lines and transformers are measured by power sensors, and the power factor angle is obtained from the system using the formula: Generate electricity flow; Among them, V represents voltage, I represents current, φ represents the square root of the power factor, and F represents the power flow; Add a timestamp to the power flow and record the result as a corresponding time-power flow pair (t, F t ), using the formula: Get the marked power flow; Among them, t r Represents the timestamp, F t represents the power flow after the mark; For the marked power flow, a time series analysis method is used to perform data integration and identify the change trend of power load, using the formula: Calculate and generate weighted average power flow; in, represents the timestamped power flow, i r represents the index in the time series, F avg represents the weighted average power flow; According to the weighted average power flow, the recorded data format is adjusted, and the formula is adopted: Convert data units to kilowatt-hours and generate power grid data monitoring records; Wherein, M represents the power grid data monitoring record.

3. The power distribution network intelligent fault response method according to claim 2, characterized in that: The steps of obtaining the multi-scale data segments are specifically as follows: According to the power grid data monitoring records, a multi-scale analysis framework is set up, using the formula: Calculate and establish time window settings; Among them, δ u represents the time scale adjustment coefficient, M u Represents a subscript adjustment value of the power grid data monitoring record, T win represents the time window; The time window setting is used to dynamically select a time window according to the volatility of the power grid data monitoring record, using the formula: Calculate the data fluctuation within the window and obtain the selected time window; Where W represents the fluctuation difference of the data in the window; The data of cables and distribution lines are split according to the selected time window, using the formula: Calculate the segmented data segments and generate data segmentation; in, Represents the time window W u A single data point in the segment, S represents the segmented data segment, and W u Represents the fluctuation difference of the data in the window, and n represents the number of data points; Extract key data features from the data segmentation using the formula: Calculate the feature set to obtain multi-scale data fragments; in, represents a single data point in a data segment, represents the average value of data segment S, F ms Represents the extracted key data features.

4. The power distribution network intelligent fault response method according to claim 3, characterized in that: The steps of obtaining the abnormal waveform are specifically as follows: The waveform comparison is performed on the multi-scale data segments using the formula: Calculate the Euclidean distance between each data segment and the reference waveform to generate a difference measure; in, represents the jth data point in the multi-scale data segment, represents the jth data point in the reference waveform, and D represents the difference measure; Based on the difference metric, a shape matching algorithm is used, using the formula: Calculate the shape basis fit of each data segment and generate the fit; Among them, Fit represents the degree of fit; According to the fit, the formula is used: Generate abnormal waveforms; Among them, α v , β v Represents the slope and position of the adjustment logic curve, and A represents the abnormal waveform.

Images