Power distribution network abnormal state sensing method and system based on data processing

Abstract

The invention provides a power distribution network abnormal state sensing method and system based on data processing, and relates to the technical field of data processing, and the method comprises the steps: obtaining a target function, and setting a differential evolution parameter; the differential evolution parameters comprise population size, iteration times, variation factors and crossover probability; the configuration of differential evolution parameters is iteratively optimized, various operation data are preprocessed, abnormal value and noise removal and normalization processing are included, and processed data are obtained; and extracting characteristic indexes reflecting the running state of the power distribution network from the processed data, wherein the characteristic indexes comprise statistical characteristics, frequency spectrum characteristics and time frequency characteristics of voltage and current. According to the invention, through real-time data acquisition, feature extraction, state evaluation and abnormity identification and positioning, comprehensive monitoring and timely early warning of the operation state of the power distribution network are realized, and the safety and stability of the power distribution network are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method and system for perceiving abnormal states of a distribution network based on data processing. Background Art

[0002] Under traditional technologies, the detection of abnormal states of a distribution network mainly relies on manual inspections and regular tests. However, when facing modern distribution network systems with large scale and wide distribution, this method gradually reveals its limitations:

[0003] Although manual inspections can provide certain on-site intuitive information, they are time-consuming, labor-intensive, and difficult to achieve real-time and comprehensive monitoring. Especially in the context of the continuous expansion of the distribution network scale, the contradiction between the inspection cycle and the detection coverage is becoming increasingly prominent, resulting in obvious lags in the discovery and handling of abnormal states, affecting the rapid response and handling efficiency of faults.

[0004] Highly dependent on the personal experience and intuitive judgment of operation and maintenance personnel. This subjectivity not only makes it difficult to guarantee the accuracy and reliability of detection results, but is also easily interfered by various factors such as personal experience, skill level, and external environmental conditions. In the complex and changeable operation environment of the distribution network, it is often difficult to accurately capture all potential abnormal states only by manual judgment. Only focus on limited electrical quantity indicators such as voltage and current, while ignoring the important role of non-electrical quantity information such as equipment temperature, environmental humidity, and operation historical data in abnormal state perception. This one-sided data utilization method not only limits the comprehensiveness of abnormal state perception, but also reduces the accuracy of detection.

[0005] Lack of intelligent data processing and analysis means, and it is difficult to deeply mine and pattern recognize the massive data collected. At the same time, with the continuous changes in the structure and operation mode of the distribution network, the detection means and judgment criteria are difficult to adapt to this dynamic change, resulting in the inability to effectively identify and handle abnormal states in some special cases, reducing the flexibility and adaptability of detection. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for perceiving abnormal states of a distribution network based on data processing, which realizes the rapid and accurate perception of abnormal states of the distribution network.

[0007] To solve the above technical problem, the technical solution of the present invention is as follows:

[0008] In the first aspect, a method for perceiving abnormal states of a distribution network based on data processing, the method includes:

[0009] Real-time obtain various operation data in the distribution network, including voltage, current, power factor, frequency, and equipment status information;

[0010] Obtain the objective function and set the differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor, and crossover probability; iteratively optimize the configuration of the differential evolution parameters, preprocess various operation data, including removing outliers and noise, and performing normalization processing to obtain processed data;

[0011] Extract characteristic indexes reflecting the operation state of the distribution network from the processed data, including statistical characteristics, spectral characteristics, and time-frequency characteristics of voltage and current;

[0012] Construct an evaluation model based on the characteristic indexes to evaluate the current state of the distribution network and obtain the real-time evaluation result of the distribution network state;

[0013] According to the real-time evaluation result of the distribution network state, obtain the result of anomaly identification and location through pattern recognition;

[0014] Generate warning information according to the result of anomaly identification and location, including anomaly type and anomaly location.

[0015] Further, obtain the objective function and set the differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor, and crossover probability; iteratively optimize the configuration of the differential evolution parameters, preprocess various operation data, including removing outliers and noise, and performing normalization processing to obtain processed data, including:

[0016] Define an objective function and set the population size, upper limit of the number of iterations, mutation factor, and crossover probability;

[0017] Randomly generate a set of parameter candidate sets as the initial population according to the preset population size;

[0018] Evaluate the processing effect of each set of parameter candidate sets in the initial population to obtain the evaluation result, and determine the parameter set according to the evaluation result, and use the determined parameter set as the parent generation;

[0019] Perform mutation operation on the determined parameter set to generate new parameter combinations, and perform crossover operation on the generated new parameter combinations. Select some parameters for exchange according to the preset crossover probability. Through mutation and crossover operations, generate a set of new parameter candidate solutions;

[0020] Substitute the new parameter candidate solutions into the objective function for re-evaluation to obtain the evaluation result, and update the population according to the evaluation result;

[0021] Repeat the steps of parameter mutation, crossover, objective function evaluation, and population update until the preset number of iterations is reached to obtain the final parameter configuration;

[0022] Normalize various operation data according to the final parameter configuration to obtain the processed data.

[0023] Further, extract characteristic indexes reflecting the operation state of the distribution network from the processed data, including statistical characteristics, spectral characteristics and time-frequency characteristics of voltage and current, including:

[0024] Obtain the time series data of electrical parameters from the processed data, including voltage, current, power and frequency;

[0025] According to the time series data, calculate the time-domain characteristics of voltage and current signals, including mean value, maximum value, minimum value and peak value, and form the time-domain analysis result;

[0026] Perform Fourier transform and wavelet transform on voltage and current signals respectively. Through Fourier transform, convert the signal from the time domain to the frequency domain, and extract spectral information such as main frequency, harmonic components and spectral energy. Through wavelet transform, obtain the frequency components of the signal at different time points to get the frequency-domain analysis result;

[0027] According to the time-domain and frequency-domain analysis results, extract statistical characteristics, including mean value, skewness and kurtosis; extract characteristics from spectral analysis, including main frequency, harmonic content and spectral entropy; extract characteristics from time-frequency analysis, including wavelet coefficients and time-frequency energy distribution.

[0028] Further, the characteristic indexes include statistical characteristics, spectral characteristics and time-frequency characteristics of voltage and current; according to the characteristic indexes, construct an evaluation model to evaluate the current state of the distribution network and obtain the real-time evaluation result of the distribution network state, including:

[0029] Divide the characteristic indexes into a training set and a test set, and select the corresponding evaluation model;

[0030] Configure the parameters of the evaluation model. The parameters include the depth of the decision tree, and use the training set to train the evaluation model;

[0031] During the training process, the evaluation model learns the mapping relationship between the characteristic indexes and the distribution network state to obtain the trained evaluation model;

[0032] Use the test set to verify the trained evaluation model. By calculating evaluation indexes, including accuracy rate and recall rate, and according to the evaluation indexes, obtain the verification result of the evaluation model;

[0033] According to the verification result of the evaluation model, adjust the depth parameter of the evaluation model, retrain the evaluation model with the adjusted parameters, and verify again. Continuously repeat the process of adjustment, training and verification until the performance of the evaluation model reaches the preset standard to obtain the trained and verified evaluation model;

[0034] Integrate the trained and validated evaluation model into the distribution network monitoring system, and obtain feature data in real time from the distribution network, including statistical features, spectral features, and time-frequency features of voltage and current;

[0035] Input the feature data obtained in real time into the evaluation model for real-time evaluation to obtain the real-time evaluation results of the distribution network status, including status classification, abnormal type, and abnormal degree information.

[0036] Furthermore, according to the real-time evaluation results of the distribution network status, obtain the results of anomaly recognition and location through pattern recognition, including:

[0037] Extract the feature indicators reflecting the current status of the distribution network from the real-time evaluation results of the distribution network status, including statistical features, spectral features, and time-frequency features of voltage and current;

[0038] Match the feature indicators with the preset abnormal pattern library to obtain the matching results, including the matching degree with each abnormal type;

[0039] Identify and determine the abnormal type in the current distribution network according to the matching results and the preset threshold;

[0040] According to the abnormal type in the current distribution network, determine the source data of the anomaly and the corresponding monitoring points, analyze the propagation and distribution of the source data of the anomaly in the distribution network, determine the location where the anomaly occurs, and obtain the results of anomaly recognition and location.

[0041] Furthermore, match the feature indicators with the preset abnormal pattern library to obtain the matching results, including the matching degree with each abnormal type, including:

[0042] Traverse the feature patterns of each abnormal type in the abnormal pattern library;

[0043] For each abnormal type, calculate the similarity between the extracted feature indicators and the corresponding feature patterns, and sort the similarities to obtain the matching results.

[0044] Furthermore, traverse the feature patterns of each abnormal type in the abnormal pattern library, including:

[0045] Traverse each abnormal type in the abnormal pattern library, and perform the following steps for the currently traversed abnormal type:

[0046] Obtain the feature pattern of this abnormal type, and the feature pattern is a regular expression;

[0047] In the log or data stream to be detected, use the compiled regular expression for searching and matching. If a matching item is found, record the matching position, content, and abnormal type;

[0048] For each match, generate corresponding exception reports or warnings based on the exception type and the matched content;

[0049] After completing the matching of the current exception type, continue to traverse the next exception type until all exception types have been traversed.

[0050] Furthermore, the warning information includes the exception type and the exception location.

[0051] Furthermore, generate warning information based on the results of exception identification and location, including:

[0052] Determine the basic structure of the warning information based on the result data of exception identification and location;

[0053] Create a warning information object or data structure to store the basic structure;

[0054] Convert the exception type identifier into a text description, and fill in the device name, line number, and specific node of the exception location into the warning information according to the result of exception location;

[0055] Configure the sending parameters, perform the sending operation, and pass the warning information to the specified recipient.

[0056] Furthermore, the basic structure includes a warning title, an exception type description, and an exception location description.

[0057] Furthermore, convert the exception type identifier into a text description, and fill in the device name, line number, and specific node of the exception location into the warning information, including:

[0058] Create an exception type mapping table, which is used to map the internal identifier of the exception type to the corresponding text description;

[0059] When the exception identification result is obtained, extract the exception type identifier therein, use the exception type identifier as an index to look up the corresponding text description in the exception type mapping table. If a matching text description is found, save it as part of the warning information; if no matching item is found, record a default or general exception type description;

[0060] Obtain the detailed information of the exception location from the exception location module, including the device name, line number, and specific node;

[0061] Create a warning information template, which contains placeholders for filling in the exception type and exception location information, and fill in the query result of the exception type text description into the corresponding placeholder position in the warning information template;

[0062] According to the anomaly location result, fill in the corresponding positions in the warning information template with information such as device name, line number, and specific node.

[0063] Furthermore, configure the sending parameters, perform the sending operation, and transmit the warning information to the specified recipients, including:

[0064] Determine the list of potential recipients of the warning information and their contact information;

[0065] Set the initial temperature T0, termination temperature Tf, cooling coefficient α, and the number of iterations L at each temperature of the simulated annealing algorithm;

[0066] Define an evaluation function to evaluate the pros and cons of the sending strategy;

[0067] Randomly generate an initial sending strategy, including the selected recipients, sending channels, and sending times; calculate the objective function value of the initial solution; set the current temperature T to the initial temperature T0, enter a loop until the current temperature T drops below the termination temperature Tf; for the current temperature T, perform L iterations; based on the current strategy, randomly generate a new sending strategy; calculate the objective function value of the new strategy and determine whether to accept the new sending strategy, and update the current strategy and its objective function value; after completing all iterations at the current temperature, reduce the temperature according to the cooling coefficient α; when the temperature drops below the termination temperature Tf, exit the simulated annealing process, and output the current final strategy and its objective function value as the optimized sending strategy;

[0068] According to the optimized sending strategy, configure the sending parameters, perform the sending operation, and send the warning information to the specified recipients according to the optimized sending strategy.

[0069] In a second aspect, a distribution network abnormal state perception system based on data processing includes:

[0070] An acquisition module, which is used to acquire various operation data in the distribution network in real time, including voltage, current, power factor, frequency, and device status information; obtain the objective function, and set the differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor, and crossover probability; iteratively optimize the configuration of the differential evolution parameters, preprocess various operation data, including removing outliers and noise, and performing normalization processing, to obtain the processed data; extract characteristic indicators reflecting the operation state of the distribution network from the processed data, including statistical characteristics, spectral characteristics, and time-frequency characteristics of voltage and current;

[0071] A processing module, which is used to construct an evaluation model according to the characteristic indicators to evaluate the current state of the distribution network and obtain the real-time evaluation result of the distribution network state;

[0072] An evaluation module, configured to obtain the results of abnormal recognition and location through pattern recognition according to the real-time evaluation results of the distribution network status;

[0073] A warning module, configured to generate warning information according to the results of abnormal recognition and location, including the abnormal type and the abnormal location.

[0074] In a third aspect, a computing device includes:

[0075] One or more processors;

[0076] A storage device, configured to store one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the method described above.

[0077] In a fourth aspect, a computer-readable storage medium stores a program, which when executed by a processor, implements the method described above.

[0078] The above solution of the present invention has at least the following beneficial effects:

[0079] A variety of operation data in the distribution network are obtained in real time, which not only covers basic electrical parameters such as voltage, current, power factor, and frequency, but also includes non-electrical quantity data such as equipment status information. This comprehensive data acquisition method provides a rich and accurate information foundation for abnormal state perception, effectively improving the comprehensiveness and accuracy of detection. By defining the objective function and carefully setting the differential evolution parameters (such as population size, number of iterations, mutation factor, and crossover probability), the parameter configuration can be iteratively optimized, and the collected variety of operation data can be finely preprocessed. This includes removing outliers and noise, as well as performing normalization processing, so as to obtain a cleaner and more regular data set. Such preprocessing steps improve the accuracy and efficiency of subsequent feature extraction and state evaluation. From the preprocessed data, feature indicators reflecting the operation state of the distribution network are further extracted. These indicators not only include the statistical characteristics of voltage and current, but also cover deeper information such as spectral characteristics and time-frequency characteristics. Through such multi-dimensional feature extraction, the recognition of abnormal states is made more accurate and comprehensive.

[0080] Based on the extracted feature indicators, an evaluation model is constructed to perform real-time evaluation on the current state of the distribution network. This model can comprehensively consider various factors and quickly output the real-time evaluation results of the distribution network state. According to the real-time evaluation results, through advanced pattern recognition technology, the abnormal types can be quickly identified and the locations where the abnormalities occur can be accurately located. This function greatly shortens the time for fault troubleshooting, improves the efficiency of fault handling, and reduces the impact of faults on the operation of the distribution network. Finally, according to the results of abnormal identification and location, warning information can be automatically generated, including key information such as abnormal types and abnormal locations. These warning information can be sent to the operation and maintenance personnel in a timely manner, enabling them to respond quickly and take measures to effectively prevent or reduce the losses caused by faults. Brief Description of the Drawings

[0081] Figure 1 is a schematic flow chart of a method for perceiving abnormal states of a distribution network based on data processing provided by an embodiment of the present invention.

[0082] Figure 2 is a schematic diagram of a system for perceiving abnormal states of a distribution network based on data processing provided by an embodiment of the present invention. Detailed Embodiments

[0083] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0084] As Figure 1 shown, an embodiment of the present invention proposes a method for perceiving abnormal states of a distribution network based on data processing, and the method includes the following steps:

[0085] Step 11, obtain various operation data in the distribution network in real time, including voltage, current, power factor, frequency, and equipment status information;

[0086] Step 12, obtain the objective function and set the differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor, and crossover probability; iteratively optimize the configuration of the differential evolution parameters, and preprocess the various operation data, including removing outliers and noise, and performing normalization processing, to obtain the processed data;

[0087] Step 13, extract the feature indicators reflecting the operation state of the distribution network from the processed data, including statistical features, spectral features, and time-frequency features of voltage and current;

[0088] Step 14: Construct an evaluation model based on the characteristic indicators to evaluate the current state of the distribution network and obtain the real-time evaluation result of the distribution network state;

[0089] Step 15: According to the real-time evaluation result of the distribution network state, obtain the result of anomaly recognition and location through pattern recognition;

[0090] Step 16: Generate warning information according to the result of anomaly recognition and location, including the anomaly type and the anomaly location.

[0091] In the embodiment of the present invention, by collecting various data such as voltage, current, power factor, frequency, and equipment status information in real time, the comprehensive monitoring of the operation state of the distribution network is realized. The acquisition of real-time data enables the system to quickly respond to any changes in the distribution network, providing the possibility for timely discovery and handling of potential problems. By removing outliers and noise, the accuracy and reliability of the data are improved. The normalization process makes the data with different dimensions comparable. Using the differential evolution algorithm to iteratively optimize the parameter configuration ensures the intelligence and self-adaptability of the data preprocessing process and improves the processing efficiency. Extracting the statistical features, spectral features, and time-frequency features of voltage and current from the processed data realizes the in-depth mining of the operation state of the distribution network. These characteristic indicators can accurately characterize the operation state of the distribution network.

[0092] The evaluation model constructed based on the characteristic indicators can evaluate the current state of the distribution network in real time, providing timely decision support for the operation and maintenance personnel. The model can also predict the future state based on historical data, helping to discover potential problems in advance and take measures. Through the pattern recognition technology, the anomaly types in the distribution network can be quickly identified, improving the efficiency of fault troubleshooting. At the same time, the location where the anomaly occurs can be accurately located, providing accurate guidance for fault repair. According to the result of anomaly recognition and location, warning information including the anomaly type and the anomaly location is generated in time, providing timely warnings for the operation and maintenance personnel. The generation of warning information enables the operation and maintenance personnel to actively intervene and take measures in time to prevent the expansion of the fault, ensuring the safe and stable operation of the distribution network.

[0093] In a preferred embodiment of the present invention, step 11 of obtaining various operation data in the distribution network in real time, including voltage, current, power factor, frequency, and equipment status information, may include:

[0094] Install high-precision voltage and current sensors, such as voltage transformers (VT) and current transformers (CT), at key nodes and branches of the distribution network to measure voltage and current values in real time. Using smart meters or power analyzers, these devices can simultaneously measure the power factor (PF) and grid frequency (f), reflecting the utilization efficiency of electrical energy and the stability of the power grid. By deploying temperature sensors, humidity sensors, vibration sensors, etc., the operating status of equipment such as transformers, switch cabinets, and lines can be monitored in real time, including parameters such as temperature, humidity, and vibration. Configure a data collector (DCU) to be responsible for collecting the raw data from each sensor and performing preliminary format conversion and data packaging. Establish a stable and reliable communication network, such as optical fiber communication, wireless communication (such as LoRa, 4G / 5G), to ensure that data can be transmitted to the data center in real time and accurately. Deploy a data receiving server in the data center to be responsible for receiving the data packets from the collector and performing parsing and verification to ensure the integrity and accuracy of the data. Decode the received data according to the communication protocol to extract specific information such as voltage, current, power factor, frequency, and equipment status.

[0095] In a preferred embodiment of the present invention, in step 12, obtain an objective function and set differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor, and crossover probability; iteratively optimize the configuration of the differential evolution parameters, and perform preprocessing on various operating data, including removing outliers and noise, and performing normalization processing, to obtain processed data, which may include:

[0096] Step 122, define an objective function, and set the population size, upper limit of the number of iterations, mutation factor, and crossover probability;

[0097] Step 123, according to the preset population size, randomly generate a set of parameter candidate sets as the initial population;

[0098] Step 124, evaluate the processing effect of each set of parameter candidate sets in the initial population to obtain an evaluation result, and according to the evaluation result, determine a parameter set, and use the determined parameter set as the parent generation;

[0099] Step 125, perform a mutation operation on the determined parameter set to generate new parameter combinations, and perform a crossover operation on the generated new parameter combinations. According to the preset crossover probability, select some parameters for exchange. Through mutation and crossover operations, generate a set of new parameter candidate solutions;

[0100] Step 126, substitute the new parameter candidate solutions into the objective function for re-evaluation to obtain an evaluation result, and update the population according to the evaluation result;

[0101] Step 127, repeatedly execute the steps of parameter mutation, crossover, objective function evaluation, and population update until the preset number of iterations is reached to obtain the final parameter configuration;

[0102] Step 128, according to the final parameter configuration, perform normalization processing on various running data to obtain the processed data.

[0103] In the embodiment of the present invention, the objective function should be able to accurately reflect the quality of the preprocessing effect. For example, it can be defined as minimizing the error between the processed data and the original data, while considering the removal effect of outliers and noise, and the degree of data normalization. Set the population size: for example, set it to 50, indicating that there are initially 50 sets of parameter candidate sets. Upper limit of the number of iterations: for example, set it to 100, indicating that the algorithm runs at most 100 generations. The mutation factor is set to 0.5, controlling the amplitude of the mutation operation. Crossover probability: for example, set it to 0.7, determining the proportion of parameters from different parents in the new parameter combination in the crossover operation. According to the preset population size, randomly generate 50 sets of parameter candidate sets, each set containing the parameters required for preprocessing, such as denoising thresholds, normalization ranges, etc. Evaluate the preprocessing effect of each set of parameter candidate sets by calculating the objective function value. For example, according to the objective function value, all parameter candidate sets can be sorted or ranked, and the parameter sets ranked higher perform better in the preprocessing effect because their objective function values are better (for example, the error is smaller). According to the evaluation results, select the parameter set with the best performance as the parent generation.

[0104] Perform a mutation operation on the parent parameter set to generate new parameter combinations. According to the crossover probability, select some parameters for exchange to generate new parameter candidate solutions. Substitute the new parameter candidate solutions into the objective function and calculate their effects. According to the evaluation results, update the population and retain the parameter set with the best performance. Repeatedly execute the steps of parameter mutation, crossover, objective function evaluation, and population update until the preset number of iterations (such as 100 generations) is reached. Use the final parameter configuration obtained by iterative evolution to preprocess various running data, including removing outliers and noise, and performing normalization processing.

[0105] Preprocess the voltage and current data in the distribution network. Remove outliers and noise from the data. Normalize the data to the range of [0, 1] or [-1, 1]. Assume there is a distribution network dataset containing voltage data. Optimize the preprocessing parameters through differential evolution to remove noise and perform normalization. Define it as maximizing the SNR of the processed data. Population size: 50; Upper limit of iteration times: 100; Mutation factor: 0.5; Crossover probability: 0.7. Randomly generate 50 groups of parameter candidate sets, each group containing parameters such as the denoising threshold (e.g., the threshold of wavelet denoising), the upper and lower limits of normalization, etc. Perform preprocessing on each group of parameter candidate sets, calculate the objective function value after processing, and select the parameter set with the highest SNR as the parent generation. Mutate the parent generation parameter set, such as changing the denoising threshold by adding random noise. According to the crossover probability, select some parameters for exchange to generate new parameter candidate solutions. Repeat the mutation, crossover, evaluation, and update steps until 100 iterations are reached. Use the final parameter configuration obtained through iterative evolution to denoise and normalize the voltage data to obtain high-quality processed data.

[0106] Traditional selection of data preprocessing parameters relies on expert experience or the trial-and-error method, while the differential evolution algorithm can automatically search the parameter space to find the final parameter configuration, thus reducing the influence of human intervention and subjective judgment. Through iterative optimization, the differential evolution algorithm can find the final parameters for removing outliers and noise, making the processed data cleaner and more accurate. At the same time, the normalization process also enables the data to be compared and analyzed on a unified scale, improving the comparability and consistency of the data. High-quality data is the basis for model training. The data preprocessed through differential evolution optimization can better reflect the true characteristics and laws of the data, thereby improving the training effect and prediction accuracy of the model. The differential evolution algorithm does not depend on the specific type and distribution of the data and can process various operating data, such as time series data, sensor data, etc., and has wide applicability.

[0107] In a preferred embodiment of the present invention, the calculation formula of the objective function is:

[0108]

[0109] where F represents the objective function value; a1, a2, b1, b2, γ represent weight coefficients; μ represents the average value of the processed signal; σ represents the standard deviation of the noise; σ d represents the standard deviation of the processed data; N represents the total number of data points in the parameter set; i represents the index variable; x i+1 represents the (i + 1)-th data point in the parameter set; x i represents the i-th data point in the parameter set; χ n represents the chi-square statistic; A represents the amplitude of a specific target frequency.

[0110] In the embodiments of the present invention, weight coefficients a1, a2, b1, b2, and γ are determined. These weight coefficients are used to balance different parts in the objective function to reflect the expectation for the data preprocessing effect. Apply preprocessing parameters (such as denoising, normalization, etc.) to a given data set, and then calculate the average value μ of the processed signal. This average value reflects the overall level of the processed data. The calculation formula for the average value of the processed signal is: where x' i represents the i-th data point in the processed parameter set. Similarly, for the noise part in the processed data, calculate the standard deviation σ. This value is used to measure the degree of noise fluctuation. Suppose there is a noise data set where n i is the i-th noise data point extracted from the processed data, and M is the total number of noise data points. The calculation formula for the standard deviation σ of the noise is: where n i represents the i-th noise data point extracted from the processed data; n j represents the j-th noise data point extracted from the processed data. Calculate the standard deviation σ of the processed data d as: According to the formula calculate the smoothness of the processed data. The smaller this value is, the smoother the data is, that is, the smaller the difference between adjacent data points. According to the characteristics of the data and the preprocessing requirements, calculate the chi-square statistic χ n . This statistic is usually used to measure the data against a certain expected distribution. If one of the goals of preprocessing is to enhance or suppress signals at a specific frequency, then the amplitude A of that frequency in the processed data needs to be calculated. Substitute the values obtained above into the objective function formula to calculate the objective function value F. This value reflects the comprehensive evaluation of the data preprocessing effect under the current preprocessing parameter configuration.

[0111] Suppose we have a simple processed data set and it is known that the standard deviation of the noise σ = 1, the chi-square statistic the amplitude A of the specific target frequency = 10, the weight coefficients a1 = 1, a2 = 0.5, b1 = 0.1, b2 = 0.2, γ = 1. Calculate the average value of the processed signal Calculate the standard deviation of the processed data Smoothness term Calculate the objective function value

[0112] By maximizing (i.e., signal-to-noise ratio), the a1 coefficient excitation algorithm increases the mean value μ of the processed signal and reduces the standard deviation σ of the noise, thereby enhancing the signal quality. The a2 coefficient penalizes the square of the standard deviation σ d of the processed data, which helps reduce the data fluctuations and then improves the data stability. The b1 coefficient is inversely proportional to the sum of the squares of the differences between adjacent data points (i.e., the smoothness term), which prompts the algorithm to generate a smoother data sequence, thereby eliminating or reducing the mutations and irregularities in the data. By penalizing the chi-square statistic with the b2 coefficient, it can ensure that the processed data fits better with the expected distribution, and then improve the goodness of fit of the data. The γ coefficient is proportional to the amplitude A of a specific target frequency, which allows the algorithm to specifically focus on and enhance the response of a specific frequency in the data, which is particularly important for application scenarios such as spectrum analysis and signal detection. The weight coefficients a1, a2, b1, b2, γ in the formula can all be adjusted according to actual needs, providing the algorithm with extremely high flexibility and customizability. By adjusting these coefficients, it can easily adapt to different application scenarios and data characteristics. The objective function comprehensively considers multiple dimensions such as signal quality, data stability, smoothness, goodness of fit, and specific frequency response, providing a comprehensive and integrated performance evaluation index for data processing and analysis.

[0113] In a preferred embodiment of the present invention, in step 13 above, extracting the characteristic indexes reflecting the operation state of the distribution network from the processed data, including the statistical characteristics, spectral characteristics, and time-frequency characteristics of voltage and current, may include:

[0114] Step 131, obtaining the time series data of electrical parameters from the processed data, including voltage, current, power, and frequency;

[0115] Step 132, calculating the time-domain characteristics of the voltage and current signals according to the time series data, including the mean value, maximum value, minimum value, and peak value, to form the time-domain analysis result;

[0116] Step 133, respectively performing Fourier transform and wavelet transform on the voltage and current signals. Through the Fourier transform, the signal is transformed from the time domain to the frequency domain, and spectral information such as the main frequency, harmonic components, and spectral energy is extracted. Through the wavelet transform, the frequency components of the signal at different time points are obtained to get the frequency-domain analysis result;

[0117] Step 134, extracting the statistical characteristics according to the time-domain and frequency-domain analysis results, including the mean value, skewness, and kurtosis; extracting the characteristics from the spectrum analysis, including the main frequency, harmonic content, and spectral entropy; extracting the characteristics from the time-frequency analysis, including wavelet coefficients and time-frequency energy distribution.

[0118] In an embodiment of the present invention, time series data of electrical parameters is obtained from the processed data, which includes key parameters such as voltage, current, power, and frequency. For the voltage signal V(t) and the current signal I(t), assume there is a time series containing N data points, where t i represents the time point; V i and I i respectively represent the voltage and current values at the time point t i . Calculate the time-domain characteristics of the voltage and current signals, such as mean value, maximum value, minimum value, and peak value. The formula for calculating the voltage mean value is: Current mean value: Voltage maximum value: V max = max(V1, V2, …, V N ); Current maximum value: I max = max(I1, I2, …, I N ); Voltage minimum value: V min = min(V1, V2, …, V N ); Current minimum value: I min = min(I1, I2, …, I N ); The voltage peak refers to the maximum absolute value in the signal, but depending on the specific application scenario, it may also refer to the positive peak or negative peak. Assume it refers to the maximum absolute value V peak = max(|V1|, |V2|, …, |V N |); Similarly for the current peak I peak = max(|I1|, |I2|, …, |I N |).

[0119] These time-domain characteristics can intuitively reflect the overall level and fluctuation of the signal, providing basic data for the state assessment of the distribution network. Perform Fourier transform on the voltage and current signals respectively to convert the signals from the time domain to the frequency domain. In the frequency domain, we extract spectral information such as the main frequency, harmonic components, and spectral energy. These information helps to identify the frequency components and potential harmonic problems in the signal. Apply wavelet transform to analyze the signal to obtain the frequency components of the signal at different time points. Through wavelet transform, characteristics such as time-frequency energy distribution and wavelet coefficients can be obtained, and these characteristics can reveal the dynamic changes of the signal in time and frequency. Based on the time-domain and frequency-domain analysis results, further extract statistical characteristics such as mean value, skewness, and kurtosis. These statistical characteristics can reflect the probability distribution and fluctuation characteristics of the signal, providing a basis for the anomaly detection and state assessment of the distribution network. Extract characteristics such as the main frequency, harmonic content, and spectral entropy from the spectral analysis. Extract characteristics such as wavelet coefficients and time-frequency energy distribution from the time-frequency analysis. These comprehensive characteristics can comprehensively reflect the complexity and dynamics of the operation state of the distribution network.

[0120] By monitoring the spectral characteristics of voltage and current, harmonic problems and frequency anomalies in the distribution network can be detected in a timely manner. For example, when unexpected high-frequency components appear in the spectrum, it may indicate equipment failures or non-linear loads in the distribution network. Using time-domain and frequency-domain characteristics, we can quantitatively evaluate the operating state of the distribution network. For example, by calculating the mean and fluctuation range of voltage and current, we can evaluate the voltage quality and current stability of the distribution network. Combining time-series data and statistical characteristics, we can establish a load forecasting model. For example, by analyzing the time-domain characteristics of historical load data and current electrical parameters, we can predict the load change trend in the future for a period of time. Using the time-frequency characteristics obtained by wavelet transform, we can accurately locate and diagnose faults in the distribution network. For example, when a fault occurs at a certain location in the distribution network, the wavelet coefficients and time-frequency energy distribution will change, providing a basis for fault location.

[0121] Through comprehensive feature extraction in the time domain, frequency domain, and time-frequency domain, the operating state of the distribution network can be evaluated comprehensively and from multiple perspectives. This helps to understand the actual situation of the power grid more accurately. Time-domain characteristics (such as mean, maximum value, minimum value, peak value) can reflect the overall level and instantaneous changes of the signal, helping to quickly detect abnormal fluctuations and potential faults. Frequency-domain characteristics (such as main frequency, harmonic components, spectral energy) can reveal the frequency components and harmonic problems in the signal, helping to locate the fault source and judge the fault type. Time-frequency characteristics (such as wavelet coefficients, time-frequency energy distribution) can consider the changes in both time and frequency simultaneously, having unique advantages in detecting non-stationary signals and transient faults. By deeply analyzing electrical parameters such as voltage and current, the load conditions, power distribution, and frequency stability of the power grid can be understood, providing data support for the optimal operation of the power grid. For example, according to the harmonic content in spectral analysis, the parameters of the filter can be adjusted to reduce the impact of harmonics on the power grid; according to the time-frequency energy distribution in time-frequency analysis, the dispatching strategy of the power grid can be optimized to improve energy utilization efficiency. Combining time-series data and statistical characteristics, a more accurate load forecasting model can be established. This helps to understand future electricity demand in advance, providing a scientific basis for the planning and dispatching of the power grid. The development of feature extraction and data analysis technologies provides strong support for the intelligentization of the power grid. By real-time monitoring and analyzing the operating data of the power grid, dynamic perception and intelligent decision-making of the power grid state can be realized.

[0122] In a preferred embodiment of the present invention, in step 14 above, the characteristic indicators include statistical characteristics, spectral characteristics, and time-frequency characteristics of voltage and current; according to the characteristic indicators, an evaluation model is constructed to evaluate the current state of the distribution network, and a real-time evaluation result of the distribution network state is obtained, including:

[0123] Divide the characteristic indicators into a training set and a test set, and select the corresponding evaluation model;

[0124] Configure the parameters of the evaluation model, where the parameters include the depth of the decision tree, and use the training set to train the evaluation model;

[0125] During the training process, the evaluation model learns the mapping relationship between the feature indicators and the distribution network state, and obtains the trained evaluation model;

[0126] Use the test set to verify the trained evaluation model. By calculating evaluation metrics, including accuracy and recall rate, and based on the evaluation metrics, obtain the verification result of the evaluation model;

[0127] According to the verification result of the evaluation model, adjust the depth parameter of the evaluation model, and use the adjusted parameters to retrain the evaluation model, and verify again. Continuously repeat the process of adjustment, training and verification until the performance of the evaluation model reaches the preset standard, and obtain the trained and verified evaluation model;

[0128] Integrate the trained and verified evaluation model into the distribution network monitoring system, and obtain feature data in real time from the distribution network, including statistical features, spectral features and time-frequency features of voltage and current;

[0129] Input the feature data obtained in real time into the evaluation model for real-time evaluation, and obtain the real-time evaluation result of the distribution network state, including state classification, abnormal type and abnormal degree information.

[0130] In the embodiment of the present invention, the feature index data set (including statistical features, spectral features and time-frequency features of voltage and current) is divided into a training set and a test set. The training set is used to train the model, and the test set is used to verify the model performance. Select the decision tree as the evaluation model and configure its parameters, such as the depth of the decision tree. The depth parameter determines the complexity of the decision tree and the risk of overfitting. Use the training set to train the evaluation model. During the training process, the evaluation model learns the mapping relationship between the feature indicators and the distribution network state. Use the test set to verify the trained evaluation model. Evaluate the performance of the model by calculating evaluation metrics (such as accuracy and recall rate). According to the verification result, adjust the depth parameter of the evaluation model, and use the adjusted parameters to retrain the model. Repeat the process of adjustment, training and verification until the performance of the model reaches the preset standard. Integrate the trained and verified evaluation model into the distribution network monitoring system to realize the function of obtaining feature data in real time from the distribution network. Input the feature data obtained in real time into the evaluation model for real-time evaluation, and obtain the real-time evaluation result of the distribution network state.

[0131] The evaluation model can classify the state of the distribution network into normal, abnormal, or a specific type of fault state based on the input feature data. For example, the model may classify the state as "low voltage", "current overload", "harmonic pollution", etc. When the distribution network is in an abnormal state, the evaluation model can further identify the type of abnormality. For example, through spectral feature analysis, the evaluation model may identify whether the harmonic problem is caused by equipment failure or voltage fluctuation caused by load change. The evaluation model can also evaluate the degree of abnormality according to the severity of the feature data. For example, through the peak value and fluctuation range in the time domain features, the model can determine whether the abnormality degree of voltage or current has reached a dangerous level, thus triggering corresponding alarms or protection measures.

[0132] Suppose a set of feature data is obtained in real time in the distribution network monitoring system, including the mean value, maximum value, minimum value, peak value of voltage, and the main frequency and harmonic content in the spectrum, etc. Input these feature data into the trained and verified decision tree evaluation model, and the evaluation model may output the following evaluation results:

[0133] State classification: Abnormal;

[0134] Type of abnormality: Voltage fluctuation;

[0135] Degree of abnormality: Severe (the voltage peak exceeds the safety threshold).

[0136] Based on these evaluation results, the distribution network monitoring system can immediately trigger an alarm and notify the operation and maintenance personnel to take corresponding measures to solve the problem, thus ensuring the safe and stable operation of the distribution network.

[0137] By comprehensively using the statistical features, spectral features, and time-frequency features of voltage and current, the evaluation model can capture the operating state information of the distribution network more comprehensively. After being trained with a training set and verified with a test set, the performance of the evaluation model is optimized, enabling it to more accurately map the relationship between feature indicators and the state of the distribution network, thereby improving the accuracy of state evaluation. After the evaluation model is integrated into the distribution network monitoring system, it can obtain feature data in real time and input them into the model for evaluation. This enables operation and maintenance personnel to promptly understand the current state of the distribution network, including information such as state classification, abnormal type, and abnormal degree, providing the possibility for rapid response and handling of potential problems. By analyzing the change trend of feature indicators and the abnormal information output by the model, potential faults in the distribution network can be predicted. At the same time, by combining advanced methods such as time-frequency analysis, the location and cause of the fault can be more accurately located, providing strong support for the rapid elimination of faults. According to the evaluation results, operation and maintenance personnel can formulate reasonable operation and maintenance strategies, such as adjusting equipment parameters, optimizing load distribution, arranging maintenance plans, etc., to improve the operation efficiency and reliability of the distribution network. The construction and application of the evaluation model are important components of the intelligent development of the distribution network. Through real-time state evaluation and intelligent decision support, refined management and optimized control of the distribution network can be achieved, promoting the distribution network to develop towards a more intelligent, efficient, and reliable direction. Accurate state evaluation and timely fault handling can reduce the power outage time and scope, and improve the continuity and stability of power supply.

[0138] In a preferred embodiment of the present invention, step 15, according to the real-time evaluation result of the distribution network state, obtaining the result of abnormal identification and location through pattern recognition, may include:

[0139] Step 151, extracting feature indicators reflecting the current state of the distribution network from the real-time evaluation result of the distribution network state, including the statistical features, spectral features, and time-frequency features of voltage and current;

[0140] Step 152, matching the feature indicators with a preset abnormal pattern library to obtain a matching result, including the matching degree with each abnormal type;

[0141] Step 153, identifying and determining the abnormal type in the current distribution network according to the matching result and a preset threshold;

[0142] Step 154, according to the abnormal type in the current distribution network, determining the source data of the abnormality and the corresponding monitoring points, and analyzing the propagation and distribution of the source data of the abnormality in the distribution network to determine the location where the abnormality occurs, obtaining the result of abnormal identification and location.

[0143] In the embodiments of the present invention, characteristic indicators reflecting the current state of the distribution network are extracted from the real-time evaluation results of the distribution network state. These indicators include statistical characteristics of voltage and current (such as mean, standard deviation, maximum value, minimum value, etc.), spectral characteristics (such as main frequency, harmonic content, etc.), and time-frequency characteristics (such as wavelet coefficients, time-frequency energy distribution, etc.). The extracted characteristic indicators are matched with a preset abnormal pattern library. The abnormal pattern library is constructed based on historical data and empirical knowledge and contains characteristic patterns of various abnormal types (such as voltage fluctuations, current overloads, harmonic pollution, etc.). According to the matching results and preset thresholds, the abnormal types in the current distribution network are identified and determined. If the matching degree of a certain abnormal pattern exceeds the preset threshold, it is considered that an abnormality of this type has occurred in the distribution network. According to the abnormal types in the current distribution network, the source data of the abnormality and the corresponding monitoring points are determined. The source data refers to the initial data that causes the abnormality, such as voltage or current data at a specific location. The propagation and distribution of the source data of the abnormality in the distribution network are analyzed, and the specific location where the abnormality occurs is determined by tracking the flow direction and influence range of the abnormal data. The results of abnormal identification and location are presented to the operation and maintenance personnel in an intuitive manner, including information such as abnormal types, abnormal locations, and degrees of abnormality, so that the operation and maintenance personnel can quickly respond and handle them.

[0144] Through pattern matching, the system can identify the abnormal types occurring in the distribution network, such as voltage fluctuations, current overloads, harmonic pollution, etc. For example, if the characteristic indicators show that the mean and standard deviation of the voltage suddenly increase, and obvious harmonic components appear in the spectrum, the system may identify that this is an abnormal type where voltage fluctuations and harmonic pollution occur simultaneously. By analyzing the source data and propagation path, the system can locate the specific location where the abnormality occurs, such as a certain substation, a certain feeder, or a specific electrical equipment. For example, if the abnormal data is mainly concentrated on the outgoing side of a certain substation and the propagation range of this data in the distribution network is small, the system may determine that the source of the abnormality is inside the substation or on the feeder directly connected to it. Based on the abnormal identification and location results output by the system, the operation and maintenance personnel can quickly locate the location where the abnormality occurs and take corresponding treatment measures, such as adjusting equipment parameters, replacing faulty components, optimizing load distribution, etc. This not only improves the operation and maintenance efficiency but also reduces the power outage time and scope caused by untimely abnormal handling, improving the continuity and stability of power supply.

[0145] Suppose a set of characteristic data is obtained in real time in the distribution network monitoring system. After being processed by the evaluation model, the real-time evaluation result of the distribution network state is obtained. The system matches this result with the preset abnormal pattern library and finds that the matching degree with the abnormal pattern of "voltage fluctuation + harmonic pollution" is the highest and exceeds the preset threshold. The system further analyzes the source data and finds that the abnormality is mainly concentrated on the outgoing side of a certain substation, and the spread range of this data in the distribution network is small. Combining the topological structure and historical data of the distribution network, the system judges that the abnormal source may be located in a certain transformer inside the substation or on the feeder directly connected to it. According to the system's prompt, the operation and maintenance personnel quickly locate the position where the abnormality occurs and take corresponding treatment measures, such as adjusting the parameters of the transformer, replacing aging components, etc. After treatment, the distribution network resumes normal operation, avoiding power outages caused by abnormalities.

[0146] By extracting the statistical features, spectral features, and time-frequency features of voltage and current from the real-time evaluation result, the current state of the distribution network can be more comprehensively reflected. Matching these characteristic indicators with the preset abnormal pattern library and utilizing the powerful classification and recognition capabilities of pattern recognition technology can more accurately identify the abnormal types in the distribution network and reduce the situations of false alarms and missed alarms. After identifying the abnormal type, by tracing the source data of the abnormality and its propagation and distribution in the distribution network, the specific location where the abnormality occurs can be quickly located. The accurate abnormal identification and location results provide strong support for the rapid handling of faults. Targeted treatment measures can be quickly taken, such as adjusting equipment parameters, replacing faulty components, etc., to promptly restore the normal operation state of the distribution network and reduce the power outage time and scope. Timely abnormal identification and fault handling can reduce the failure rate of the distribution network and improve its reliability and stability. This can not only ensure the normal power consumption needs of users but also reduce the economic losses and social impacts caused by faults. Accurate abnormal identification and rapid fault handling can reduce the power outage time and frequency of users and improve the continuity and stability of power supply.

[0147] In another preferred embodiment of the present invention, in step 152 above, matching the characteristic indicators with the preset abnormal pattern library to obtain a matching result, including the matching degree with each abnormal type, may include:

[0148] Step 1523, traversing the characteristic patterns of each abnormal type in the abnormal pattern library;

[0149] Step 1524, for each abnormal type, calculating the similarity between the extracted characteristic indicators and the corresponding characteristic pattern, and sorting the similarities to obtain the matching result.

[0150] In an embodiment of the present invention, each abnormal type in a preset abnormal pattern library is traversed. The abnormal pattern library usually contains characteristic patterns of multiple known abnormal types, and these characteristic patterns are constructed based on means such as historical data, expert experience, and simulation. For each abnormal type, the system needs to align the extracted characteristic indicators with the corresponding characteristic patterns. This includes ensuring the consistency of the characteristic indicators and the characteristic patterns in terms of dimension, unit of measurement, and value range. On the basis of characteristic alignment, the similarity between the extracted characteristic indicators and the characteristic patterns of each abnormal type is calculated. Let the extracted characteristic indicator vector be X = (x1, x2, …, x n ), where x i represents the value of the i-th characteristic, and n is the number of characteristics. At the same time, let the characteristic pattern vector of a certain abnormal type be y = (y1, y2, …, y n ), where y i represents the pattern value of this abnormal type on the i-th characteristic. Then, the cosine similarity sim(x, y) between the extracted characteristic indicator x and the characteristic pattern y of this abnormal type can be calculated by the following formula:

[0151]

[0152] The system sorts each abnormal type according to the calculated similarity to obtain the abnormal type that best matches the extracted characteristic indicators. The basis for sorting is the magnitude of the similarity. The higher the similarity, the closer the extracted characteristic indicators are to the characteristic pattern of this abnormal type. Finally, the matching results are output, including the matching degree (i.e., similarity) with each abnormal type and the sorting information. This helps the operation and maintenance personnel quickly understand the current state of the distribution network and make corresponding decisions.

[0153] Suppose a set of characteristic data is obtained in real time in the distribution network monitoring system, including the mean value, standard deviation, main frequency, harmonic content, etc. of the voltage. At the same time, the preset abnormal mode library contains three abnormal types: voltage fluctuation, current overload, and harmonic pollution. The system first traverses the three abnormal types in the abnormal mode library: voltage fluctuation, current overload, and harmonic pollution. For each abnormal type, the system ensures that the extracted characteristic indicators (such as the mean value, standard deviation, etc. of the voltage) are consistent with the characteristic patterns in the abnormal mode library in terms of dimension and unit. Calculate the similarity between the extracted characteristic indicators and the characteristic patterns of each abnormal type. For example, for the voltage fluctuation abnormal type, it may be found that the extracted voltage mean value and standard deviation are relatively close to the voltage fluctuation characteristic pattern, so a relatively high similarity is calculated. The system sorts the abnormal types according to the similarity. Suppose the system calculates that the similarity with the voltage fluctuation abnormal type is the highest, followed by harmonic pollution, and finally current overload. The system outputs the matching result, showing that the similarity of the voltage fluctuation abnormal type is the highest, so it is judged that the distribution network may currently have a voltage fluctuation abnormality. At the same time, the system also provides the similarity and sorting information of each abnormal type for the operation and maintenance personnel to refer to.

[0154] By matching with the preset abnormal mode library, the system can more accurately identify the abnormal types in the distribution network, reducing the situations of misdiagnosis and missed diagnosis. The matching result can quickly point out the possible problem areas or abnormal types in the distribution network, helping the operation and maintenance personnel quickly locate and conduct fault troubleshooting. The automated matching process shortens the time for the operation and maintenance personnel to manually analyze and diagnose, improving the overall operation and maintenance efficiency. The matching result provides the similarity values and sorting information of each abnormal type, providing more comprehensive and objective data support, which helps to make more informed decisions. Through the accurate matching result, resources can be allocated more targeted, such as giving priority to dealing with the abnormal types with high similarity, so as to optimize the use efficiency of resources.

[0155] In a preferred embodiment of the present invention, in step 16 above, according to the results of abnormal identification and location, warning information is generated, including the abnormal type and the abnormal location, which may include:

[0156] Determine the most likely type of anomaly according to probability ranking. If the monitoring system provides location information (such as GPS coordinates, device numbers, etc.), obtain it directly; otherwise, it may be necessary to infer based on the data source of the characteristic indicators or the network topology. Use the obtained location information, combined with the results of anomaly recognition, to determine the specific location where the anomaly occurred. Verify the accuracy of the positioning result by comparing it with other data sources (such as on-site inspections, historical records, etc.). According to the results of anomaly recognition, clearly indicate the type of anomaly (such as overload, short circuit, voltage fluctuation, etc.). According to the results of anomaly positioning, provide the specific location where the anomaly occurred (such as device number, geographical location coordinates, area name, etc.), including the severity, occurrence time, duration, and possible affected range of the anomaly. Select an appropriate warning format (such as text message, email, SMS, system pop-up window, etc.) according to the needs of the system or personnel receiving the warning information. Send the constructed warning information to the relevant operation and maintenance personnel through the selected communication channel.

[0157] In a preferred embodiment of the present invention, traverse the characteristic patterns of each type of anomaly in the anomaly pattern library, including:

[0158] Traverse each type of anomaly in the anomaly pattern library, and perform the following steps for the currently traversed type of anomaly:

[0159] Obtain the characteristic pattern of this type of anomaly, and the characteristic pattern is a regular expression;

[0160] In the log or data stream to be detected, use the compiled regular expression for searching and matching. If a matching item is found, record the position, content, and type of anomaly of the match;

[0161] For each matching item, generate a corresponding anomaly report or warning according to the type of anomaly and the matching content;

[0162] After completing the matching of the current type of anomaly, continue to traverse the next type of anomaly until all types of anomalies have been traversed; The warning information includes the type of anomaly and the location of the anomaly.

[0163] In an embodiment of the present invention, the feature patterns defined by regular expressions can accurately match and identify abnormal patterns in logs or data streams. This method is more flexible and accurate than simple text search or rule-based methods, and can effectively reduce false positives and false negatives. Once a match is detected, the system can automatically record the location, content, and type of the abnormality of the match, and generate corresponding abnormality reports or warnings. This automated processing can greatly reduce the burden of manual monitoring and improve the efficiency and response speed of abnormality handling. By traversing each type of abnormal pattern in the abnormal pattern library, the system can cover a wider range of abnormal situations. This means that whether it is a known common abnormality or a newly emerging unknown abnormality, as long as its feature pattern is defined in the pattern library, it may be detected and reported in a timely manner. Generating corresponding abnormality reports or warnings according to the type of abnormality and the content of the match makes the reports more targeted and informative. This helps relevant personnel quickly understand the details of the abnormality and thus make correct handling and decisions. Timely detection and handling of abnormalities are important means to ensure the stability and security of the system. By traversing the abnormal pattern library and using regular expressions for matching, intervention can be carried out at the initial stage of the occurrence of the abnormality to prevent the spread of the abnormality and cause more serious consequences.

[0164] In a preferred embodiment of the present invention, according to the results of abnormal recognition and positioning, early warning information is generated, including:

[0165] Determine the basic structure of the early warning information according to the result data of abnormal recognition and positioning;

[0166] Create an early warning information object or data structure for storing the basic structure;

[0167] Convert the abnormal type identifier into a text description, and fill the device name, line number, and specific node of the abnormal location into the early warning information according to the result of abnormal positioning;

[0168] Configure the sending parameters, execute the sending operation, and transmit the early warning information to the designated recipient; the basic structure includes an early warning title, a description of the abnormal type, and a description of the abnormal location.

[0169] In the embodiments of the present invention, by automatically generating warning information and immediately sending it to relevant personnel, it can ensure that anomalies are detected and processed in a timely manner. This rapid response mechanism can significantly reduce the time from the occurrence of an anomaly to the resolution of the problem and reduce potential losses. The warning information includes a description of the anomaly type and a description of the anomaly location, providing clear instructions to the recipient. This enables relevant personnel to quickly understand the nature and location of the problem, and thus take targeted measures. Since the warning information has classified and located the anomaly, the processing personnel can save the time for preliminary analysis and location and directly solve the problem. This greatly improves the efficiency of anomaly handling. Accurate anomaly type and location information helps to reduce the possibility of misoperation. The processing personnel can avoid blind operations without understanding the specific situation based on the accurate description in the warning information, thereby reducing the risk of further causing problems. By detecting and processing anomalies in a timely manner, the system downtime can be reduced and the overall reliability of the system can be improved.

[0170] In a preferred embodiment of the present invention, the anomaly type identifier is converted into a text description, and according to the result of anomaly location, the device name, line number, and specific node of the anomaly location are filled into the warning information, including:

[0171] Create an anomaly type mapping table, which is used to map the internal identifier of the anomaly type to the corresponding text description;

[0172] When the anomaly recognition result is obtained, extract the anomaly type identifier therein, use the anomaly type identifier as an index to look up the corresponding text description in the anomaly type mapping table. If a matching text description is found, save it as part of the warning information; if no matching item is found, record a default or general anomaly type description;

[0173] Obtain the detailed information of the anomaly location from the anomaly location module, including the device name, line number, and specific node;

[0174] Create a warning information template, which contains placeholders for filling in the anomaly type and anomaly location information, and fill the queried anomaly type text description into the corresponding placeholder position in the warning information template;

[0175] According to the anomaly location result, fill in the information such as the device name, line number, and specific node into the corresponding positions in the warning information template.

[0176] In an embodiment of the present invention, the anomaly type identifier is converted into a human-readable text description through a mapping table, which greatly improves the readability of the warning information. This enables the recipient to quickly understand the nature of the anomaly without referring to additional documents or materials. Using a unified mapping table and warning information template ensures the standardization and consistency of the anomaly description. Regardless of which system or module generates the warning, its format and content will follow the same standard, facilitating subsequent processing and analysis. The design of the mapping table and template allows for the easy addition of new anomaly types and location information. When a new anomaly type or device is introduced into the system, only the mapping table and template need to be updated, without modifying the entire warning generation logic. By filling in details such as the device name, line number, and specific node, the warning information provides accurate anomaly location. This helps to quickly locate the source of the problem, reduce the troubleshooting time, and improve the efficiency of fault repair. Since the warning information contains rich context information (such as anomaly type and location), the processing personnel can make response decisions faster without waiting for additional information collection or confirmation.

[0177] In a preferred embodiment of the present invention, send parameters are configured and a sending operation is performed to deliver the warning information to a specified recipient, including:

[0178] Determine the list of potential recipients of the warning information and their contact information;

[0179] Set the initial temperature T0, the termination temperature Tf, the cooling coefficient α, and the number of iterations L at each temperature of the simulated annealing algorithm;

[0180] Define an evaluation function to evaluate the quality of the sending strategy;

[0181] Randomly generate an initial sending strategy, including the selected recipients, sending channels, and sending times; calculate the objective function value of the initial solution; set the current temperature T to the initial temperature T0, and enter a loop until the current temperature T drops below the termination temperature Tf; for the current temperature T, perform L iterations; based on the current strategy, randomly generate a new sending strategy; calculate the objective function value of the new strategy, and determine whether to accept the new sending strategy, and update the current strategy and its objective function value; after completing all iterations at the current temperature, lower the temperature according to the cooling coefficient α; when the temperature drops below the termination temperature Tf, exit the simulated annealing process, and output the current final strategy and its objective function value as the optimized sending strategy;

[0182] According to the optimized sending strategy, configure the sending parameters and perform the sending operation to send the warning information to the specified recipient according to the optimized sending strategy.

[0183] In the embodiments of the present invention, through the simulated annealing algorithm, an approximate optimal solution can be found among numerous possible transmission strategies. This means that early warning information can be sent more accurately to the most suitable recipients, improving the efficiency and pertinence of information transmission. The simulated annealing algorithm has a powerful global search ability and can avoid falling into local optimal solutions. This ensures that the transmission strategy can still find a globally better solution when considering various factors (such as recipient availability, transmission channel reliability, transmission time sensitivity, etc.). Since the simulated annealing algorithm allows accepting worse solutions with a certain probability, the transmission strategy can remain flexible and adaptable when facing a complex and changeable environment. For example, when some recipients are unreachable or some transmission channels fail, the algorithm can adjust the strategy to adapt to these situations. By optimizing the transmission strategy, unnecessary transmission operations can be avoided, thereby reducing the waste of network bandwidth, server resources, etc. This helps to reduce operating costs and improve the overall efficiency of the system. The accurate and timely transmission of early warning information is crucial for the user experience. By optimizing the transmission strategy through the simulated annealing algorithm, it can be ensured that users receive relevant information when they most need it, thereby enhancing user satisfaction and trust in the system. Among them, the calculation formula of the evaluation function f(x) is:

[0184]

[0185] Among them, w1, w2, and w3 are weight coefficients; e c is the efficiency of the transmission channel; r i is the efficiency of recipient i, and the range is between [0, 1]; N is the number of recipients; i is the index value of the recipient; t is the current hour, and t ∈ [0, 24].

[0186] When specifically applied, it specifically includes:

[0187] For the current temperature T, perform L iterations; based on the current strategy, randomly generate a new transmission strategy (the recipient, transmission channel, or transmission time can be randomly changed), and calculate the objective function value of the new strategy. Determine whether to accept the new strategy:

[0188] If the objective function value of the new strategy is better than the current strategy, accept the new strategy; if the objective function value of the new strategy is not better than the current strategy, accept the new strategy with a certain probability, and this probability is set to where ΔE is the difference between the objective function values of the new strategy and the current strategy, and T is the current temperature. If the new strategy is accepted, set it as the current strategy and update the objective function value.

[0189] As Figure 2 shown, the embodiments of the present invention also provide a distribution network abnormal state perception system based on data processing, including:

[0190] An acquisition module for real-time acquisition of various operation data in a distribution network, including voltage, current, power factor, frequency, and equipment status information; an acquisition objective function for setting differential evolution parameters; the differential evolution parameters including population size, number of iterations, mutation factor, and crossover probability; iteratively optimizing the configuration of the differential evolution parameters, preprocessing the various operation data, including removing outliers and noise, and performing normalization processing to obtain processed data; extracting characteristic indicators reflecting the operation status of the distribution network from the processed data, including statistical characteristics, spectral characteristics, and time-frequency characteristics of voltage and current;

[0191] A processing module for constructing an evaluation model based on the characteristic indicators to evaluate the current status of the distribution network and obtain a real-time evaluation result of the distribution network status;

[0192] An evaluation module for obtaining the results of anomaly identification and location through pattern recognition according to the real-time evaluation result of the distribution network status;

[0193] A warning module for generating warning information, including anomaly type and anomaly location, according to the results of anomaly identification and location.

[0194] It should be noted that this system corresponds to the above method. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0195] An embodiment of the present invention further provides a computing device, including: a processor and a memory storing a computer program. When the computer program is run by the processor, it executes the method as described above. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0196] An embodiment of the present invention further provides a computer-readable storage medium storing instructions. When the instructions are run on a computer, the computer is caused to execute the method as described above. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0197] The above are the preferred embodiments of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for sensing abnormal state of distribution network based on data processing, characterized in that: The method comprises: Obtain various operating data in the distribution network in real time, including voltage, current, power factor, frequency and equipment status information; Obtaining the objective function and setting differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor and crossover probability; iteratively optimizing the configuration of differential evolution parameters, preprocessing various operating data, including removing outliers and noise, and performing normalization processing to obtain processed data; Extract characteristic indicators reflecting the operating status of the distribution network from the processed data, including statistical characteristics, spectrum characteristics and time-frequency characteristics of voltage and current; According to the characteristic indicators, an evaluation model is constructed to evaluate the current state of the distribution network and obtain real-time evaluation results of the distribution network state; Based on the real-time evaluation results of the distribution network status, the results of abnormal identification and positioning are obtained through pattern recognition; Generate early warning information based on the results of anomaly identification and positioning.

2. The method for sensing abnormal state of distribution network based on data processing according to claim 1 is characterized in that: Obtaining an objective function and setting differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor and crossover probability; Iteratively optimize the configuration of differential evolution parameters, pre-process various operating data, including removing outliers and noise, and performing normalization to obtain processed data, including: Define an objective function and set the population size, upper limit of iterations, mutation factor and crossover probability; According to the preset population size, a set of parameter candidate sets is randomly generated as the initial population; The processing effect of each parameter candidate set in the initial population is evaluated to obtain an evaluation result, and the parameter set is determined according to the evaluation result, and the determined parameter set is used as the parent generation.

3. The method for sensing abnormal state of distribution network based on data processing according to claim 2 is characterized in that: Obtaining an objective function and setting differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor and crossover probability; Iteratively optimize the configuration of differential evolution parameters, pre-process various operating data, including removing outliers and noise, and normalizing to obtain processed data, including: Perform mutation operation on the determined parameter set to generate a new parameter combination, and perform crossover operation on the generated new parameter combination. According to the preset crossover probability, select some parameters for exchange, and generate a new set of parameter candidate solutions through mutation and crossover operations; Substitute the new parameter candidate solution into the objective function for re-evaluation to obtain the evaluation result, and update the population according to the evaluation result; Repeat the steps of parameter variation, crossover, objective function evaluation and population update until the preset number of iterations is reached to obtain the final parameter configuration; According to the final parameter configuration, various operating data are normalized to obtain processed data.

4. The method for sensing abnormal state of distribution network based on data processing according to claim 3 is characterized in that: Extract characteristic indicators reflecting the operating status of the distribution network from the processed data, including statistical characteristics, spectrum characteristics and time-frequency characteristics of voltage and current, including: Obtain time series data of electrical parameters from the processed data, including voltage, current, power, and frequency; According to the time series data, the time domain characteristics of the voltage and current signals are calculated, including the mean, maximum, minimum, and peak values, to form the time domain analysis results; The voltage and current signals are subjected to Fourier transform and wavelet transform respectively. The signals are converted from the time domain to the frequency domain through Fourier transform to extract spectrum information. The frequency components of the signals at different time points are obtained through wavelet transform to obtain the frequency domain analysis results. The spectrum information includes the main frequency, harmonic components and spectrum energy. According to the results of time domain and frequency domain analysis, statistical features are extracted, including mean, skewness, and kurtosis; features are extracted from spectrum analysis, including main frequency, harmonic content, and spectrum entropy; features are extracted from time-frequency analysis, including wavelet coefficients and time-frequency energy distribution.

5. The method for sensing abnormal state of distribution network based on data processing according to claim 4 is characterized in that: The characteristic indicators include statistical characteristics of voltage and current as well as spectrum characteristics and time-frequency characteristics; According to the characteristic indicators, an evaluation model is constructed to evaluate the current state of the distribution network and obtain real-time evaluation results of the distribution network state, including: Divide the feature indicators into training sets and test sets, and select the corresponding evaluation model; Configure the parameters of the evaluation model, including the depth of the decision tree, and use the training set to train the evaluation model; During the training process, the evaluation model learns the mapping relationship between the characteristic indicators and the distribution network status to obtain a post-training evaluation model; Use the test set to verify the trained evaluation model, calculate the evaluation indicators, including accuracy and recall, and obtain the verification results of the evaluation model based on the evaluation indicators; According to the verification results of the evaluation model, the depth parameters of the evaluation model are adjusted, and the evaluation model is retrained with the adjusted parameters and verified again. The adjustment, training and verification process is repeated until the performance of the evaluation model reaches the preset standard, and a trained and verified evaluation model is obtained; Integrate the trained and verified evaluation model into the distribution network monitoring system to obtain feature data from the distribution network in real time, including statistical features, spectrum features, and time-frequency features of voltage and current; The characteristic data acquired in real time is input into the evaluation model for real-time evaluation to obtain the real-time evaluation results of the distribution network status, including status classification, abnormality type, and abnormality degree information.

6. The method for sensing abnormal state of distribution network based on data processing according to claim 5 is characterized in that: Based on the real-time evaluation results of the distribution network status, pattern recognition is used to obtain the results of abnormal identification and positioning, including: Extract characteristic indicators reflecting the current state of the distribution network from the real-time evaluation results of the distribution network state, including statistical characteristics, spectrum characteristics and time-frequency characteristics of voltage and current; Match the characteristic indicators with the preset abnormal pattern library to obtain matching results, including the matching degree with each abnormal type; According to the matching results and the preset threshold, identify and determine the abnormal type in the current distribution network; According to the abnormal type in the current distribution network, the source data of the abnormality and the corresponding monitoring point are determined, and the propagation and distribution of the source data of the abnormality in the distribution network are analyzed to determine the location where the abnormality occurs, and obtain the results of abnormality identification and positioning.

7. The method for sensing abnormal state of distribution network based on data processing according to claim 6 is characterized in that: Match the characteristic indicators with the preset abnormal pattern library to obtain the matching results, including the degree of matching with each abnormal type, including: Traverse the characteristic pattern of each abnormal type in the abnormal pattern library; For each anomaly type, the similarity between the extracted feature indicators and the corresponding feature patterns is calculated, and the similarities are sorted to obtain the matching results.

8. The method for sensing abnormal state of distribution network based on data processing according to claim 7 is characterized in that: Traverse the characteristic patterns of each exception type in the exception pattern library, including: Traverse each exception type in the exception pattern library and perform the following steps for the currently traversed exception type: Get the characteristic pattern of the exception type, which is a regular expression; In the log or data stream to be detected, the compiled regular expression is used to search and match. If a match is found, the matching location, content, and exception type are recorded. For each matching item, a corresponding exception report or warning is generated based on the exception type and matching content; After completing the matching of the current exception type, continue to traverse the next exception type until all exception types have been traversed.

9. The method for sensing abnormal state of distribution network based on data processing according to claim 8, characterized in that: The warning information includes the abnormal type and abnormal location.

10. The method for sensing abnormal state of distribution network based on data processing according to claim 9, characterized in that: Generate warning information based on the results of abnormal identification and positioning, including: Determine the basic structure of the warning information based on the result data of abnormal identification and location; Create an early warning information object or data structure to store the basic structure; Convert the abnormal type identifier into a text description, and fill the device name, line number and specific node of the abnormal location into the warning information according to the abnormal location result; Configure sending parameters, execute sending operations, and pass the warning information to the specified recipients.

11. The method for sensing abnormal state of distribution network based on data processing according to claim 10, characterized in that: The basic structure includes the warning title, the exception type description and the exception location description.

12. The method for sensing abnormal state of distribution network based on data processing according to claim 11, characterized in that: Convert the abnormal type identifier into a text description. According to the abnormal location result, fill the device name, line number and specific node of the abnormal location into the warning information, including: Create an exception type mapping table, which is used to map the internal identifier of the exception type to the corresponding text description; When the anomaly recognition result is obtained, the anomaly type identifier is extracted, and the anomaly type identifier is used as an index to find the corresponding text description in the anomaly type mapping table. If a matching text description is found, it is saved as part of the warning information; if no match is found, a default or general anomaly type description is recorded; Obtain detailed information about the abnormal location from the abnormal location module, including device name, line number, and specific node; Create an early warning information template, which contains placeholders for filling in the abnormal type and abnormal location information, and fill the queried abnormal type text description into the corresponding placeholder position in the early warning information template; According to the abnormal location results, fill in the equipment name, line number and specific node into the corresponding position in the warning information template.

13. The method for sensing abnormal state of distribution network based on data processing according to claim 12, characterized in that: Configure sending parameters, execute sending operations, and deliver warning information to the specified recipients, including: Identify a list of potential recipients of warning information and their contact information; Set the initial temperature T0, the end temperature Tf, the temperature reduction coefficient α and the number of iterations L at each temperature of the simulated annealing algorithm; Define an evaluation function to evaluate the quality of the sending strategy; Randomly generate an initial sending strategy, including the selected receiver, sending channel and sending time; calculate the objective function value of the initial solution; set the current temperature T to the initial temperature T0, enter the loop until the current temperature T drops below the termination temperature Tf; for the current temperature T, perform L iterations; based on the current strategy, randomly generate a new sending strategy; calculate the objective function value of the new strategy, and determine whether to accept the new sending strategy, and update the current strategy and its objective function value; after completing all iterations at the current temperature, reduce the temperature according to the cooling coefficient α; when the temperature drops below the termination temperature Tf, exit the simulated annealing process, and output the current final strategy and its objective function value as the optimized sending strategy; According to the optimized sending strategy, configure the sending parameters, execute the sending operation, and send the warning information to the designated recipient according to the optimized sending strategy.

14. A distribution network abnormal state perception system based on data processing, the system implementing the method according to any one of claims 1 to 13, characterized in that: include: The acquisition module is used to obtain various operating data in the distribution network in real time, including voltage, current, power factor, frequency and equipment status information; Obtaining the objective function and setting differential evolution parameters; the differential evolution parameters include population size, number of iterations, mutation factor and crossover probability; iteratively optimizing the configuration of differential evolution parameters, preprocessing various operating data, including removing outliers and noise, and performing normalization processing to obtain processed data; Extract characteristic indicators reflecting the operating status of the distribution network from the processed data, including statistical characteristics, spectrum characteristics and time-frequency characteristics of voltage and current; A processing module is used to construct an evaluation model based on characteristic indicators to evaluate the current state of the distribution network and obtain a real-time evaluation result of the distribution network state; An evaluation module is used to obtain abnormality identification and location results through pattern recognition based on the real-time evaluation results of the distribution network status; The early warning module is used to generate early warning information based on the results of abnormality identification and positioning, including abnormality type and abnormal location.

15. A computing device, characterized in that: include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 13.

16. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which implements the method according to any one of claims 1 to 13 when executed by a processor.

Images