Power distribution network real-time operation state sensing method and system based on composite disaster

Abstract

The invention discloses a power distribution network real-time operation state sensing method and system based on composite disasters, and the method comprises the steps: S1, collecting a multi-source data set of a power distribution network of an environment index, an equipment operation index and a human factor index, and carrying out the preliminary screening of the multi-source data set; s2, performing data dimension reduction on the multi-source data set subjected to preliminary screening processing based on an auto-encoder to obtain a feature data set; s3, establishing a GMM-based power distribution network real-time operation state sensing model, and inputting the feature data set into the power distribution network real-time operation state sensing model for clustering analysis to obtain an optimal classification result; and S4, collecting a multi-source data set of the real-time power distribution network, and repeating the steps to obtain a composite disaster classification result. By integrating data from different sources, comprehensive and diversified information is obtained, and the real-time operation state and the disaster possibility of the power distribution network are predicted more accurately.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid disaster perception and early warning, and in particular relates to a method and system for perceiving the real-time operating status of a distribution network based on compound disasters. Background Art

[0002] In today's society, power supply is an important foundation for modern life and economic development. However, the power grid system faces many challenges, including compound disasters caused by multiple factors such as natural disasters, man-made damage and equipment aging. These challenges pose a huge threat to the safe operation and reliability of the power grid. In this case, it is crucial to detect and warn of possible disaster events in a timely and accurate manner. However, the current power grid detection and warning technology has some shortcomings in dealing with compound disasters.

[0003] First, current power grid detection and early warning technologies usually only monitor and warn of a single type of disaster, such as monitoring systems for power grid equipment failures such as circuit breaks and short circuits, or monitoring systems for natural disasters such as lightning and storms. This single type of monitoring and early warning technology is often unable to cope with compound disasters, because compound disasters are often caused by the superposition of multiple factors, and current technology cannot accurately determine what type of disaster has occurred, nor can it provide corresponding response measures.

[0004] Secondly, existing power grid detection and early warning technologies often only use a single data source for feature extraction, and lack the ability to conduct comprehensive analysis and feature extraction of multi-source data. When a compound disaster occurs, different types of data sources (such as sensor data, meteorological data, equipment operation data, etc.) may provide different aspects of information. The comprehensive use of these data sources can provide a more comprehensive understanding of the operating status of the power grid and possible disaster risks. However, current technologies often cannot effectively integrate and analyze multi-source data, resulting in an inability to accurately judge the occurrence and impact of compound disasters.

[0005] In addition, existing power grid detection and early warning technologies also have some limitations in terms of algorithms and models. Traditional monitoring and early warning methods often use rules or empirical models for judgment. This method often lacks the ability to effectively utilize large-scale data and in-depth analysis, and is difficult to cope with complex and changing compound disaster scenarios.

[0006] The Chinese patent with the publication number CN103914951A discloses a method for mining and early warning analysis of natural disaster data based on GIS power grid. First, the government provides information on possible disasters such as earthquakes, floods, snowstorms, lightning strikes, gales, wildfires, landslides and mudslides in advance, and uses hierarchical analysis, artificial neural networks, logistic regression, statistics and other methods to establish a multi-correlation composite disaster evaluation model to evaluate and analyze various types of disaster information in the past. The evaluation results are comprehensively compared and analyzed, and the example data mining technology is used to provide emergency decisions for various disasters. Finally, the consequences of the disaster and emergency decisions are visualized on the key control lines and equipment information of the power grid. The method of the invention mentions the evaluation and analysis of natural disaster information that has occurred in the past, but there is no detailed description on how to comprehensively use multi-source data for analysis. Summary of the invention

[0007] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a method and system for perceiving the real-time operating status of a distribution network based on compound disasters. By integrating data from different sources, comprehensive and diverse information is obtained to more accurately predict the real-time operating status of the distribution network and the possibility of disasters.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] On the one hand, the present invention provides a method for sensing the real-time operating status of a distribution network based on a composite disaster, comprising the following steps:

[0010] S1: Collect multi-source data sets of distribution network including environmental indicators, equipment operation indicators and human factor indicators, and perform preliminary screening and processing on the multi-source data sets.

[0011] S2: Perform data dimension reduction based on the autoencoder on the multi-source data set that has been preliminarily screened to obtain a feature data set.

[0012] S3: Establish a GMM-based distribution network real-time operation status perception model, input the feature data set into the distribution network real-time operation status perception model for cluster analysis, and obtain the optimal classification result.

[0013] S4: Collect multi-source data sets of real-time distribution network and repeat the above steps to obtain composite disaster classification results.

[0014] Preferably, the step S1 specifically comprises:

[0015] S11: Collect multi-source data sets of distribution network including environmental indicators, equipment operation indicators and human factor indicators and pre-process the multi-source data sets.

[0016] S12: Calculate the covariance matrix of the preprocessed multi-source data set, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors.

[0017] S13: Select the number of principal components to be retained based on the Kaiser criterion for the eigenvalues.

[0018] S14: Projecting the original multi-source data set onto the selected principal component to obtain the multi-source data set after preliminary screening processing.

[0019] Preferably, the step S2 is specifically:

[0020] S21: Construct an autoencoder model based on a neural network, input a multi-source data set that has been preliminarily screened and processed into the autoencoder model, and optimize the autoencoder model parameters by minimizing the reconstruction error.

[0021] S22: When the training reaches the set maximum number of iterations or the reconstruction error reaches the set threshold, stop training the autoencoder model.

[0022] Preferably, the step S3 is specifically:

[0023] S31: Establish a real-time operation status perception model of the distribution network based on GMM, specifically:

[0024]

[0025] Where P(x) is the real-time operating state perception model of the distribution network, π i is the mixing coefficient, N(x|μ i ,Σ i ) is the probability density function of each Gaussian distribution in the distribution network real-time operation status perception model, M is the total number of Gaussian distributions, μ i is the center of the i-th Gaussian distribution, Σ i is the covariance of the i-th Gaussian distribution, i∈[1,M].

[0026] S32: Use the feature data set to train the distribution network real-time operation status perception model and initialize the distribution network real-time operation status perception model parameters.

[0027] S33: Calculate the probability that each data sample in the feature data set belongs to a given Gaussian distribution.

[0028] S34: Update the parameters of the Gaussian distribution based on step S33.

[0029] S35: Check whether the new Gaussian parameter is less than the set threshold or whether the gain of the likelihood function is less than the set threshold. If it is less than the set threshold, repeat steps S33 and S34 until the distribution network real-time operation status perception model converges, and stop the distribution network real-time operation status perception model training.

[0030] On the other hand, the present invention provides a distribution network real-time operation status perception system based on compound disasters, including a data acquisition module, a data dimension reduction module, a cluster analysis module and a real-time perception module.

[0031] The data acquisition module is used to collect multi-source data sets of the distribution network including environmental indicators, equipment operation indicators and human factor indicators, and to perform preliminary screening and processing on the multi-source data sets.

[0032] The data dimension reduction module is used to perform data dimension reduction on the multi-source data set that has been preliminarily screened based on the autoencoder to obtain a feature data set.

[0033] The clustering analysis module is used to establish a distribution network real-time operation status perception model based on GMM, input the feature data set into the distribution network real-time operation status perception model for clustering analysis, and obtain the optimal classification result.

[0034] The real-time perception module is used to collect multi-source data sets of the real-time distribution network and repeat the above steps to obtain the composite disaster classification results.

[0035] Preferably, the data acquisition module realizes the collection of multi-source data sets of the distribution network including environmental indicators, equipment operation indicators and human factor indicators, and the preliminary screening and processing of the multi-source data sets is specifically as follows:

[0036] S11: Collect multi-source data sets of distribution network including environmental indicators, equipment operation indicators and human factor indicators and pre-process the multi-source data sets.

[0037] S12: Calculate the covariance matrix of the preprocessed multi-source data set, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors.

[0038] S13: Select the number of principal components to be retained based on the Kaiser criterion for the eigenvalues.

[0039] S14: Projecting the original multi-source data set onto the selected principal component to obtain the multi-source data set after preliminary screening processing.

[0040] Preferably, the data dimension reduction module implements data dimension reduction based on the autoencoder for the multi-source data set that has been preliminarily screened, and obtains the feature data set specifically as follows:

[0041] S21: Construct an autoencoder model based on a neural network, input a multi-source data set that has been preliminarily screened and processed into the autoencoder model, and optimize the autoencoder model parameters by minimizing the reconstruction error.

[0042] S22: When the training reaches the set maximum number of iterations or the reconstruction error reaches the set threshold, stop training the autoencoder model.

[0043] Preferably, the cluster analysis module realizes the establishment of a distribution network real-time operation state perception model based on GMM, inputs the feature data set into the distribution network real-time operation state perception model for cluster analysis, and obtains the optimal classification result as follows:

[0044] S31: Establish a real-time operation status perception model of the distribution network based on GMM, specifically:

[0045]

[0046] Where P(x) is the real-time operating state perception model of the distribution network, π i is the mixing coefficient, N(x|μ i ,Σ i ) is the probability density function of each Gaussian distribution in the distribution network real-time operation status perception model, M is the total number of Gaussian distributions, μ i is the center of the i-th Gaussian distribution, Σ i is the covariance of the i-th Gaussian distribution, i∈[1,M].

[0047] S32: Use the feature data set to train the distribution network real-time operation status perception model and initialize the distribution network real-time operation status perception model parameters.

[0048] S33: Calculate the probability that each data sample in the feature data set belongs to a given Gaussian distribution.

[0049] S34: Update the parameters of the Gaussian distribution based on step S33.

[0050] S35: Check whether the new Gaussian parameter is less than the set threshold or whether the gain of the likelihood function is less than the set threshold. If it is less than the set threshold, repeat steps S33 and S34 until the distribution network real-time operation status perception model converges, and stop the distribution network real-time operation status perception model training.

[0051] On the other hand, the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements a method for sensing the real-time operation status of a distribution network based on complex disasters as described in any embodiment of the present invention.

[0052] On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for sensing the real-time operating status of a distribution network based on complex disasters as described in any embodiment of the present invention.

[0053] Compared with the prior art, the present invention has the following technical effects:

[0054] 1. Traditional distribution network monitoring technology often only focuses on a single type of disaster, such as equipment failure or natural disasters. The multi-source data collected by the present invention covers multiple aspects such as weather, geography, equipment operation and human factors, making the monitoring system more comprehensive and able to comprehensively consider the impact of multiple factors on the operating status of the distribution network. Data dimension reduction and feature extraction through preliminary screening processing and autoencoder and other technologies help to discover hidden relationships and important features between data, thereby providing more accurate disaster warning information and more accurately judging the operating status of the distribution network and potential disaster risks.

[0055] 2. The present invention provides a real-time operation status perception model based on GMM, which can quickly perform cluster analysis on feature data sets, and realize real-time monitoring and perception of the operation status of the distribution network. This enables the system to detect abnormal situations in a timely manner and make corresponding responses, thereby improving the real-time performance and reliability of the power grid. Through comprehensive analysis of multi-source data, the system can not only monitor and warn of a single type of disaster, but also identify complex disaster situations. This enables distribution network managers to better understand the operation status of the power grid and provides more decision-making support for distribution network managers. Distribution network managers can adjust operation strategies in a timely manner, optimize resource allocation, reduce disaster risks, and improve the disaster resistance of the power grid based on the real-time monitoring results and warning information provided by the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of a method for sensing the real-time operation status of a distribution network based on composite disasters described in the present invention. DETAILED DESCRIPTION

[0057] In order to make the objectives, technical solutions and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in combination with specific embodiments of the present application and with reference to the accompanying drawings.

[0058] Embodiment 1

[0059] This embodiment provides a method for sensing the real-time operation status of a distribution network based on a composite disaster. Figure 1 As shown, the following steps are included:

[0060] S1: Collect multi-source data sets of the distribution network of environmental indicators, equipment operation indicators and human factor indicators, and perform preliminary screening and processing on the multi-source data sets. Specifically, environmental indicators include temperature data, humidity data, wind speed data, rainfall data, lightning frequency data, wind direction data, radiation intensity data, visibility data, evaporation data, etc., equipment operation indicators include cable temperature data, transformer load data, power supply line voltage data, power supply line current data, equipment vibration data, battery charging and discharging data, etc., and human factor indicators include personnel inspection record data, operation log data, illegal operation record data, equipment operation authorization record data, etc. There are many ways to obtain the above data information, which can be obtained through weather stations, sensors, SCADA systems, RFID technology, system logs and other methods, which are not limited in this embodiment.

[0061] As a preferred implementation of this embodiment, the step S1 is specifically as follows:

[0062] S11: Collect multi-source data sets of distribution network of environmental indicators, equipment operation indicators and human factor indicators and pre-process the multi-source data sets, wherein the pre-processing includes data cleaning, data integration and standardization. Among them, data cleaning includes processing missing values, outliers and duplicate data. Ensure the integrity and accuracy of the data. Data integration is to integrate environmental indicators, equipment operation indicators and human factor indicator data from different data sources into the data set for subsequent analysis and processing. Standardization is to standardize each indicator so that different indicators are comparable. In this embodiment, the Z-score standardization method is used for processing, and its specific calculation formula is:

[0063]

[0064] In the formula, X * is the standardized transformed multi-source dataset, X is the original multi-source dataset, μ is the mean of the multi-source dataset, and σ is the standard deviation of the multi-source dataset.

[0065] S12: Calculate the covariance matrix of the multi-source data set preprocessed by the result. The covariance matrix is ​​composed of covariance and is used to measure the correlation and change trend between data. The covariance is a statistic that measures the relationship between two feature data. The specific calculation formula is:

[0066]

[0067] In the formula, Cov(x,y) is the covariance, n is the number of samples of feature data, x i ,y i are the sample values ​​corresponding to the two feature data, is the sample mean of the two feature data.

[0068] The covariances between all feature data are placed in the covariance matrix according to the corresponding positions. If there are n feature data, the covariance matrix is ​​an n*n symmetric matrix, where the elements in the i-th row and j-th column are the covariances between the i-th and j-th feature data. Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​represent the variance in the data set, while the eigenvectors represent the main directions in the data set.

[0069] S13: The number of principal components to be retained is selected based on the Kaiser criterion for the eigenvalues. The principal components with eigenvalues ​​greater than 1 are retained based on the Kaiser criterion because these principal components contain more information than the original variables.

[0070] S14: Project the original multi-source data set onto the selected principal component to obtain a multi-source data set after preliminary screening. Through this step, the data is transformed into a new feature space, in which each dimension represents a principal component instead of the original variable of the original data. This helps to reduce the dimension of the data and retain the key information of the data.

[0071] S2: Perform data dimension reduction based on the autoencoder on the multi-source data set that has been preliminarily screened to obtain a feature data set.

[0072] As a preferred implementation of this embodiment, step S2 is specifically as follows:

[0073] S21: Construct an autoencoder model based on a neural network, input a multi-source data set that has been preliminarily screened and processed into the autoencoder model, and optimize the autoencoder model parameters by minimizing the reconstruction error. An autoencoder is an unsupervised learning model consisting of an encoder and a decoder. The encoder compresses the input data into a representation of the latent space, while the decoder reconstructs the representation back to the original data. The goal of the autoencoder is to minimize the reconstruction error, that is, the difference between the original data and the reconstructed data. Specifically, the calculation formula for the reconstruction error is:

[0074]

[0075] Where N is the number of data samples, i∈[1,N], w i is the weight factor, x i is the original data, y i To reconstruct the data.

[0076] S22: When the training reaches the set maximum number of iterations or the reconstruction error reaches the set threshold, stop training the autoencoder model to avoid overfitting and ensure that the autoencoder model will not continue to optimize indefinitely during the training process. Specifically, the maximum number of iterations can generally be set to 1000, and the reconstruction error threshold can be set to 0.001.

[0077] S3: Establish a GMM-based distribution network real-time operation status perception model, input the feature data set into the distribution network real-time operation status perception model for cluster analysis, and obtain the optimal classification result.

[0078] As a preferred implementation of this embodiment, step S3 is specifically as follows:

[0079] S31: GMM is a statistical model used to model data. It assumes that the data is composed of multiple Gaussian distributions. A distribution network real-time operation status perception model based on GMM is established. Specifically:

[0080]

[0081] Where P(x) is the real-time operating state perception model of the distribution network, π i is the mixing coefficient, N(x|μ i ,Σ i ) is the probability density function of each Gaussian distribution in the distribution network real-time operation status perception model, M is the total number of Gaussian distributions, μ i is the center of the i-th Gaussian distribution, Σ i is the covariance of the i-th Gaussian distribution, i∈[1,M].

[0082] S32: Use the feature data set to train the distribution network real-time operation status perception model, and initialize the distribution network real-time operation status perception model parameters, including the covariance matrix and mean of each Gaussian distribution.

[0083] S33: Calculate the probability that each data sample in the feature data set belongs to a given Gaussian distribution through the expectation maximization algorithm.

[0084] S34: Based on step S33, the parameters of the Gaussian distribution are updated, that is, the probability that each data sample belongs to each Gaussian distribution is calculated using the maximum likelihood estimation method, and the parameters of each Gaussian distribution are updated.

[0085] S35: Detect whether the new Gaussian parameter is less than the set threshold or whether the gain of the likelihood function is less than the set threshold. If it is less than the set threshold, repeat S33 and S34 until the distribution network real-time operation state perception model converges, and stop the distribution network real-time operation state perception model training. Specifically, the set threshold can be set to 0.001 based on experience or determined using a cross-validation method, and the gain of the likelihood function can be set to 0.01.

[0086] S4: Collect multi-source data sets of real-time distribution network, and repeat the above steps to obtain composite disaster classification results. The composite disaster classification results can be determined according to the actual distribution network area, and may include normal operation, strong wind and rainfall, lightning storm, wildfire, equipment icing, equipment short circuit, etc.

[0087] In order to verify the effectiveness and superiority of the method provided in this embodiment, some specific cases are provided below:

[0088] Taking a certain park distribution network as an example, a real-time operation status perception method of a distribution network based on compound disasters provided in this embodiment is tested. The Ubuntu operating system is selected, a Python environment is built, and Scikit-learn and TensorFlow are selected as the main frameworks. Various indicator data are collected as shown in Table 1. The classification accuracy of the method described in this embodiment and the K-means clustering algorithm is shown in Table 2. The classification accuracy of the method described in this embodiment for compound disasters is significantly higher than that of the K-means clustering algorithm.

[0089] Table 1 Index data unit table

[0090]

[0091]

[0092] Table 2 Algorithm comparison table

[0093] algorithm Classification accuracy The algorithm described in this embodiment 87.4％ K-means clustering algorithm 75.0％

[0094] Embodiment 2

[0095] Accordingly, this embodiment provides a real-time operation status perception system of a distribution network based on compound disasters, including a data acquisition module, a data dimension reduction module, a cluster analysis module and a real-time perception module.

[0096] The data acquisition module is used to collect multi-source data sets of the distribution network including environmental indicators, equipment operation indicators and human factor indicators, and to perform preliminary screening and processing on the multi-source data sets. This module is used to implement the function of step S1 in embodiment 1, which will not be described in detail here.

[0097] The data dimension reduction module is used to perform data dimension reduction on the multi-source data set that has been preliminarily screened based on the autoencoder to obtain a feature data set. This module is used to implement the function of step S2 in embodiment 1, and will not be repeated here.

[0098] The clustering analysis module is used to establish a distribution network real-time operation status perception model based on GMM, input the feature data set into the distribution network real-time operation status perception model for clustering analysis, and obtain the optimal classification result. This module is used to implement the function of step S3 in Example 1 and will not be repeated here.

[0099] The real-time perception module is used to collect multi-source data sets of the real-time power distribution network, and repeat the above steps to obtain the composite disaster classification results. This module is used to implement the function of step S4 in embodiment 1, which will not be repeated here.

[0100] Embodiment 3

[0101] This embodiment provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements a method for sensing the real-time operating status of a distribution network based on complex disasters as described in any embodiment of the present invention.

[0102] Embodiment 4

[0103] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a method for sensing the real-time operation status of a distribution network based on a composite disaster as described in any embodiment of the present invention is implemented.

[0104] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0105] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0106] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0107] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0108] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for sensing the real-time operation status of a distribution network based on compound disasters, characterized in that: The following steps are involved: S1: Collect multi-source data sets of distribution network including environmental indicators, equipment operation indicators and human factor indicators, and perform preliminary screening and processing on the multi-source data sets; S2: Perform data dimension reduction based on the autoencoder on the multi-source data set that has been preliminarily screened to obtain a feature data set; S3: Establish a distribution network real-time operation status perception model based on GMM, input the feature data set into the distribution network real-time operation status perception model for cluster analysis, and obtain the optimal classification result; S4: Collect multi-source data sets of real-time distribution network and repeat the above steps to obtain composite disaster classification results.

2. A method for sensing the real-time operation status of a distribution network based on a composite disaster according to claim 1, characterized in that: The step S1 is specifically as follows: S11: Collecting multi-source data sets of distribution network including environmental indicators, equipment operation indicators and human factor indicators and preprocessing the multi-source data sets; S12: Calculate the covariance matrix of the preprocessed multi-source data set, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors; S13: Select the number of principal components to be retained based on the Kaiser criterion for the eigenvalues; S14: Projecting the original multi-source data set onto the selected principal component to obtain the multi-source data set after preliminary screening processing.

3. A method for sensing the real-time operation status of a distribution network based on a composite disaster according to claim 1, characterized in that: The step S2 is specifically as follows: S21: Build an autoencoder model based on a neural network, input a multi-source data set that has been preliminarily screened into the autoencoder model, and optimize the autoencoder model parameters by minimizing the reconstruction error; S22: When the training reaches the set maximum number of iterations or the reconstruction error reaches the set threshold, stop training the autoencoder model.

4. A method for sensing the real-time operation status of a distribution network based on a composite disaster according to claim 1, characterized in that: The step S3 is specifically as follows: S31: Establish a real-time operation status perception model of the distribution network based on GMM, specifically: Where P(x) is the real-time operating state perception model of the distribution network, π i is the mixing coefficient, N(x|μ i ,Σ i ) is the probability density function of each Gaussian distribution in the distribution network real-time operation status perception model, M is the total number of Gaussian distributions, μ i is the center of the i-th Gaussian distribution, Σ i is the covariance of the i-th Gaussian distribution, i∈[1,M]; S32: using the feature data set to train the distribution network real-time operation state perception model and initialize the distribution network real-time operation state perception model parameters; S33: Calculate the probability that each data sample in the feature data set belongs to a given Gaussian distribution; S34: Update the parameters of the Gaussian distribution based on step S33; S35: Check whether the new Gaussian parameter is less than the set threshold or whether the gain of the likelihood function is less than the set threshold. If it is less than the set threshold, repeat steps S33 and S34 until the distribution network real-time operation status perception model converges, and stop the distribution network real-time operation status perception model training.

5. A distribution network real-time operation status perception system based on compound disasters, characterized in that: It includes data acquisition module, data dimension reduction module, cluster analysis module and real-time perception module; A data acquisition module is used to collect multi-source data sets of the distribution network including environmental indicators, equipment operation indicators and human factor indicators, and to perform preliminary screening and processing on the multi-source data sets; The data dimension reduction module is used to perform data dimension reduction on the multi-source data set that has been preliminarily screened based on the autoencoder to obtain a feature data set; The clustering analysis module is used to establish a distribution network real-time operation status perception model based on GMM, input the feature data set into the distribution network real-time operation status perception model for clustering analysis, and obtain the optimal classification result; The real-time perception module is used to collect multi-source data sets of the real-time distribution network and repeat the above steps to obtain the composite disaster classification results.

6. A distribution network real-time operation status perception system based on compound disasters according to claim 5, characterized in that: The data acquisition module collects multi-source data sets of distribution network including environmental indicators, equipment operation indicators and human factor indicators. The preliminary screening and processing of multi-source data sets are as follows: S11: Collecting multi-source data sets of distribution network including environmental indicators, equipment operation indicators and human factor indicators and preprocessing the multi-source data sets; S12: Calculate the covariance matrix of the preprocessed multi-source data set, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors; S13: Select the number of principal components to be retained based on the Kaiser criterion for the eigenvalues; S14: Projecting the original multi-source data set onto the selected principal component to obtain the multi-source data set after preliminary screening processing.

7. A distribution network real-time operation status perception system based on compound disasters according to claim 5, characterized in that: The data dimension reduction module implements data dimension reduction based on the autoencoder for the multi-source data set that has been preliminarily screened, and obtains the feature data set as follows: S21: Build an autoencoder model based on a neural network, input a multi-source data set that has been preliminarily screened into the autoencoder model, and optimize the autoencoder model parameters by minimizing the reconstruction error; S22: When the training reaches the set maximum number of iterations or the reconstruction error reaches the set threshold, stop training the autoencoder model.

8. The real-time operation status perception system of a distribution network based on compound disasters according to claim 5 is characterized in that: The cluster analysis module establishes a distribution network real-time operation status perception model based on GMM, inputs the feature data set into the distribution network real-time operation status perception model for cluster analysis, and obtains the optimal classification results as follows: S31: Establish a real-time operation status perception model of the distribution network based on GMM, specifically: Where P(x) is the real-time operating state perception model of the distribution network, π i is the mixing coefficient, N(x|μ i ,Σ i ) is the probability density function of each Gaussian distribution in the distribution network real-time operation status perception model, M is the total number of Gaussian distributions, μ i is the center of the i-th Gaussian distribution, Σ i is the covariance of the i-th Gaussian distribution, i∈[1,M]; S32: using the feature data set to train the distribution network real-time operation state perception model and initialize the distribution network real-time operation state perception model parameters; S33: Calculate the probability that each data sample in the feature data set belongs to a given Gaussian distribution; S34: Update the parameters of the Gaussian distribution based on step S33; S35: Check whether the new Gaussian parameter is less than the set threshold or whether the gain of the likelihood function is less than the set threshold. If it is less than the set threshold, repeat steps S33 and S34 until the distribution network real-time operation status perception model converges, and stop the distribution network real-time operation status perception model training.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements a method for sensing the real-time operation status of a distribution network based on complex disasters as described in any one of claims 1 to 4 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements a method for sensing the real-time operation status of a distribution network based on complex disasters as described in any one of claims 1 to 4.

Images