A method for abnormal fault identification and emergency response in power grid based on artificial intelligence

Abstract

The present invention relates to the technical field of power grid fault identification, and discloses a power grid abnormal fault identification and emergency disposal method based on artificial intelligence, the method comprising: preliminary marking of abnormal data on pre-processed power grid data; constructing an abnormal power grid data marking correction model to correct the power grid data after the initial marking of abnormal data; performing wavelet decomposition on the corrected power grid data to obtain power grid data feature information; constructing a power grid abnormality identification model to identify abnormal power grid faults, and performing emergency disposal according to the power grid abnormal fault identification results. The present invention calculates the feedback information represented by the abnormal data based on the preliminary marking results of the abnormal data, selects corresponding correction processing actions based on the feedback information and in combination with the reinforcement learning strategy of the action space to correct the abnormal data, improves the accuracy of subsequent power grid data abnormality identification, and performs emergency disposal according to the power grid fault category identification results.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault identification, and in particular to an artificial intelligence-based power grid abnormal fault identification and emergency handling method. Background Art

[0002] With the development of social economy, the power grid has become an important infrastructure for modern urban production, life and social order maintenance. However, faults often occur during the operation of the power grid, which may lead to serious consequences such as power outages and fires, causing great inconvenience and losses to people's production and life. Rapidly and accurately identifying abnormal power grid faults and performing emergency disposal have become important issues that need to be solved urgently. In response to this problem, the present invention proposes a method for identifying and emergency handling of abnormal power grid faults based on artificial intelligence. By quickly and accurately identifying abnormal power grid faults and performing emergency disposal, the incidence of power grid accidents can be effectively reduced, and the safety and reliability of the power grid can be guaranteed. Summary of the invention

[0003] In view of this, the present invention provides an artificial intelligence-based power grid abnormal fault identification and emergency disposal method, the purpose of which is to: 1) perform local change rate abnormality marking according to the local change rate coefficient of the transmission power at different transmission times in the power grid data, and set the attenuation distance mean and quantity parameters of the adaptive attenuation to simulate density clustering processing, and perform distance abnormality marking, thereby realizing preliminary marking of abnormal data of the power grid data, and according to the preliminary marking result of the abnormal data, calculate the feedback information represented by the abnormal data, select the corresponding correction processing action based on the feedback information and the reinforcement learning strategy of the action space to correct the abnormal data, avoid the error in the abnormal identification of the power grid data due to the error in the recording of the transmission power, and improve the accuracy of the subsequent abnormal identification of the power grid data; 2) perform wavelet decomposition processing on the corrected power grid data to obtain the characteristic information of the power grid data at multiple decomposition scales, and perform multi-dimensional mapping and fusion processing on the characteristic information of the power grid data, and perform probability calculation in combination with the fusion characteristics of the power grid data to obtain the fault categories existing in different power grid equipment in the power grid, and perform emergency processing to realize accurate fault identification of the power grid equipment, and then perform emergency disposal.

[0004] To achieve the above object, the present invention provides an artificial intelligence-based power grid abnormal fault identification and emergency handling method, comprising the following steps:

[0005] S1: Collecting power grid data and preprocessing it to obtain preprocessed power grid data, and preliminarily marking abnormal data on the preprocessed power grid data, wherein an abnormal marking method based on power grid data clustering is the main implementation method of the preliminary marking of abnormal data;

[0006] S2: Constructing an abnormal power grid data annotation correction model to correct the power grid data after the abnormal data is initially annotated to obtain the corrected power grid data, wherein the abnormal power grid data annotation model takes the power grid data after the abnormal data is initially annotated as input and takes the corrected power grid data as output, wherein the feedback-based deep reinforcement learning strategy is the main implementation method of the abnormal power grid data annotation correction model;

[0007] S3: performing wavelet decomposition on the corrected power grid data to obtain power grid data characteristic information, wherein the Mallat algorithm is the main implementation method of the wavelet decomposition;

[0008] S4: constructing a power grid anomaly recognition model to identify power grid abnormal faults, and taking emergency measures according to the power grid abnormal fault recognition results, wherein the power grid anomaly recognition model takes power grid data feature information as input and takes power grid data abnormality recognition results as output.

[0009] As a further improvement method of the present invention:

[0010] Optionally, collecting and preprocessing power grid data in step S1 includes:

[0011] Collect power grid data, which is the time series data of power transmission power sent by different power grid equipment during the transmission and distribution process. The types of power grid equipment include booster stations, step-down stations and distribution stations. The collected power grid data The representation is:

[0012] X={ x n (t)=( x n ( t 1 ), x n ( t 2 ),..., x n ( t m ),..., x n ( t M ))|n ∈ [1, N]} ;

[0013] in:

[0014] represents the transmission power time series data of the nth power grid device in the power grid, t represents the time series information, and N represents the total number of power grid devices in the power grid;

[0015] Indicates the transmission power time series data The transmission power at M transmission moments, Indicates the nth power grid equipment in the power grid at the transmission time The transmission power, the time interval between adjacent transmission times is ;

[0016] The collected power grid data X is preprocessed, where the transmission power time series data in the power grid data X The preprocessing formula is:

[0017] ;

[0018] ;

[0019] in:

[0020] Indicates the transmission power time series data The preprocessing results;

[0021] Indicates the transmission power The preprocessing results are Indicates the transmission power time series data The maximum transmission power in Indicates the transmission power time series data The minimum transmission power in

[0022] The preprocessed power grid data is preliminarily labeled for abnormal data to obtain power grid data after the preliminarily labeled abnormal data.

[0023] Optionally, in step S1, preliminary marking of abnormal data on the preprocessed power grid data includes:

[0024] The abnormal data of the pre-processed power grid data are preliminarily marked, and the transmission power time series data in the pre-processed power grid data is The initial annotation process of abnormal data is as follows:

[0025] S11: Calculate transmission power time series data The local rate of change coefficient of any preprocessing result in The local rate of change coefficient is:

[0026] ;

[0027] in:

[0028] express The local rate of change coefficient of

[0029] Respectively represent the preset maximum local change rate and minimum local change rate;

[0030] S12: Transmission power time series data according to local change rate coefficient Any preprocessing result in the local change rate abnormality mark is The abnormal marking result of the local change rate is:

[0031] α n ( t m )= 1, L n ( t m )∉[0,1] 0, L n ( t m )∈[0,1] ;

[0032] in:

[0033] express The local change rate of abnormal marking results, if , then it means There are abnormal local rates of change;

[0034] S13: Calculate the transmission power time series data The distance vector between any preprocessing result and the K nearest preprocessing results in The distance vector between the K nearest preprocessing results is:

[0035] dis [ x n ' ( t m )]= [ dis 1 [ x n ' ( t m )], dis 2 [ x n ' ( t m )],..., dis k [ x n ' ( t m )],..., dis K [ x n ' ( t m )]] T

[0036] in:

[0037] dis [ x n ' ( t m )] express The distance vector between the K nearest preprocessing results;

[0038] T stands for transpose;

[0039] dis k [ x n ' ( t m )] express The distance between the preprocessing results and the kth nearest preprocessing result; in the embodiment of the present invention, the distance between the preprocessing results in the power transmission time series data is the absolute value of the difference between the preprocessing results;

[0040] S14: Constructing transmission power time series data The distance matrix is:

[0041] A n (t)=[ dis [ x n ' ( t 1 )], dis [ x n ' ( t 2 )],..., dis [ x n ' ( t m )],..., dis [ x n ' ( t M )] ]

[0042] in:

[0043] Indicates the transmission power time series data The distance matrix of

[0044] Calculate the distance matrix The mean of each row of the matrix is ​​calculated and adaptively decayed to obtain the mean decay distance of the K-row matrix, the distance matrix The mean attenuation distance of the k-th row matrix in is:

[0045] ;

[0046] μ k = 1 M ∑ m=1 M dis k [ x n ' ( t m )] ;

[0047] ;

[0048] in:

[0049] Represents the distance matrix The mean attenuation distance of the k-th row matrix in ;

[0050] represents an exponential function with a natural constant as base;

[0051] represents the adaptive attenuation coefficient;

[0052] S15: Generate the quantity parameter corresponding to each row of the matrix according to the mean attenuation distance of the K rows of the matrix, where the quantity parameter of the k-th row of the matrix is:

[0053] ;

[0054] in:

[0055] Represents the number parameter of the k-th row matrix;

[0056] Indicates the transmission power time series data Zhongyu The distance between The number of preprocessing results;

[0057] S16: Extract the mean attenuation distance and quantity parameters of each row of the matrix to form K groups of density clustering parameters:

[0058] {( μ k ' , num k )|k ∈ [1, K]} ;

[0059] in:

[0060] represents the density clustering parameter of the kth group;

[0061] Based on each group of density clustering parameters, the transmission power time series data Perform clustering processing to obtain the clustering results corresponding to each set of density clustering parameters, and calculate the Davidson-Borg index of the clustering results corresponding to each set of density clustering parameters, and select the clustering result with the smallest Davidson-Borg index as the transmission power time series data The final clustering result contains several clusters, each of which contains several transmission power time series data. In the embodiment of the present invention, the clustering processing method used is (Density-Based Spatial Clustering of Applications with Noise, density-based spatial clustering of noise applications) algorithm, the mean attenuation distance is used as the Eps parameter in the DBSCAN algorithm, and the number parameter is used as the MinPts parameter in the DBSCAN algorithm;

[0062] S17: Calculate the number of preprocessing results in each cluster in the final clustering result, and mark all preprocessing results in the cluster whose number of preprocessing results is lower than a preset threshold as having distance anomalies;

[0063] S18: Combine the marking results of steps S12 and S17, extract the preprocessing results with both distance anomaly and local change rate anomaly and mark them as abnormal data, and obtain the transmission power time series data in the preprocessed power grid data Preliminary annotation results of abnormal data In the embodiment of the present invention, the preliminary marking results of abnormal data of N power grid devices are used as the power grid data with preliminary marking of abnormal data.

[0064] Optionally, constructing an abnormal power grid data annotation correction model in step S2 includes:

[0065] Constructing an abnormal power grid data annotation correction model, wherein the abnormal power grid data annotation model takes the power grid data with preliminary annotation of abnormal data as input and takes the corrected power grid data as output, wherein the abnormal power grid data annotation model includes an input layer, a feedback layer and a correction layer, wherein the input layer is used to receive the power grid data with preliminary annotation of abnormal data, the feedback layer is used to output feedback information of the preliminary annotation result of abnormal data based on a feedback mechanism, and the correction layer is used to correct the preliminary annotation result of abnormal data based on the feedback information to obtain the corrected power grid data;

[0066] The constructed abnormal power grid data annotation correction model is optimized for model parameters, where the model parameter optimization process is as follows:

[0067] S21: Obtain the preliminary labeling results of abnormal data in the D group of power grid data and the corresponding manual correction results to form a training data set data:

[0068] data ={(x(d), y(d))|d ∈ [1, D]} ;

[0069] in:

[0070] Indicates the preliminary labeling results of the dth group of abnormal data obtained And the corresponding manual correction results ;

[0071] S22: Extract feedback information of the preliminary labeling results of abnormal data in the training data set data, and manual correction actions, where the preliminary labeling results of abnormal data The feedback information is , manual correction results The corresponding manual correction action is ;

[0072] S23: The manual correction actions corresponding to the manual correction results of group D constitute a correction processing action space ;

[0073] S24: Constructing the training optimization objective function of the abnormal power grid data annotation correction model:

[0074] ;

[0075] in:

[0076] represents the training optimization objective function for constructing the abnormal power grid data annotation and correction model, represents the model parameter to be optimized; in the embodiment of the present invention, the model parameter to be optimized in the abnormal power grid data annotation correction model is a mapping relationship between feedback information and manual correction action;

[0077] Indicates that Input to model-based parameters In the abnormal power grid data annotation correction model, the correction result of the model output;

[0078] S25: Initially generate the initial position coordinates of H sparrows, where the initial position coordinates of the hth sparrow are , set the current iteration number of the initial position coordinate to e, the initial value of e is 0, and the maximum value is E, then the e-th iteration result of the initial position coordinate of the h-th sparrow is ; The position coordinates of each sparrow correspond to a set of model parameters;

[0079] S26: performing iterative processing combined with fractional calculus on the position coordinates of each sparrow, wherein the iterative processing formula is:

[0080] ;

[0081] β h e = 1 2 exp[-0.47 dis avg ( q h (e))- dis min ( q h (e)) dis max ( q h (e))- dis min ( q h (e)) ] ;

[0082] in:

[0083] Indicates location coordinates The fractional coefficients of

[0084] Indicates location coordinates The average distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration, Indicates location coordinates The maximum distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration, Indicates location coordinates The minimum distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration;

[0085] It means that the position coordinates of H sparrows obtained in the e-th iteration are used as the input of the training optimization objective function, and the position coordinates that make the training optimization objective function value reach the minimum are selected;

[0086] S27: Let e=e+1, and return to step S26 until the maximum number of iterations is reached; and the position coordinates of the H sparrows that have reached the maximum number of iterations are used as the input of the training optimization objective function, and the position coordinates that make the training optimization objective function value reach the minimum are selected as the model parameter optimization result, and the abnormal power grid data annotation correction model is constructed;

[0087] The abnormal power grid data annotation correction model is used to correct the power grid data after the abnormal data is initially annotated.

[0088] Optionally, in step S2, the abnormal power grid data annotation correction model is used to correct the power grid data after the abnormal data is initially annotated, including:

[0089] The abnormal power grid data annotation correction model is used to correct the power grid data after the abnormal data is initially annotated. The initial annotation results of the abnormal data in the power grid data are The correction process is:

[0090] The input layer receives the preliminary annotation results of abnormal data ;

[0091] The feedback layer initially labels the abnormal data based on the feedback mechanism. Output feedback information, including preliminary annotation results of abnormal data The feedback information is:

[0092] F n = max n (t)[1+ 1 ln(M) p n ln( p n )] M ;

[0093] in:

[0094] Indicates the preliminary labeling results of abnormal data Feedback information;

[0095] Indicates the preliminary labeling results of abnormal data The longest continuous length of abnormal data in the

[0096] Indicates the preliminary labeling results of abnormal data The proportion of abnormal data in

[0097] Based on the feedback information, the correction layer takes correction actions corresponding to the feedback information to preliminarily mark the abnormal data. The abnormal data marked as abnormal in the correction process, in which the initial marking results of abnormal data The correction formula is:

[0098] ;

[0099] ;

[0100] in:

[0101] Indicates the preliminary labeling results of abnormal data The correction result will be As the corrected transmission power time series data of the nth grid device in the grid;

[0102] Indicates the preliminary labeling results of abnormal data Correction processing actions for abnormal data in Indicates that a corrective action is taken Preliminary labeling results for abnormal data The abnormal data is corrected and other data are retained to form the preliminary labeling results of abnormal data. The correction result of

[0103] Indicates that feedback information will be Mapping to the correction processing action space , the resulting corrected processing action;

[0104] Obtain the corrected transmission power time series data of N power grid devices in the power grid to form corrected power grid data:

[0105] X * ={ x n * (t)|n ∈ [1, N]} ;

[0106] in:

[0107] Indicates the corrected power grid data.

[0108] Optionally, in step S3, performing wavelet decomposition on the corrected power grid data includes:

[0109] Corrected grid data Perform wavelet decomposition, where the wavelet decomposition process is:

[0110] S31: Building a Low-Pass Filter and high pass filter ;

[0111] S32: Set the corrected grid data The current wavelet decomposition scale of any corrected transmission power time series data is j, the initial value of j is 0, and the maximum value is J. Then the corrected transmission power time series data The decomposition result at wavelet decomposition scale j is ,in ;

[0112] S33: Decomposition of any corrected transmission power time series data by wavelet decomposition scale j+1, wherein the corrected transmission power time series data The decomposition formula is:

[0113] ;

[0114] S34: let j=j+1, return to step S33, until the maximum value of the wavelet decomposition scale is reached, and the decomposition results of the transmission power time series data after arbitrary correction at J wavelet decomposition scales are obtained;

[0115] S35: Decomposition results of the corrected transmission power time series data of N power grid devices at J wavelet decomposition scales are used as power grid data feature information.

[0116] Optionally, in step S4, a power grid abnormality identification model is constructed to identify abnormal power grid faults, and emergency treatment is performed according to the abnormal power grid fault identification result, including:

[0117] Construct a power grid anomaly recognition model to identify power grid abnormal faults, wherein the power grid anomaly recognition model takes power grid data feature information as input and takes power grid data anomaly recognition results as output, wherein the power grid anomaly recognition model includes an input layer, a feature information mapping layer, a fusion layer and an output layer, wherein the input layer is used to receive power grid data feature information, the feature information mapping layer is used to map the power grid data feature information, the fusion layer is used to fuse the mapping results of multi-dimensional power grid data feature information to obtain power grid data fusion features, and the output layer performs probability calculation on the power grid data fusion features to obtain the fault categories existing in different power grid equipment in the power grid as power grid anomaly recognition results, and performs emergency disposal on the corresponding power grid equipment according to the power grid abnormal fault recognition results;

[0118] The grid abnormal fault identification process based on the grid abnormality identification model is as follows:

[0119] S41: The input layer receives the grid data characteristic information;

[0120] S42: The characteristic information mapping layer performs mapping processing on the characteristic information of the power grid data, wherein the decomposition result mapping processing formula of the J wavelet decomposition scales corresponding to the nth power grid equipment in the characteristic information of the power grid data is:

[0121] ;

[0122] ;

[0123] in:

[0124] Indicates the mapping processing result of the nth power grid device in the power grid data characteristic information;

[0125] Decomposition results The mapping vector of

[0126] represents the mapping parameter matrix;

[0127] represents an activation function; in the embodiment of the present invention, the selected activation function is a ReLU function;

[0128] S43: The fusion layer fuses the mapping results of the multi-dimensional power grid data feature information to obtain the power grid data fusion feature of each power grid device, wherein the power grid data fusion feature of the nth power grid device is:

[0129] ;

[0130] in:

[0131] Represents the grid data fusion characteristics of the nth grid device;

[0132] S44: The output layer performs probability calculation on the fusion features of the power grid data to obtain the fault categories of different power grid devices in the power grid, where the probability calculation result of the fault category of the nth power grid device is:

[0133] ;

[0134] in:

[0135] represents the probability that the fault category of the nth power grid equipment is the sth fault category, s ∈ [1,S] , S represents the total number of fault categories of power grid equipment;

[0136] represents the parameter matrix corresponding to the sth fault category;

[0137] The fault category with the highest probability is taken as the fault category of the nth power grid equipment.

[0138] In order to solve the above problem, the present invention provides an electronic device, the electronic device comprising:

[0139] A memory storing at least one instruction;

[0140] Communication interface, enabling electronic equipment to communicate; and

[0141] The processor executes the instructions stored in the memory to implement the above-mentioned artificial intelligence-based power grid abnormal fault identification and emergency handling method.

[0142] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned artificial intelligence-based power grid abnormal fault identification and emergency response method.

[0143] Compared with the prior art, the present invention proposes an artificial intelligence-based power grid abnormal fault identification and emergency handling method, which has the following advantages:

[0144] First, this scheme proposes an abnormal data annotation method to calculate the transmission power time series data The local rate of change coefficient of any preprocessing result in The local rate of change coefficient is:

[0145] ;

[0146] in: express The local rate of change coefficient of Respectively represent the preset maximum local change rate and minimum local change rate; according to the local change rate coefficient, the transmission power time series data Any preprocessing result in the local change rate abnormality mark is The abnormal marking result of the local change rate is:

[0147] α n ( t m )= 1, L n ( t m )∉[0,1] 0, L n ( t m )∈[0,1] ;

[0148] in: express The local change rate of abnormal marking results, if , then it means There is an abnormal local change rate; the transmission power time series data is calculated The distance vector between any preprocessing result and the K nearest preprocessing results in The distance vector between the K nearest preprocessing results is:

[0149] dis [ x n ' ( t m )]= [ dis 1 [ x n ' ( t m )], dis 2 [ x n ' ( t m )],..., dis k [ x n ' ( t m )],..., dis K [ x n ' ( t m )]] T ;

[0150] in: dis [ x n ' ( t m )] express The distance vector between the K nearest preprocessing results; T represents transpose; dis k [ x n ' ( t m )] express The distance between the kth nearest preprocessing result and the transmission power time series data The distance matrix is:

[0151] A n (t)=[ dis [ x n ' ( t 1 )], dis [ x n ' ( t 2 )],..., dis [ x n ' ( t m )],..., dis [ x n ' ( t M )] ] ;

[0152] in: Indicates the transmission power time series data The distance matrix of ; calculate the distance matrix The mean of each row of the matrix is ​​calculated and adaptively decayed to obtain the mean decay distance of the K-row matrix, the distance matrix The mean attenuation distance of the k-th row matrix in is:

[0153] ;

[0154] μ k = 1 M ∑ m=1 M dis k [ x n ' ( t m )] ;

[0155] ;

[0156] in: Represents the distance matrix The mean attenuation distance of the k-th row matrix in ; represents an exponential function with a natural constant as base; Represents the adaptive attenuation coefficient; according to the mean attenuation distance of the K-row matrix, the quantity parameter corresponding to each row of the matrix is ​​generated, where the quantity parameter of the k-th row matrix is:

[0157] ;

[0158] in: Represents the number parameter of the k-th row matrix; Indicates the transmission power time series data Zhongyu The distance between The number of preprocessing results; extract the mean attenuation distance and quantity parameters of each row matrix to form K groups of density clustering parameters:

[0159] {( μ k ' , num k )|k ∈ [1, K]} ;

[0160] in: Represents the kth group of density clustering parameters; based on each group of density clustering parameters, the transmission power time series data Perform clustering processing to obtain the clustering results corresponding to each set of density clustering parameters, and calculate the Davidson-Borg index of the clustering results corresponding to each set of density clustering parameters, and select the clustering result with the smallest Davidson-Borg index as the transmission power time series data The final clustering result contains several clusters, each of which contains several transmission power time series data. Calculate the number of preprocessing results in each cluster in the final clustering result, and mark all preprocessing results in the cluster whose number of preprocessing results is lower than the preset threshold as having distance anomalies; extract the preprocessing results with both distance anomalies and local change rate anomalies and mark them as abnormal data, and obtain the transmission power time series data in the preprocessed power grid data. Preliminary annotation results of abnormal data This scheme marks the local change rate anomaly according to the local change rate coefficient of the power transmission power at different transmission times in the power grid data, sets the attenuation distance mean and quantity parameters of the adaptive attenuation to simulate density clustering processing, and marks the distance anomaly, thereby realizing the preliminary marking of abnormal data in the power grid data.

[0161] At the same time, this scheme proposes an abnormal data correction strategy and a power grid abnormal fault identification method, which uses the abnormal power grid data annotation correction model to correct the power grid data after the abnormal data is initially annotated. The correction process is as follows: the input layer receives the preliminary annotation results of abnormal data ; The feedback layer preliminarily labels the abnormal data based on the feedback mechanism Output feedback information, including preliminary annotation results of abnormal data The feedback information is:

[0162] F n = max n (t)[1+ 1 ln(M) p n ln( p n )] M ;

[0163] in: Indicates the preliminary labeling results of abnormal data Feedback information; Indicates the preliminary labeling results of abnormal data The longest continuous length of abnormal data in the Indicates the preliminary labeling results of abnormal data The correction layer takes correction actions corresponding to the feedback information based on the feedback information to preliminarily mark the abnormal data. The abnormal data marked as abnormal in the correction process, in which the initial marking results of abnormal data The correction formula is:

[0164] ;

[0165] ;

[0166] in: Indicates the preliminary labeling results of abnormal data The correction result will be As the corrected transmission power time series data of the nth grid device in the grid; Indicates the preliminary labeling results of abnormal data Correction processing actions for abnormal data in Indicates that a corrective action is taken Preliminary labeling results for abnormal data The abnormal data is corrected and other data are retained to form the preliminary labeling results of abnormal data. The correction result of Indicates that feedback information will be Mapping to the correction processing action space , the obtained correction processing action; obtain the corrected transmission power time series data of N power grid devices in the power grid to form the corrected power grid data:

[0167] X * ={ x n * (t)|n ∈ [1, N]} ;

[0168] in: Represents the corrected power grid data. This solution calculates the feedback information represented by the abnormal data based on the preliminary labeling results of the abnormal data, selects the corresponding correction processing action based on the feedback information and combined with the reinforcement learning strategy of the action space to correct the abnormal data, so as to avoid errors in the recognition of abnormal power grid data due to errors in the recording of transmission power, and improve the accuracy of subsequent abnormal power grid data recognition; the corrected power grid data is processed by wavelet decomposition to obtain the characteristic information of power grid data at multiple decomposition scales, and the characteristic information of power grid data is multi-dimensionally mapped and fused, and probability calculation is performed in combination with the fusion characteristics of power grid data to obtain the fault categories of different power grid equipment in the power grid, and emergency processing is performed to achieve accurate fault identification of power grid equipment, and then emergency disposal is performed. BRIEF DESCRIPTION OF THE DRAWINGS

[0169] Figure 1 A flowchart of an artificial intelligence-based power grid abnormal fault identification and emergency handling method provided by an embodiment of the present invention;

[0170] Figure 2 A schematic diagram of the structure of an electronic device for implementing an artificial intelligence-based power grid abnormal fault identification and emergency handling method provided by one embodiment of the present invention.

[0171] In the figure: 1 electronic device, 10 processor, 11 memory, 12 program, 13 communication interface.

[0172] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0173] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0174] The embodiment of the present application provides a method for identifying abnormal faults and emergency handling of power grids based on artificial intelligence. The execution subject of the method for identifying abnormal faults and emergency handling of power grids based on artificial intelligence includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for identifying abnormal faults and emergency handling of power grids based on artificial intelligence can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0175] Embodiment 1:

[0176] S1: Collect and preprocess the power grid data to obtain the preprocessed power grid data, and preliminarily mark the abnormal data on the preprocessed power grid data.

[0177] The step S1 collects power grid data and performs preprocessing, including:

[0178] Collect power grid data, which is the time series data of power transmission power sent by different power grid equipment during the transmission and distribution process. The types of power grid equipment include booster stations, step-down stations and distribution stations. The collected power grid data The representation is:

[0179] X={ x n (t)=( x n ( t 1 ), x n ( t 2 ),..., x n ( t m ),..., x n ( t M ))|n ∈ [1, N]} ;

[0180] in:

[0181] represents the transmission power time series data of the nth power grid device in the power grid, t represents the time series information, and N represents the total number of power grid devices in the power grid;

[0182] Indicates the transmission power time series data The transmission power at M transmission moments, Indicates the nth power grid equipment in the power grid at the transmission time The transmission power, the time interval between adjacent transmission times is ;

[0183] The collected power grid data X is preprocessed, where the power transmission power time series data in the power grid data X The preprocessing formula is:

[0184] ;

[0185] ;

[0186] in:

[0187] Indicates the transmission power time series data The preprocessing results;

[0188] Indicates the transmission power The preprocessing results are Indicates the transmission power time series data The maximum transmission power in Indicates the transmission power time series data The minimum transmission power in

[0189] The preprocessed power grid data is preliminarily labeled for abnormal data to obtain power grid data after the preliminarily labeled abnormal data.

[0190] In step S1, the preprocessed power grid data is preliminarily marked with abnormal data, including:

[0191] The abnormal data of the pre-processed power grid data are preliminarily marked, and the transmission power time series data in the pre-processed power grid data is The initial annotation process of abnormal data is as follows:

[0192] S11: Calculate transmission power time series data The local rate of change coefficient of any preprocessing result in The local rate of change coefficient is:

[0193] ;

[0194] in:

[0195] express The local rate of change coefficient of

[0196] Respectively represent the preset maximum local change rate and minimum local change rate;

[0197] S12: Transmission power time series data according to local change rate coefficient Any preprocessing result in the local change rate abnormality mark is The abnormal marking result of the local change rate is:

[0198] α n ( t m )= 1, L n ( t m )∉[0,1] 0, L n ( t m )∈[0,1] ;

[0199] in:

[0200] express The local change rate of abnormal marking results, if , then it means There are abnormal local rates of change;

[0201] S13: Calculate the transmission power time series data The distance vector between any preprocessing result and the K nearest preprocessing results in The distance vector between the K nearest preprocessing results is:

[0202] dis [ x n ' ( t m )]= [ dis 1 [ x n ' ( t m )], dis 2 [ x n ' ( t m )],..., dis k [ x n ' ( t m )],..., dis K [ x n ' ( t m )]] T ;

[0203] in:

[0204] dis [ x n ' ( t m )] express The distance vector between the K nearest preprocessing results;

[0205] T stands for transpose;

[0206] dis k [ x n ' ( t m )] express The distance to the kth nearest preprocessing result;

[0207] S14: Constructing transmission power time series data The distance matrix is:

[0208] A n (t)=[ dis [ x n ' ( t 1 )], dis [ x n ' ( t 2 )],..., dis [ x n ' ( t m )],..., dis [ x n ' ( t M )] ] ;

[0209] in:

[0210] Indicates the transmission power time series data The distance matrix of

[0211] Calculate the distance matrix The mean of each row of the matrix is ​​calculated and adaptively decayed to obtain the mean decay distance of the K-row matrix, the distance matrix The mean attenuation distance of the k-th row matrix in is:

[0212] ;

[0213] μ k = 1 M ∑ m=1 M dis k [ x n ' ( t m )] ;

[0214] ;

[0215] in:

[0216] Represents the distance matrix The mean attenuation distance of the k-th row matrix in ;

[0217] represents an exponential function with a natural constant as base;

[0218] represents the adaptive attenuation coefficient;

[0219] S15: Generate the quantity parameter corresponding to each row of the matrix according to the mean attenuation distance of the K rows of the matrix, where the quantity parameter of the k-th row of the matrix is:

[0220] ;

[0221] in:

[0222] Represents the number parameter of the k-th row matrix;

[0223] Indicates the transmission power time series data Zhongyu The distance between The number of preprocessing results;

[0224] S16: Extract the mean attenuation distance and quantity parameters of each row of the matrix to form K groups of density clustering parameters:

[0225] {( μ k ' , num k )|k ∈ [1, K]} ;

[0226] in:

[0227] represents the density clustering parameter of the kth group;

[0228] Based on each group of density clustering parameters, the transmission power time series data Perform clustering processing to obtain the clustering results corresponding to each set of density clustering parameters, and calculate the Davidson-Borg index of the clustering results corresponding to each set of density clustering parameters, and select the clustering result with the smallest Davidson-Borg index as the transmission power time series data The final clustering result contains several clusters, each of which contains several transmission power time series data. The preprocessing results in ;

[0229] S17: Calculate the number of preprocessing results in each cluster in the final clustering result, and mark all preprocessing results in the cluster whose number of preprocessing results is lower than a preset threshold as having distance anomalies;

[0230] S18: Combine the marking results of steps S12 and S17, extract the preprocessing results with both distance anomaly and local change rate anomaly and mark them as abnormal data, and obtain the transmission power time series data in the preprocessed power grid data Preliminary annotation results of abnormal data .

[0231] S2: Construct an abnormal power grid data annotation correction model to correct the power grid data after the abnormal data is initially annotated to obtain the corrected power grid data. The abnormal power grid data annotation model takes the power grid data after the abnormal data is initially annotated as input and takes the corrected power grid data as output.

[0232] The abnormal power grid data annotation correction model is constructed in step S2, including:

[0233] Constructing an abnormal power grid data annotation correction model, wherein the abnormal power grid data annotation model takes the power grid data with preliminary annotation of abnormal data as input and takes the corrected power grid data as output, wherein the abnormal power grid data annotation model includes an input layer, a feedback layer and a correction layer, wherein the input layer is used to receive the power grid data with preliminary annotation of abnormal data, the feedback layer is used to output feedback information of the preliminary annotation result of abnormal data based on a feedback mechanism, and the correction layer is used to correct the preliminary annotation result of abnormal data based on the feedback information to obtain the corrected power grid data;

[0234] The constructed abnormal power grid data annotation correction model is optimized for model parameters, where the model parameter optimization process is as follows:

[0235] S21: Obtain the preliminary labeling results of abnormal data in the D group of power grid data and the corresponding manual correction results to form a training data set data:

[0236] data ={(x(d), y(d))|d ∈ [1, D]} ;

[0237] in:

[0238] Indicates the preliminary labeling results of the dth group of abnormal data obtained And the corresponding manual correction results ;

[0239] S22: Extract feedback information of the preliminary labeling results of abnormal data in the training data set data, and manual correction actions, where the preliminary labeling results of abnormal data The feedback information is , manual correction results The corresponding manual correction action is ;

[0240] S23: The manual correction actions corresponding to the manual correction results of group D constitute a correction processing action space ;

[0241] S24: Constructing the training optimization objective function of the abnormal power grid data annotation correction model:

[0242] ;

[0243] in:

[0244] represents the training optimization objective function for constructing the abnormal power grid data annotation and correction model, represents the model parameter to be optimized; in the embodiment of the present invention, the model parameter to be optimized in the abnormal power grid data annotation correction model is a mapping relationship between feedback information and manual correction action;

[0245] Indicates that Input to model-based parameters In the abnormal power grid data annotation correction model, the correction result of the model output;

[0246] S25: Initially generate the initial position coordinates of H sparrows, where the initial position coordinates of the hth sparrow are , set the current iteration number of the initial position coordinate to e, the initial value of e is 0, and the maximum value is E, then the e-th iteration result of the initial position coordinate of the h-th sparrow is ; The position coordinates of each sparrow correspond to a set of model parameters;

[0247] S26: performing iterative processing combined with fractional calculus on the position coordinates of each sparrow, wherein the iterative processing formula is:

[0248] ;

[0249] β h e = 1 2 exp[-0.47 dis avg ( q h (e))- dis min ( q h (e)) dis max ( q h (e))- dis min ( q h (e)) ] ;

[0250] in:

[0251] Indicates location coordinates The fractional coefficients of

[0252] Indicates location coordinates The average distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration, Indicates location coordinates The maximum distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration, Indicates location coordinates The minimum distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration;

[0253] It means that the position coordinates of H sparrows obtained in the e-th iteration are used as the input of the training optimization objective function, and the position coordinates that make the training optimization objective function value reach the minimum are selected;

[0254] S27: Let e=e+1, and return to step S26 until the maximum number of iterations is reached; and the position coordinates of the H sparrows that have reached the maximum number of iterations are used as the input of the training optimization objective function, and the position coordinates that make the training optimization objective function value reach the minimum are selected as the model parameter optimization result, and the abnormal power grid data annotation correction model is constructed;

[0255] The abnormal power grid data annotation correction model is used to correct the power grid data after the abnormal data is initially annotated.

[0256] In the step S2, the abnormal power grid data annotation correction model is used to correct the power grid data after the abnormal data is initially annotated, including:

[0257] The abnormal power grid data annotation correction model is used to correct the power grid data after the abnormal data is initially annotated. The initial annotation results of the abnormal data in the power grid data are The correction process is:

[0258] The input layer receives the preliminary annotation results of abnormal data ;

[0259] The feedback layer initially labels the abnormal data based on the feedback mechanism. Output feedback information, including preliminary annotation results of abnormal data The feedback information is:

[0260] F n = max n (t)[1+ 1 ln(M) p n ln( p n )] M ;

[0261] in:

[0262] Indicates the preliminary labeling results of abnormal data Feedback information;

[0263] Indicates the preliminary labeling results of abnormal data The longest continuous length of abnormal data in the

[0264] Indicates the preliminary labeling results of abnormal data The proportion of abnormal data in

[0265] Based on the feedback information, the correction layer takes correction actions corresponding to the feedback information to preliminarily mark the abnormal data. The abnormal data marked as abnormal in the correction process, in which the initial marking results of abnormal data The correction formula is:

[0266] ;

[0267] ;

[0268] in:

[0269] Indicates the preliminary labeling results of abnormal data The correction result will be As the corrected transmission power time series data of the nth grid device in the grid;

[0270] Indicates the preliminary labeling results of abnormal data Correction processing actions for abnormal data in Indicates that a corrective action is taken Preliminary labeling results for abnormal data The abnormal data is corrected and other data are retained to form the preliminary labeling results of abnormal data. The correction result of

[0271] Indicates that feedback information will be Mapping to the correction processing action space , the resulting corrected processing action;

[0272] Obtain the corrected transmission power time series data of N power grid devices in the power grid to form corrected power grid data:

[0273] X * ={ x n * (t)|n ∈ [1, N]} ;

[0274] in:

[0275] Indicates the corrected power grid data.

[0276] S3: Perform wavelet decomposition on the corrected power grid data to obtain power grid data characteristic information.

[0277] In the step S3, the corrected power grid data is subjected to wavelet decomposition, including:

[0278] Corrected grid data Perform wavelet decomposition, where the wavelet decomposition process is:

[0279] S31: Building a Low-Pass Filter and high pass filter ;

[0280] S32: Set the corrected grid data The current wavelet decomposition scale of any corrected transmission power time series data is j, the initial value of j is 0, and the maximum value is J. Then the corrected transmission power time series data The decomposition result at wavelet decomposition scale j is ,in ;

[0281] S33: Decomposition of any corrected transmission power time series data by wavelet decomposition scale j+1, wherein the corrected transmission power time series data The decomposition formula is:

[0282] ;

[0283] S34: let j=j+1, return to step S33, until the maximum value of the wavelet decomposition scale is reached, and the decomposition results of the transmission power time series data after arbitrary correction at J wavelet decomposition scales are obtained;

[0284] S35: Decomposition results of the corrected transmission power time series data of N power grid devices at J wavelet decomposition scales are used as power grid data feature information.

[0285] S4: constructing a power grid anomaly recognition model to identify power grid abnormal faults, and taking emergency measures according to the power grid abnormal fault recognition results, wherein the power grid anomaly recognition model takes power grid data feature information as input and takes power grid data abnormality recognition results as output.

[0286] In the step S4, a power grid abnormality identification model is constructed to identify abnormal power grid faults, and emergency treatment is performed according to the abnormal power grid fault identification results, including:

[0287] Construct a power grid anomaly recognition model to identify power grid abnormal faults, wherein the power grid anomaly recognition model takes power grid data feature information as input and takes power grid data anomaly recognition results as output, wherein the power grid anomaly recognition model includes an input layer, a feature information mapping layer, a fusion layer and an output layer, wherein the input layer is used to receive power grid data feature information, the feature information mapping layer is used to map the power grid data feature information, the fusion layer is used to fuse the mapping results of multi-dimensional power grid data feature information to obtain power grid data fusion features, and the output layer performs probability calculation on the power grid data fusion features to obtain the fault categories existing in different power grid equipment in the power grid as power grid anomaly recognition results, and performs emergency disposal on the corresponding power grid equipment according to the power grid abnormal fault recognition results;

[0288] The grid abnormal fault identification process based on the grid abnormality identification model is as follows:

[0289] S41: The input layer receives the grid data characteristic information;

[0290] S42: The characteristic information mapping layer performs mapping processing on the characteristic information of the power grid data, wherein the decomposition result mapping processing formula of the J wavelet decomposition scales corresponding to the nth power grid equipment in the characteristic information of the power grid data is:

[0291] ;

[0292] ;

[0293] in:

[0294] Indicates the mapping processing result of the nth power grid device in the power grid data characteristic information;

[0295] Decomposition results The mapping vector of

[0296] represents the mapping parameter matrix;

[0297] represents an activation function; in the embodiment of the present invention, the selected activation function is a ReLU function;

[0298] S43: The fusion layer fuses the mapping results of the multi-dimensional power grid data feature information to obtain the power grid data fusion feature of each power grid device, wherein the power grid data fusion feature of the nth power grid device is:

[0299] ;

[0300] in:

[0301] Represents the grid data fusion characteristics of the nth grid device;

[0302] S44: The output layer performs probability calculation on the fusion features of the power grid data to obtain the fault categories of different power grid devices in the power grid, where the probability calculation result of the fault category of the nth power grid device is:

[0303] ;

[0304] in:

[0305] represents the probability that the fault category of the nth power grid equipment is the sth fault category, s ∈ [1,S] , S represents the total number of fault categories of power grid equipment;

[0306] represents the parameter matrix corresponding to the sth fault category;

[0307] The fault category with the highest probability is taken as the fault category of the nth power grid equipment.

[0308] Embodiment 2:

[0309] like Figure 2 , is a schematic diagram of the structure of an electronic device for implementing an artificial intelligence-based power grid abnormal fault identification and emergency handling method provided by an embodiment of the present invention.

[0310] The electronic device 1 may include a processor 10 , a memory 11 , a communication interface 13 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10 , such as a program 12 .

[0311] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 may not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the program 12, etc., but also be used to temporarily store data that has been output or is to be output.

[0312] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (such as programs 12 for realizing abnormal fault identification and emergency handling of power grids based on artificial intelligence), and calls data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0313] The communication interface 13 may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices, and to achieve connection and communication between internal components of the electronic device.

[0314] The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0315] Figure 2 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 2 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0316] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to various components. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0317] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0318] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0319] The program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0320] Collect and preprocess the power grid data to obtain preprocessed power grid data, and preliminarily mark abnormal data on the preprocessed power grid data;

[0321] Constructing an abnormal power grid data annotation correction model to correct the power grid data after the abnormal data is initially annotated, and obtaining the corrected power grid data;

[0322] Perform wavelet decomposition on the corrected power grid data to obtain power grid data characteristic information;

[0323] Construct a power grid anomaly recognition model to identify power grid abnormal faults, and take emergency measures based on the power grid abnormal fault recognition results.

[0324] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figures 1 to 2 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0325] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0326] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0327] The above are only preferred 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 identifying abnormal faults in power grid and handling them in an emergency based on artificial intelligence, characterized in that: The method comprises: S1: Collect and preprocess the power grid data to obtain the preprocessed power grid data, and preliminarily mark the abnormal data on the preprocessed power grid data; S2: constructing an abnormal power grid data annotation correction model to correct the power grid data after the abnormal data is initially annotated to obtain the corrected power grid data, wherein the abnormal power grid data annotation correction model takes the power grid data after the abnormal data is initially annotated as input and takes the corrected power grid data as output; The abnormal power grid data annotation correction model is used to correct the power grid data after the abnormal data is initially annotated, where the initial annotation result of the abnormal data in the power grid data is x″ n The amendment process for (t) is: The input layer receives the initial annotation result x″ of abnormal data n (t); The feedback layer preliminarily labels the abnormal data based on the feedback mechanism x″ n (t) Output feedback information, including the initial labeling result x″ of abnormal data n The feedback information of (t) is: in: F n Represents the preliminary labeling result of abnormal data x″ n (t) feedback information; max n (t) represents the preliminary labeling result of abnormal data x″ n (t) The longest continuous length of abnormal data; p n Represents the preliminary labeling result of abnormal data x″ n The proportion of abnormal data in (t); Based on the feedback information, the correction layer takes the correction processing action corresponding to the feedback information to preliminarily mark the abnormal data. n The abnormal data marked as abnormal in (t) are corrected, where the initial labeling result x″ of the abnormal data is n (t) The correction formula is: π n ＝π * (F n )； in: Indicates the preliminary labeling result of abnormal data x″ n The result of the correction process of (t) is As the corrected transmission power time series data of the nth grid device in the grid; π n Represents the preliminary labeling result x″ of abnormal data n (t) The correction processing action of abnormal data, π n (x″ n (t)) represents the correction processing action π n Preliminary labeling results of abnormal data x″ n The abnormal data in (t) are corrected, and other data are retained to form the preliminary labeling result x″ of the abnormal data n (t) The correction result; π * (F n ) indicates that the feedback information F n Mapping to the correction processing action space π * , the resulting corrected processing action; Obtain the corrected transmission power time series data of N power grid devices in the power grid to form corrected power grid data: in: X * Indicates the corrected power grid data; S3: performing wavelet decomposition on the corrected power grid data to obtain characteristic information of the power grid data; S4: constructing a power grid anomaly recognition model to identify power grid abnormal faults, and taking emergency measures according to the power grid abnormal fault recognition results, wherein the power grid anomaly recognition model takes power grid data feature information as input and takes power grid data abnormality recognition results as output.

2. The method for identifying abnormal faults and emergency handling of power grids based on artificial intelligence according to claim 1, characterized in that: The step S1 collects power grid data and performs preprocessing, including: Collect power grid data, where the power grid data is the time series data of power transmission power sent by different power grid equipment during the transmission and distribution process. The types of power grid equipment include step-up stations, step-down stations and distribution stations. The representation of the collected power grid data X is as follows: X={x n (t)=(x n (t1),x n (t2),...,x n (t m ),...,x n (t M ))|n∈[1,N]}; in: x n (t) represents the transmission power time series data of the nth grid device in the grid, t represents the time series information, and N represents the total number of grid devices in the grid; (x n (t1),x n (t2),...,x n (t m ),...,x n (t M )) represents the transmission power time series data x n (t) is the transmission power at M transmission times, x n (t m ) represents the nth power grid equipment in the power grid at the transmission time t m The transmission power of the power transmission system is Δt, and the time interval between adjacent transmission times is Δt; The collected power grid data X is preprocessed, wherein the power transmission power time series data x in the power grid data X is n The preprocessing formula of (t) is: x′ n (t)=(x′ n (t1),x′ n (t2),...,x′ n (t m ),...,x′ n (t M )); in: x' n (t) represents the transmission power time series data x n (t) preprocessing results; x' n (t m ) represents the transmission power x n (t m ) is the preprocessing result, max{x n (t)} represents the transmission power time series data x n (t), min{x n (t)} represents the transmission power time series data x n (t) the minimum transmission power; The preprocessed power grid data is preliminarily labeled for abnormal data to obtain power grid data after the preliminarily labeled abnormal data.

3. The method for identifying abnormal faults and handling emergencies in power grids based on artificial intelligence according to claim 2, characterized in that: In step S1, the pre-processed power grid data is preliminarily marked with abnormal data, including: The abnormal data of the pre-processed power grid data are preliminarily marked, among which the transmission power time series data x' in the pre-processed power grid data n The initial annotation process of abnormal data (t) is as follows: S11: Calculate the transmission power time series data x' n (t) is the local rate of change coefficient of any preprocessing result, where x' n (t m ) is: in: L n (t m ) represents x' n (t m )’s local rate of change coefficient; L max ,L min Respectively represent the preset maximum local change rate and minimum local change rate; S12: Transmission power time series data x' according to the local change rate coefficient n Any preprocessing result in (t) is marked as abnormal in local change rate, where x' n (t m )’s local change rate abnormality marking result is: in: α n (t m ) represents x' n (t m )’s local change rate abnormal marking result, if α n (t m )=1, then x' n (t m ) There is an abnormal local rate of change; S13: Calculate the transmission power time series data x' n The distance vector between any preprocessing result in (t) and the K nearest preprocessing results, where x' n (t m ) and the distance vector between the K nearest preprocessing results is: dis[x' n (t m )]=[dis1[x' n (t m )],dis2[x' n (t m )],...,dis k [x' n (t m )],...,dis K [x' n (t m )]] T ; in: dis[x' n (t m )] represents x' n (t m ) and the distance vector between the K nearest preprocessing results; T stands for transpose; dis k [x' n (t m )] represents x′ n (t m ) and the distance between the kth nearest preprocessing result; S14: Constructing transmission power time series data x′ n The distance matrix of (t): A n (t)=[dis[x′ n (t1)],dis[x′ n (t2)],...,dis[x′ n (t m )],...,dis[x′ n (t M )]]; in: A n (t) represents the transmission power time series data x' n (t) distance matrix; Calculate the distance matrix A n The mean of each row of the matrix in (t) is adaptively attenuated to obtain the mean attenuation distance of the K-row matrix; S15: Generate a quantity parameter corresponding to each row of the matrix according to the mean attenuation distance of the K rows of the matrix; S16: extracting the attenuation distance mean and quantity parameter of each row matrix to form K groups of density clustering parameters; Based on each group of density clustering parameters, the transmission power time series data x' n (t) Perform clustering processing to obtain the clustering results corresponding to each set of density clustering parameters, and calculate the Davidson-Borg index of the clustering results corresponding to each set of density clustering parameters, and select the clustering result with the smallest Davidson-Borg index as the transmission power time series data x' n (t), where the final clustering result contains several clusters, each of which contains several transmission power time series data x' n (t) Preprocessing results; S17: Calculate the number of preprocessing results in each cluster in the final clustering result, and mark all preprocessing results in the cluster whose number of preprocessing results is lower than a preset threshold as having distance anomalies; S18: Combine the marking results of steps S12 and S17, extract the preprocessing results with both distance anomaly and local change rate anomaly and mark them as abnormal data, and obtain the transmission power time series data x' in the preprocessed power grid data n (t) The initial labeling result of abnormal data x' n '(t).

4. The method for identifying abnormal faults and handling emergencies in power grids based on artificial intelligence according to claim 1, characterized in that: The abnormal power grid data annotation correction model is constructed in step S2, including: Constructing an abnormal power grid data annotation correction model, wherein the abnormal power grid data annotation correction model takes the power grid data with preliminary annotation of abnormal data as input and takes the corrected power grid data as output, wherein the abnormal power grid data annotation model includes an input layer, a feedback layer and a correction layer, wherein the input layer is used to receive the power grid data with preliminary annotation of abnormal data, the feedback layer is used to output feedback information of the preliminary annotation result of abnormal data based on a feedback mechanism, and the correction layer is used to correct the preliminary annotation result of abnormal data based on the feedback information to obtain the corrected power grid data; The constructed abnormal power grid data annotation correction model is optimized for model parameters, where the model parameter optimization process is as follows: S21: Obtain the preliminary labeling results of abnormal data in the D group of power grid data and the corresponding manual correction results to form a training data set data: data={(x(d),y(d))|d∈[1,D]}; in: (x(d), y(d)) represents the initial labeling result x(d) of the dth group of abnormal data and the corresponding manual correction result y(d); S22: extracting feedback information of the preliminary labeling result of abnormal data in the training data set data and manual correction actions, wherein the feedback information of the preliminary labeling result of abnormal data x(d) is F(d), and the manual correction action corresponding to the manual correction result y(d) is π(d); S23: The manual correction actions corresponding to the manual correction results of group D constitute a correction processing action space π * ; S24: Constructing the training optimization objective function of the abnormal power grid data annotation correction model: in: Loss(θ) represents the training optimization objective function for constructing the abnormal power grid data annotation and correction model, and θ represents the model parameters to be optimized; G θ (π * ,x(d)) represents the correction result output by the model when x(d) is input into the abnormal power grid data annotation correction model based on the model parameter θ; ||·|| represents the L1 norm; S25: Initially generate the initial position coordinates of H sparrows, where the initial position coordinates of the hth sparrow are q h (0), set the current iteration number of the initial position coordinate to e, the initial value of e is 0, the maximum value is E, then the initial position coordinate of the hth sparrow in the eth iteration result is q h (e); where the position coordinates of each sparrow correspond to a set of model parameters; S26: performing iterative processing combined with fractional calculus on the position coordinates of each sparrow, wherein the iterative processing formula is: in: Represents the position coordinate q h (e) fractional coefficients; dis avg (q h (e)) represents the position coordinate q h (e) The average distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration, dis max (q h (e)) represents the position coordinate q h (e) The maximum distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration, dis min (q h (e)) represents the position coordinate q h (e) The minimum distance to the position coordinates of the remaining H-1 sparrows obtained in the e-th iteration; q best (e) represents taking the position coordinates of H sparrows obtained in the e-th iteration as the input of the training optimization objective function, and selecting the position coordinates that minimize the value of the training optimization objective function; S27: Let e=e+1, and return to step S26 until the maximum number of iterations is reached; and the position coordinates of the H sparrows that have reached the maximum number of iterations are used as the input of the training optimization objective function, and the position coordinates that make the training optimization objective function value reach the minimum are selected as the model parameter optimization result, and the abnormal power grid data annotation correction model is constructed; The abnormal power grid data annotation correction model is used to correct the power grid data after the abnormal data is initially annotated.

5. The method for identifying abnormal faults and handling emergencies in power grids based on artificial intelligence according to claim 1, characterized in that: In the step S3, the corrected power grid data is subjected to wavelet decomposition, including: The corrected power grid data X * Perform wavelet decomposition, where the wavelet decomposition process is: S31: construct a low-pass filter G1(·) and a high-pass filter G2(·); S32: Set the corrected grid data X * The current wavelet decomposition scale of any corrected transmission power time series data is j, the initial value of j is 0, and the maximum value is J. Then the corrected transmission power time series data The decomposition result at wavelet decomposition scale j is in S33: Decomposition of any corrected transmission power time series data by wavelet decomposition scale j+1, wherein the corrected transmission power time series data The decomposition formula is: S34: let j=j+1, return to step S33, until the maximum value of the wavelet decomposition scale is reached, and the decomposition results of the transmission power time series data after arbitrary correction at J wavelet decomposition scales are obtained; S35: Decomposition results of the corrected transmission power time series data of N power grid devices at J wavelet decomposition scales are used as power grid data feature information.

6. The method for identifying abnormal faults and handling emergencies in power grids based on artificial intelligence according to claim 1, characterized in that: In the step S4, a power grid abnormality identification model is constructed to identify abnormal power grid faults, and emergency treatment is performed according to the abnormal power grid fault identification results, including: Construct a power grid anomaly recognition model to identify power grid abnormal faults, wherein the power grid anomaly recognition model takes power grid data feature information as input and takes power grid data anomaly recognition results as output, wherein the power grid anomaly recognition model includes an input layer, a feature information mapping layer, a fusion layer and an output layer, wherein the input layer is used to receive power grid data feature information, the feature information mapping layer is used to map the power grid data feature information, the fusion layer is used to fuse the mapping results of multi-dimensional power grid data feature information to obtain power grid data fusion features, and the output layer performs probability calculation on the power grid data fusion features to obtain the fault categories existing in different power grid equipment in the power grid as power grid anomaly recognition results, and performs emergency disposal on the corresponding power grid equipment according to the power grid abnormal fault recognition results; The abnormal fault identification process of the power grid based on the power grid abnormality identification model is as follows: S41: The input layer receives the grid data characteristic information; S42: The characteristic information mapping layer performs mapping processing on the characteristic information of the power grid data, wherein the decomposition result mapping processing formula of the J wavelet decomposition scales corresponding to the nth power grid equipment in the characteristic information of the power grid data is: f n =(f n (1),f n (2),...,f n (j),...,f n (J)); in: f n Indicates the mapping processing result of the nth power grid device in the power grid data characteristic information; f n (j) represents the decomposition result The mapping vector of W1 represents the mapping parameter matrix; σ(·) represents the activation function; S43: The fusion layer fuses the mapping results of the multi-dimensional power grid data feature information to obtain the power grid data fusion feature of each power grid device, wherein the power grid data fusion feature of the nth power grid device is: in: Represents the grid data fusion characteristics of the nth grid device; S44: The output layer performs probability calculation on the fusion features of the power grid data to obtain the fault categories of different power grid devices in the power grid, where the probability calculation result of the fault category of the nth power grid device is: in: represents the probability that the fault category of the nth power grid equipment is the sth fault category, s∈[1,S], S represents the total number of fault categories of the power grid equipment; w(s) represents the parameter matrix corresponding to the sth fault category; The fault category with the highest probability is taken as the fault category of the nth power grid equipment.

Images