Power system terminal behavior identification method based on deep learning

Abstract

The invention discloses a power system terminal behavior identification method based on deep learning, and relates to the technical field of power systems, and the method comprises the steps: carrying out the digital-graph conversion of a terminal data sequence, extracting feature points, and constructing a peak point core sampling array; setting a local boundary coefficient to carry out local extraction on the peak point core sampling array and carrying out feature extraction on the peak point core sampling array and the terminal sequence coefficient to obtain a terminal feature coefficient group; performing information stream conversion on the terminal feature coefficient group, constructing a terminal vector storage library, performing comparison on the terminal vector storage library to obtain a terminal code class distance, performing judgment supplementation according to the terminal code class distance, and restoring the terminal code class distance into a corresponding terminal feature stream segment to obtain a classification stream segment storage library; setting a threshold reference matrix to perform behavior screening on the classification stream segment storage library to obtain abnormal data behaviors; abnormal behaviors of the terminal equipment are accurately identified, the operation state of the terminal equipment is monitored in real time, and safe and stable operation of a power system is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method for identifying power system terminal behavior based on deep learning. Background Art

[0002] In modern society, the power system is the core of the national energy infrastructure. With the rapid development of smart grids, the number of power system terminal devices has increased dramatically, including smart meters, distributed power access equipment, electric vehicle charging piles, etc. The behavioral characteristics of these terminal devices are complex and diverse and change dynamically, making accurate identification of their behavior a key issue to ensure the safe, stable and efficient operation of the power system.

[0003] Traditional methods for identifying terminal behavior in power systems mainly rely on pre-set rules to judge terminal behavior or analyze the statistical characteristics of historical data to identify behavior. They are effective in dealing with simple behaviors with clear rules, but for complex and changeable behavior patterns, the formulation of rules is difficult and lacks flexibility, making it difficult to adapt to the ever-changing needs of power systems.

[0004] Deep learning technology can automatically learn deep feature representations from a large amount of raw data without the need for manual complex feature engineering. Therefore, applying deep learning technology to the identification of power system terminal behavior, collecting terminal data, accurately and efficiently processing terminal data, and analyzing abnormal data in terminal data can provide early warning of abnormal behavior and provide strong protection for the safe operation of the power system. To this end, a method for identifying power system terminal behavior based on deep learning is provided. Summary of the invention

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A method for identifying power system terminal behavior based on deep learning, comprising the following steps:

[0007] Step S1: Collect terminal comprehensive data and capture time to construct terminal data sequence;

[0008] Step S2: extracting the terminal data sequence by digital graph to obtain the terminal coefficient peak point, performing morphological transformation according to the terminal coefficient peak point to obtain the peak point core sampling array;

[0009] Step S3: setting local boundary coefficients to perform local extraction on the peak point core sampling matrix to obtain a boundary peak point sampling matrix, and performing joint transformation on the boundary peak point sampling matrix and performing feature extraction with the terminal sequence coefficients to obtain a terminal feature coefficient group;

[0010] Step S4: performing information flow conversion on the terminal characteristic coefficient group to obtain terminal characteristic flow segments, extracting and constructing the terminal characteristic flow segments to obtain a terminal vector repository, performing comparison and selection on the terminal vector repository to obtain a reference terminal vector, traversing and comparing the terminal vector repository through the reference terminal vector to obtain a terminal code class distance, performing determination and supplementation based on the terminal code class distance, and obtaining a classification vector repository;

[0011] Step S5: dock and restore the classification vector repository according to the terminal feature flow segment to obtain the classification flow segment repository, set the terminal data threshold for form replacement, obtain the threshold reference matrix, perform behavior screening on the classification flow segment repository according to the threshold reference matrix, and obtain abnormal data behavior.

[0012] Preferably, the process of collecting terminal comprehensive data includes:

[0013] Comprehensively collect data from power system terminals to obtain terminal comprehensive data, and time-stamp the comprehensively collected terminal comprehensive data to obtain the capture time;

[0014] The terminal comprehensive data is time-sorted according to the obtained capture time to obtain a terminal data sequence.

[0015] Preferably, the process of extracting a digitogram from a terminal data sequence includes:

[0016] Performing coefficient conversion on the terminal data sequence to obtain terminal sequence coefficients, and performing spectrum conversion on the obtained terminal sequence coefficients to obtain a terminal coefficient spectrum diagram;

[0017] According to the obtained terminal coefficient spectrum diagram, peak points are traversed to obtain the terminal coefficient peak point.

[0018] Preferably, the process of performing morphological transformation according to the terminal coefficient peak point includes:

[0019] Sorting the obtained terminal coefficient peak points to obtain coefficient peak point sorting;

[0020] Setting extraction conditions according to the obtained coefficient peak point ranking, performing peak point extraction on the coefficient peak point ranking based on the extraction conditions, and obtaining the sample coefficient peak point;

[0021] An original peak point sampling array is constructed based on the order of peak point extraction, and the original peak point sampling array is core-filled according to the obtained sample coefficient peak points to obtain a peak point core sampling array.

[0022] Preferably, the process of jointly transforming the boundary peak sampling matrix and extracting features with the terminal sequence coefficients includes:

[0023] A local boundary coefficient is set according to the peak point core sampling array, and the peak point core sampling array is locally extracted according to the local boundary coefficient to obtain a boundary peak point sampling array;

[0024] Adaptively adjust the boundary peak point sampling array to obtain linear sampling segments, and sequentially combine the linear sampling segments based on the adaptive adjustment order to obtain a linear boundary sampling group;

[0025] The obtained linear boundary sampling group is uploaded to the terminal sequence coefficient, and the terminal sequence coefficient is cyclically screened through the linear boundary sampling group to obtain the terminal characteristic coefficient group.

[0026] Preferably, the process of cyclically screening the terminal sequence coefficients through the linear boundary sampling group includes:

[0027] The starting endpoint is set according to the terminal sequence coefficient, the linear boundary sampling group is initially docked based on the starting endpoint, and the linear boundary sampling group is uploaded to the docking position;

[0028] The terminal sequence coefficients are docked and extracted through the linear boundary sampling group to obtain the segment boundary coefficients;

[0029] The sliding interval is set according to the starting endpoint, and the linear boundary sampling group is sequentially slid according to the sliding interval based on the terminal sequence coefficient to reach the next docking position, and the terminal sequence coefficient is docked and extracted through the linear boundary sampling group to obtain the segment boundary coefficient;

[0030] Based on the terminal sequence coefficient, the linear boundary sampling group is sequentially slid according to the slid spacing until the terminal sequence coefficient is covered;

[0031] The segment boundary coefficients are combined according to the order of sequential sliding to obtain the terminal characteristic coefficient group.

[0032] Preferably, the process of extracting and constructing the terminal characteristic flow segment includes:

[0033] Separate the terminal characteristic stream segments into code categories to obtain the number of code divisions;

[0034] Summarize the number of codes of the obtained terminal characteristic flow segments to obtain the comprehensive characteristic code quantity;

[0035] Based on the terminal feature flow segment, the number of code divisions is separated and extracted according to the obtained comprehensive feature code quantity to obtain the feature code frequency, and a vector is constructed according to the feature code frequency to obtain a feature code vector;

[0036] The capture period is set according to the obtained capture time, and the obtained feature code vectors are comprehensively counted based on the capture period to build a terminal vector storage library.

[0037] Preferably, the process of determining and supplementing according to the terminal code class distance includes:

[0038] Set a matching threshold, determine the deviation of the terminal code class distance according to the matching threshold, obtain a matching feature vector, and construct a matching vector set for the obtained matching feature vector;

[0039] Performing control screening on the terminal vector repository according to the matching vector set to obtain a screening vector repository;

[0040] Based on the reference terminal vector, the features of the matching vector set are supplemented according to the obtained screening vector repository to obtain a classification vector repository.

[0041] Preferably, the process of setting the terminal data threshold for form replacement includes:

[0042] Based on the classified flow segment repository, terminal data thresholds are set according to terminal comprehensive data, and the terminal data thresholds are standardized and checked to obtain standardized threshold flow segments;

[0043] Encode and segment the terminal feature flow segment based on the classified flow segment repository to obtain the terminal feature code element segment;

[0044] The obtained standard threshold flow segment is formatted and a threshold reference matrix is ​​obtained.

[0045] Preferably, the process of performing behavior screening on the classified flow segment repository according to the threshold reference matrix includes:

[0046] Based on the classification flow segment repository, analog judgment is performed on the terminal feature code element segment according to the threshold reference matrix to obtain a reference judgment result;

[0047] According to the obtained reference judgment result, the terminal feature flow segment is forward matched to obtain abnormal terminal data, and according to the obtained abnormal terminal data, behavior analogy is performed to obtain abnormal data behavior.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. The collected terminal data of the power system are combined into a terminal data sequence in chronological order. By converting it into a signal spectrum, the subtle and key feature differences in the spectrum can be captured. The feature segments in the spectrum are extracted to construct a peak core sampling array to extract features from the terminal data sequence and obtain a terminal feature coefficient group. By reducing the data dimension, the computational complexity is reduced, and the efficiency of data processing is greatly improved.

[0050] 2. Capture the similarity of terminal feature coefficient groups, and classify them according to the similarity between different terminal feature coefficient groups to obtain a classification vector repository. Classify the status of the equipment to facilitate early maintenance or replacement of key components of the equipment, avoid sudden equipment failures affecting power supply, and screen the classification vector repository by setting terminal data thresholds to obtain abnormal data. Based on the screened abnormal data, it is possible to warn of possible equipment failures in advance, gain valuable time for the maintenance and management of the power system, and ensure the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0053] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] like Figure 1 As shown, a method for identifying power system terminal behavior based on deep learning includes the following steps:

[0055] Step S1: Collect terminal comprehensive data and capture time to construct terminal data sequence;

[0056] Step S2: extracting the terminal data sequence by digital graph to obtain the terminal coefficient peak point, performing morphological transformation according to the terminal coefficient peak point to obtain the peak point core sampling array;

[0057] Step S3: setting local boundary coefficients to perform local extraction on the peak point core sampling matrix to obtain a boundary peak point sampling matrix, and performing joint transformation on the boundary peak point sampling matrix and performing feature extraction with the terminal sequence coefficients to obtain a terminal feature coefficient group;

[0058] Step S4: performing information flow conversion on the terminal characteristic coefficient group to obtain terminal characteristic flow segments, extracting and constructing the terminal characteristic flow segments to obtain a terminal vector repository, performing comparison and selection on the terminal vector repository to obtain a reference terminal vector, traversing and comparing the terminal vector repository through the reference terminal vector to obtain a terminal code class distance, performing determination and supplementation based on the terminal code class distance, and obtaining a classification vector repository;

[0059] Step S5: dock and restore the classification vector repository according to the terminal feature flow segment to obtain the classification flow segment repository, set the terminal data threshold for form replacement, obtain the threshold reference matrix, perform behavior screening on the classification flow segment repository according to the threshold reference matrix, and obtain abnormal data behavior.

[0060] It should be further explained that, in the specific implementation process, the process of collecting terminal comprehensive data includes:

[0061] Comprehensively collect data from power system terminals to obtain comprehensive terminal data;

[0062] The terminal comprehensive data includes electrical quantity data, equipment status data and environmental data, wherein the electrical quantity data includes but is not limited to voltage, current, power, and frequency, the equipment status data includes but is not limited to switch status and operating time, and the environmental data includes but is not limited to temperature and humidity;

[0063] Time-stamp the fully collected terminal comprehensive data to obtain the capture time;

[0064] Associating the obtained terminal comprehensive data with the corresponding capture time;

[0065] Time-sorting the terminal comprehensive data according to the acquired capture time to obtain a terminal data sequence;

[0066] The time sorting means sorting according to the time sequence of the capture time of each terminal comprehensive data to obtain a time series sequence, that is, the terminal data sequence. Then, according to the electrical quantity data, equipment status data and environmental data included in the terminal comprehensive data, the terminal data sequence includes an electrical quantity data sequence, an equipment status data sequence and an environmental data sequence.

[0067] Perform digital image extraction on the terminal data sequence to obtain the terminal coefficient peak point. The specific process includes:

[0068] Acquire a terminal data sequence, perform coefficient conversion on the acquired terminal data sequence, and obtain a terminal sequence coefficient;

[0069] The coefficient conversion means converting the obtained terminal data sequence into a signal form, that is, a terminal sequence coefficient. According to the electrical quantity data sequence, the equipment status data sequence and the environment data sequence included in the terminal data sequence, the terminal sequence coefficient includes an electrical quantity sequence coefficient, an equipment status sequence coefficient and an environment sequence coefficient.

[0070] Performing spectrum conversion on the obtained terminal sequence coefficients to obtain a terminal coefficient spectrum diagram;

[0071] The spectrum conversion means generating a spectrum form of a signal according to the obtained terminal sequence coefficients, that is, a terminal coefficient spectrum;

[0072] Further, according to the electrical quantity sequence coefficient, the equipment state sequence coefficient and the environment sequence coefficient included in the terminal sequence coefficient, the terminal coefficient spectrum diagram includes an electrical quantity coefficient spectrum diagram, an equipment state coefficient spectrum diagram and an environment coefficient spectrum diagram;

[0073] Perform peak point traversal according to the obtained terminal coefficient spectrum diagram to obtain the terminal coefficient peak point;

[0074] The peak point traversal means recording the spectrum at the intersection of the horizontal axis and the vertical axis as the peak point in the terminal coefficient spectrum diagram, and marking the obtained peak point as the terminal coefficient peak point.

[0075] According to the terminal coefficient peak point, the morphological transformation is performed to obtain the peak point core sampling array. The specific process includes:

[0076] The obtained terminal coefficient peak points are sorted in descending order to obtain the coefficient peak point sorting, wherein "sequence" means that the values ​​at the intersection of the terminal coefficient peak point and the ordinate are in descending order;

[0077] Setting extraction conditions according to the obtained coefficient peak point ranking, performing peak point extraction on the coefficient peak point ranking based on the extraction conditions, and obtaining the sample coefficient peak point;

[0078] The extraction condition represents a rule for selecting the terminal coefficient peak point set for the coefficient peak point sorting, and the extraction condition is to select the 1st, ath, a+5th, a+10th, a+20th, a+30th, a+40th, a+50th, a+60th, a+70th, a+80th, a+90th, a+10th, a+110th, a+120th, a+130th, a+140th, a+15th, a+16th, a+17th, a+18th, a+19th, a+20th, a+21th, a+32th, a+40th, a+50th, a+60th, a+70th, a+80th, a+90th, a+10th, a+110th, a+120th, a+130th, a+140th, a+15th, a 1st 1st The first and second digits are the first and nth digits, where n is the last digit of the terminal coefficient peak point in the coefficient peak point sorting, that is, there are n terminal coefficient peak points in the coefficient peak point sorting, and a is a positive integer greater than 1. The selection of a is related to the number of terminal coefficient peak points in the coefficient peak point sorting and must satisfy ;

[0079] Constructing an original peak point sampling matrix based on the order of peak point extraction, wherein the original peak point sampling matrix is ​​expressed as a matrix of 3 rows and 3 columns;

[0080] Perform core filling on the original peak point sampling array according to the obtained sample coefficient peak points to obtain a peak point core sampling array;

[0081] The core filling means uploading the sample coefficient peak points to the original peak point sampling array in the order in which the peak points are extracted, and replacing the elements at the original positions to obtain the peak point core sampling array, wherein the 1st sample coefficient peak point is uploaded to the 1st row and 1st column position of the original peak point sampling array, the ath sample coefficient peak point is uploaded to the 1st row and 2nd column position of the original peak point sampling array, the a+5th sample coefficient peak point is uploaded to the 1st row and 3rd column position of the original peak point sampling array, the a+10th sample coefficient peak point is uploaded to the 2nd row and 1st column position of the original peak point sampling array, the a+20th sample coefficient peak point is uploaded to the 2nd row and 2nd column position of the original peak point sampling array, and the a+15th sample coefficient peak point is uploaded to the 2nd row and 3rd column position of the original peak point sampling array. The peak point of the bit sample coefficient is uploaded to the 2nd row and 3rd column position of the original peak point sampling array. The peak point of the bit sample coefficient is uploaded to the original peak point sampling array at the 3rd row and 1st column position. The nth sample coefficient peak point is uploaded to the 3rd row and 2nd column position of the original peak point sampling array, and the nth sample coefficient peak point is uploaded to the 3rd row and 3rd column position of the original peak point sampling array.

[0082] Acquire a peak core sampling matrix, and set a local boundary coefficient according to the peak core sampling matrix, wherein the local boundary coefficient is expressed in a function form; in particular, in this embodiment, the local boundary coefficient is based on a wavelet function, wherein the wavelet function includes but is not limited to Haar wavelet, Symlets wavelet, and Morlet wavelet;

[0083] According to the obtained local boundary coefficient, the peak point core sampling array is locally extracted to obtain the boundary peak point sampling array;

[0084] The local extraction means that the local boundary coefficient is multiplied by each element in the peak point core sampling matrix to obtain the boundary peak point sampling matrix, and the corresponding position element after multiplication with the local boundary coefficient is recorded as the local coefficient peak point;

[0085] Adaptively adjusting the obtained boundary peak point sampling matrix to obtain a linear sampling segment, wherein the linear sampling segment includes a first sampling segment, a second sampling segment, and a third sampling segment;

[0086] The adaptive adjustment means splitting the boundary peak point sampling array into a group according to each row, and each split row is recorded as a linear sampling segment, then the first row is recorded as the first sampling segment, the second row is recorded as the second sampling segment, and the third row is recorded as the third sampling segment;

[0087] sequentially combining the obtained linear sampling segments based on the order of the adaptive adjustment to obtain a linear boundary sampling group;

[0088] The sequential combination means combining the first sampling segment, the second sampling segment and the third sampling segment in sequence to obtain a vector of one row and nine columns, which is recorded as a linear boundary sampling group;

[0089] The obtained linear boundary sampling group is uploaded to the terminal sequence coefficient, and the terminal sequence coefficient is cyclically screened through the linear boundary sampling group to obtain the terminal characteristic coefficient group;

[0090] It should be further explained that, in the specific implementation process, the cyclic screening process includes:

[0091] Setting a starting endpoint according to the obtained terminal sequence coefficient, wherein the starting endpoint represents the first element of the terminal sequence coefficient, that is, the element at the first sampling point of the signal;

[0092] Initially docking the linear boundary sampling group based on the starting endpoint, and uploading the linear boundary sampling group to the docking position, wherein the initial docking means aligning the elements of the first row and first column of the boundary peak sampling array with the starting endpoint, and the docking position means the aligned elements of the first row and first column;

[0093] The terminal sequence coefficients are docked and extracted through the linear boundary sampling group to obtain the segment boundary coefficients;

[0094] The docking extraction means multiplying the elements in the linear boundary sampling group with the elements at the corresponding positions of the terminal sequence coefficients, and summing the multiplication results at each position to obtain the segment boundary coefficients, and the segment boundary coefficients represent the characteristic coefficients corresponding to the starting endpoint positions;

[0095] A sliding distance is set according to the obtained starting endpoint, where the sliding distance represents the distance of the linear boundary sampling group moving to the next docking position on the terminal sequence coefficient. In this embodiment, the sliding distance is one sampling point away, that is, the next docking position one sliding distance away from the starting endpoint is the second element of the terminal sequence coefficient, that is, the element at the second sampling point of the signal;

[0096] Based on the terminal sequence coefficient, the linear boundary sampling group is sequentially slid according to the obtained sliding spacing to reach the next docking position, and the terminal sequence coefficient is docked and extracted through the linear boundary sampling group to obtain the segmented boundary coefficient, wherein the sequential sliding means moving the linear boundary sampling group to a position of a sliding spacing, and the next docking position is the second element of the terminal sequence coefficient;

[0097] Based on the terminal sequence coefficient, the linear boundary sampling group is sequentially slid according to the obtained sliding spacing until the terminal sequence coefficient is covered, that is, the linear boundary sampling group is moved once to the third element of the terminal sequence coefficient, the fourth element of the terminal sequence coefficient, the fifth element of the terminal sequence coefficient, ..., the last element of the terminal sequence coefficient, where "the last element of the terminal sequence coefficient" means that the last element of the linear boundary sampling group coincides with the last element of the terminal sequence coefficient, that is, the end point is reached;

[0098] Combining the obtained segment boundary coefficients according to the order of sequential sliding to obtain a terminal characteristic coefficient group;

[0099] The combination means that according to the order of sequential sliding, the segment boundary coefficients at the corresponding positions are obtained for each sliding interval, and the segment boundary coefficients are sequentially merged to obtain a terminal characteristic coefficient group. For example, the segment boundary coefficients obtained according to the sequential sliding are h1, h2, h3, h4, h5, h6, and h7, respectively, and the terminal characteristic coefficient group is [h1, h2, h3, h4, h5, h6, and h7].

[0100] In particular, extracting the characteristic information of the terminal sequence coefficients in matrix form can greatly speed up the calculation and improve the feature extraction speed. At the same time, local feature extraction is conducive to capturing more subtle characteristic information of the data, improving the accuracy of data analysis, and facilitating the identification of the behavior to which the data corresponds.

[0101] Acquire a terminal characteristic coefficient group, perform information flow conversion on the obtained terminal characteristic coefficient group, and obtain a terminal characteristic flow segment;

[0102] The information stream conversion means converting the terminal characteristic coefficient group into a data stream composed of binary code elements, and the encoding rules for converting the terminal characteristic coefficient group corresponding to each power system are the same;

[0103] The terminal feature flow segment is extracted and constructed to obtain the terminal vector repository. The specific process includes:

[0104] Perform code-class separation on the obtained terminal characteristic stream segment to obtain code division times, wherein the code division times include zero code frequency and non-zero code frequency;

[0105] The code class separation means that the number of occurrences of binary code elements of 1 in the terminal feature stream segment is counted to obtain the non-zero code frequency, and the number of occurrences of binary code elements of 0 is counted to obtain the zero code frequency;

[0106] Summarize the number of codes of the obtained terminal characteristic flow segments to obtain the comprehensive characteristic code quantity;

[0107] The code number summary means that the number of binary code elements in the terminal characteristic stream segment is counted to obtain the comprehensive characteristic code quantity;

[0108] Based on the terminal characteristic flow segment, the number of code divisions is separated and extracted according to the obtained comprehensive characteristic code quantity to obtain the characteristic code frequency, wherein the characteristic code frequency includes zero code frequency and non-zero code frequency;

[0109] The obtained zero code frequency is marked as F0, where , represents the frequency of zero code, and Z represents the amount of comprehensive feature code;

[0110] The obtained non-zero code frequency is marked as F1, where , Indicates the frequency of non-zero code;

[0111] A vector is constructed according to the obtained characteristic code frequency to obtain a characteristic code vector, and the obtained characteristic code vector is associated with the corresponding terminal characteristic flow segment, indicating that each terminal characteristic flow segment has a corresponding characteristic code vector;

[0112] The vector construction means that a two-dimensional vector is formed according to the obtained zero code frequency and non-zero code frequency, that is, the feature code vector, and the obtained feature code vector is marked as ,in, ;

[0113] The capture period is set according to the obtained capture time. The capture period is a time period during which the corresponding terminal data sequence is obtained. In order to make the data analysis more accurate, sufficient terminal data sequences are required. Therefore, the capture period is set to be long enough to obtain sufficient terminal data sequences for analysis reference.

[0114] Based on the capture cycle, the obtained feature code vector is comprehensively counted, and a terminal vector repository is constructed, the obtained feature code vector is uploaded to the terminal vector repository, and the feature code vector in the terminal vector repository is recorded as , i represents the number of the feature code vector in the terminal vector storage library, i=1, 2, 3, ..., v1, v1 is a positive integer.

[0115] The obtained terminal vector repository is selected by comparison to obtain a reference terminal vector, and the obtained reference terminal vector is recorded as ;

[0116] The comparison selection means that a feature code vector is randomly selected in the terminal vector repository and recorded as a reference terminal vector, which is used as a reference vector to be compared with the remaining vectors in the terminal vector repository, and the vector type is divided according to the comparison result;

[0117] Traversing and comparing the terminal vector repository according to the obtained reference terminal vector to obtain the terminal code class distance;

[0118] Furthermore, the traversal comparison compares the reference terminal vector with each feature code vector in the terminal vector repository, respectively, obtains the terminal code class distance corresponding to the reference terminal vector and each feature code vector in the terminal vector repository, and marks the obtained terminal code class distance as ,in, , represents the terminal code class distance between the reference terminal vector and the i-th feature code vector in the terminal vector repository;

[0119] In particular, the comparison of the reference terminal vector with each signature vector in the terminal vector repository does not include the comparison with the reference terminal vector itself in the terminal vector repository;

[0120] Setting a matching threshold, which is a threshold set according to the terminal code class distance, and is used to determine whether the distance between the reference terminal vector and the feature code vector in the terminal vector repository is within a specified threshold, that is, determining whether the reference terminal vector and which feature code vectors in the terminal vector repository have a high similarity through the matching threshold;

[0121] Determine the deviation of the terminal code class distance according to the obtained matching threshold, obtain a matching feature vector, construct a matching vector set for the obtained matching feature vector, and upload the obtained matching feature vector and the reference terminal vector to the matching vector set;

[0122] The deviation determination means recording the feature code vector corresponding to the terminal code class distance that is less than or equal to the matching threshold as the matching feature vector;

[0123] Performing control screening on the terminal vector repository according to the obtained matching vector set to obtain a screening vector repository;

[0124] The control screening means deleting the feature code vectors included in the matching vector set from the terminal vector repository, and the remaining feature code vectors constitute the screening vector repository;

[0125] Based on the reference terminal vector, feature supplement is performed on the matching vector set according to the obtained screening vector repository to obtain a classification vector repository;

[0126] It should be further explained that, in the specific implementation process, the feature supplement process includes:

[0127] Compare and select the matching vector set to obtain a reference terminal vector, and repeat the process of obtaining the terminal code class distance for the screening vector repository according to the obtained reference terminal vector;

[0128] Deviate the obtained terminal code class distance according to the matching threshold, upload the feature code vector with the matching feature vector as the result to the matching vector set, and screen the feature code vector with the matching feature vector as the result in the screening vector repository, and obtain the second supplemented matching vector set, which indicates that the feature value vector has the qualified similarity with the first selected reference terminal vector, that is, the terminal data sequence corresponding to the feature value vector has a high similarity and is determined to be the first class of terminal data. The second supplemented matching vector set indicates that similar feature code vectors continue to be matched in the screening vector repository until all similar feature code vectors of the first class are supplemented to the matching vector set, and a supplemented matching vector set is recorded as a classification vector repository;

[0129] Continue to perform comparison and selection in the second supplemented coincident vector set to obtain the reference terminal vector, repeat the process of obtaining the second supplemented coincident vector set, and obtain the third supplemented coincident vector set;

[0130] Until there is no feature code vector that meets the matching threshold in the screening vector repository, the feature supplement of the matching vector set is completed, and one of the remaining screening vector repositories is selected as the reference terminal vector, and another type of feature code vector is classified. The process of obtaining the classification vector repository is repeated to obtain another type of feature code vector, until all feature code vectors in the terminal vector repository are supplemented into the corresponding matching vector set, that is, the classification of the feature code vector is completed, that is, the classification of the terminal data sequence corresponding to each feature code vector is completed;

[0131] In particular, the same feature code vector is not selected for each comparison, and the feature code vector selected as the reference terminal vector cannot be selected repeatedly. Moreover, each time the matching vector set is supplemented, the matching threshold needs to be narrowed to ensure that the feature code vector in the supplemented screening vector repository is highly matched with the feature code in the matching vector set.

[0132] According to the terminal characteristic flow segment, the classification vector repository is docked and restored to obtain the classification flow segment repository;

[0133] The docking restoration means obtaining the terminal characteristic flow segment associated with each feature code vector in the classification vector repository, and replacing the corresponding feature code vector with the associated terminal characteristic flow segment to obtain the classification flow segment repository. Then, each classification flow segment repository represents a terminal data sequence with high similarity, and it cannot be ruled out that all of them are qualified terminal data sequences. It is necessary to perform abnormal identification on the terminal data sequence and obtain abnormal terminal data. Then, it can be determined that the power system terminal has abnormal behavior according to the abnormal terminal data, that is, the abnormal behavior is identified through the abnormal data.

[0134] Based on the classified flow segment repository, terminal data thresholds are set according to the obtained terminal comprehensive data, and the obtained terminal data thresholds are standardized and checked to obtain standardized threshold flow segments;

[0135] It should be further explained that, in the specific implementation process, the terminal data threshold value means setting the corresponding normal threshold value according to the terminal comprehensive data corresponding to the terminal characteristic flow segment in the classification flow segment repository, that is, the set threshold value is within the safety range and will not cause the comprehensive data generated by the abnormal behavior of the terminal, and is recorded as the terminal data threshold value, then each classification flow segment repository has a corresponding standard threshold flow segment, which is used to determine whether there is an abnormal situation in the terminal characteristic flow segment in the classification flow segment repository; the standard investigation means performing the same feature extraction on the terminal data threshold as the terminal characteristic flow segment, and obtaining the standard threshold flow segment of the same form, so as to perform abnormal investigation on the classification flow segment repository through the standard threshold flow segment, and the specific process includes:

[0136] Perform coefficient conversion on the terminal data threshold to obtain the terminal threshold coefficient;

[0137] The obtained linear boundary sampling group is uploaded to the terminal threshold coefficient, and the terminal threshold coefficient is cyclically screened through the linear boundary sampling group to obtain a terminal threshold coefficient group;

[0138] Performing information flow conversion on the obtained terminal threshold coefficient group to obtain a standard threshold flow segment;

[0139] Based on the classified flow segment repository, the obtained terminal characteristic flow segment is encoded and segmented to obtain a terminal characteristic code element segment;

[0140] The coding segmentation means grouping the terminal characteristic stream segment into groups of 7 binary symbols, and recording them as terminal characteristic symbol segments; in particular, if the last group is less than 7 binary symbols, zero symbols are added to the front end of the last group until 7 binary symbols are added;

[0141] Performing format replacement on the obtained standard threshold flow segment to obtain a threshold reference matrix;

[0142] Furthermore, the process of format replacement includes:

[0143] The standard threshold stream segment is encoded and segmented to obtain a threshold codeword segment, which means that the standard threshold stream segment is grouped into groups of 7 binary codewords, and the groups with less than 7 binary codewords are supplemented until 7 binary codewords are satisfied as a group, thereby obtaining a threshold reference matrix;

[0144] Separating the obtained threshold codeword segments within the group to obtain fixed codewords and variable codewords;

[0145] The intra-group separation means recording the first 4 binary code elements of the threshold code element segment as fixed code elements, and recording the last 3 binary code elements as variable code elements;

[0146] Performing joint matching on the obtained threshold codeword segments based on the standard threshold stream segments to obtain a joint threshold array;

[0147] The joint matching means constructing a matrix in the standard threshold stream segment according to three threshold codeword segments as a group, generating a matrix of three rows and seven columns, recorded as a joint threshold matrix, and if the last group is less than three threshold codeword segments, then the last three threshold codeword segments in the standard threshold stream segment are combined into a matrix, recorded as a joint threshold matrix;

[0148] Perform array transformation on the joint threshold matrix according to the fixed code elements and the variable code elements to obtain a threshold reference matrix;

[0149] The array conversion means that in the joint threshold array, a matrix of three rows and four columns corresponding to fixed codewords is recorded as a fixed codeword group, and a matrix of three rows and three columns corresponding to variable codewords is recorded as a variable codeword group, and row modulo 2 addition is performed on the variable codeword group based on the joint threshold array until the variable codeword group becomes a unit matrix, and at the same time, the changed value of the fixed codeword group is recorded, and a threshold adjustment matrix is ​​formed with the unit matrix, and then a matrix transposition operation is performed on the threshold adjustment matrix to obtain a threshold reference matrix, wherein the unit matrix represents a matrix of three rows and three columns, and only the elements of the first row and first column, the second row and second column, and the third row and third column are 1, and the remaining elements are 0;

[0150] Based on the classification flow segment repository, analog judgment is performed on the terminal characteristic code element segment according to the obtained threshold reference matrix to obtain a reference judgment result, wherein the reference judgment result includes a compliant code element segment and an abnormal code element segment;

[0151] It should be further explained that, in the specific implementation process, the analogy judgment process includes:

[0152] In the classification flow segment repository, the first terminal feature flow segment is sequentially compared with the threshold reference matrix according to the upload order of the terminal feature flow segments to obtain the feature threshold segment;

[0153] The sequential comparison means performing a matrix multiplication operation on the first terminal feature symbol segment of the terminal feature stream segment and the threshold reference matrix to obtain a feature threshold segment, then performing a matrix multiplication operation on the second terminal feature symbol segment and the threshold reference matrix to obtain a feature threshold segment, and continuing to perform a matrix multiplication operation on the third, fourth, ... until all terminal feature symbol segments and the threshold reference matrix to obtain a feature threshold segment, and performing statistics on the obtained feature threshold segments;

[0154] Performing segment code judgment on the obtained characteristic threshold segment to obtain a reference judgment result;

[0155] The segment code judgment indicates that when the characteristic threshold segment is a zero matrix, it is recorded as a compliant code element segment, and when the characteristic threshold segment is a non-zero matrix, it is recorded as an abnormal code element segment;

[0156] Based on the upload order of the terminal feature flow segments in the classified flow segment repository, the process of obtaining the reference judgment result by comparing the next terminal feature flow segment with the threshold reference matrix is ​​repeated until all the terminal feature flow segments in the classified flow segment repository are judged with the threshold reference matrix and the reference judgment result is obtained, and the analog judgment is completed;

[0157] Perform forward matching on the terminal feature stream segment according to the obtained abnormal code element segment to obtain abnormal terminal data, and perform behavior analogy according to the obtained abnormal terminal data to obtain abnormal data behavior;

[0158] The forward matching means that when there is an abnormal code element segment in the terminal feature stream segment, the terminal feature stream segment is recorded as abnormal terminal data, indicating that the terminal comprehensive data corresponding to the terminal feature stream segment is within an abnormal range, and the terminal comprehensive data is not within a safe range, indicating that the comprehensive data generated by the terminal behavior has an abnormal situation, and an early warning is required to avoid the safety hazards of the power system caused by the abnormal terminal data, and according to the location of the collection point of the terminal comprehensive data, the problem can be quickly located, so as to take measures to solve the problem in time, ensure the normal power consumption of users, and improve the reliability of power supply;

[0159] On the contrary, it means that the terminal characteristic flow segment is normal, that is, the terminal comprehensive data corresponding to the terminal characteristic flow segment is within the normal range and there is no abnormal data, indicating that there is no abnormality in the comprehensive data generated by the terminal behavior.

[0160] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for identifying terminal behavior of a power system based on deep learning, characterized in that: The following steps are involved: Step S1: Collect terminal comprehensive data and capture time to construct terminal data sequence; Step S2: extracting the terminal data sequence by digital graph to obtain the terminal coefficient peak point, performing morphological transformation according to the terminal coefficient peak point to obtain the peak point core sampling array; Step S3: setting local boundary coefficients to perform local extraction on the peak point core sampling matrix to obtain a boundary peak point sampling matrix, and performing joint transformation on the boundary peak point sampling matrix and performing feature extraction with the terminal sequence coefficients to obtain a terminal feature coefficient group; Step S4: performing information flow conversion on the terminal characteristic coefficient group to obtain terminal characteristic flow segments, extracting and constructing the terminal characteristic flow segments to obtain a terminal vector repository, performing comparison and selection on the terminal vector repository to obtain a reference terminal vector, traversing and comparing the terminal vector repository through the reference terminal vector to obtain a terminal code class distance, performing determination and supplementation based on the terminal code class distance, and obtaining a classification vector repository; Step S5: dock and restore the classification vector repository according to the terminal feature flow segment to obtain the classification flow segment repository, set the terminal data threshold for form replacement, obtain the threshold reference matrix, perform behavior screening on the classification flow segment repository according to the threshold reference matrix, and obtain abnormal data behavior.

2. According to a method for identifying power system terminal behavior based on deep learning according to claim 1, it is characterized in that: The process of collecting terminal comprehensive data includes: Comprehensively collect data from power system terminals to obtain terminal comprehensive data, and time-stamp the comprehensively collected terminal comprehensive data to obtain the capture time; The terminal comprehensive data is time-sorted according to the obtained capture time to obtain a terminal data sequence.

3. According to a method for identifying power system terminal behavior based on deep learning according to claim 1, it is characterized in that: The process of extracting the digital map from the terminal data sequence includes: Performing coefficient conversion on the terminal data sequence to obtain terminal sequence coefficients, and performing spectrum conversion on the obtained terminal sequence coefficients to obtain a terminal coefficient spectrum diagram; According to the obtained terminal coefficient spectrum diagram, peak points are traversed to obtain the terminal coefficient peak point.

4. According to a method for identifying power system terminal behavior based on deep learning according to claim 1, it is characterized in that: The process of morphological transformation according to the terminal coefficient peak point includes: Sorting the obtained terminal coefficient peak points to obtain coefficient peak point sorting; Setting extraction conditions according to the obtained coefficient peak point ranking, performing peak point extraction on the coefficient peak point ranking based on the extraction conditions, and obtaining the sample coefficient peak point; An original peak point sampling array is constructed based on the order of peak point extraction, and the original peak point sampling array is core-filled according to the obtained sample coefficient peak points to obtain a peak point core sampling array.

5. According to a method for identifying power system terminal behavior based on deep learning according to claim 1, it is characterized in that: The process of jointly transforming the boundary peak sampling matrix and extracting features from the terminal sequence coefficients includes: A local boundary coefficient is set according to the peak point core sampling array, and the peak point core sampling array is locally extracted according to the local boundary coefficient to obtain a boundary peak point sampling array; Adaptively adjust the boundary peak point sampling array to obtain linear sampling segments, and sequentially combine the linear sampling segments based on the adaptive adjustment order to obtain a linear boundary sampling group; The obtained linear boundary sampling group is uploaded to the terminal sequence coefficient, and the terminal sequence coefficient is cyclically screened through the linear boundary sampling group to obtain the terminal characteristic coefficient group.

6. The method for identifying terminal behavior of a power system based on deep learning according to claim 5 is characterized in that: The process of cyclically screening the terminal sequence coefficients through the linear boundary sampling group includes: The starting endpoint is set according to the terminal sequence coefficient, the linear boundary sampling group is initially docked based on the starting endpoint, and the linear boundary sampling group is uploaded to the docking position; The terminal sequence coefficients are docked and extracted through the linear boundary sampling group to obtain the segment boundary coefficients; The sliding distance is set according to the starting endpoint, and the linear boundary sampling group is sequentially slid according to the sliding distance based on the terminal sequence coefficient to reach the next docking position, and the terminal sequence coefficient is docked and extracted through the linear boundary sampling group to obtain the segment boundary coefficient; Based on the terminal sequence coefficient, the linear boundary sampling group is sequentially slid according to the slid spacing until the terminal sequence coefficient is covered; The segment boundary coefficients are combined according to the order of sequential sliding to obtain the terminal characteristic coefficient group.

7. The method for identifying terminal behavior of a power system based on deep learning according to claim 1, characterized in that: The process of extracting and constructing terminal feature flow segments includes: Separate the terminal characteristic stream segments into code categories to obtain the number of code divisions; Summarize the number of codes of the obtained terminal characteristic flow segments to obtain the comprehensive characteristic code quantity; Based on the terminal feature flow segment, the number of code divisions is separated and extracted according to the obtained comprehensive feature code quantity to obtain the feature code frequency, and a vector is constructed according to the feature code frequency to obtain a feature code vector; The capture period is set according to the obtained capture time, and the obtained feature code vectors are comprehensively counted based on the capture period to build a terminal vector storage library.

8. The method for identifying terminal behavior of a power system based on deep learning according to claim 1, characterized in that: The process of determining and supplementing based on the terminal code class distance includes: Set a matching threshold, determine the deviation of the terminal code class distance according to the matching threshold, obtain a matching feature vector, and construct a matching vector set for the obtained matching feature vector; Performing control screening on the terminal vector repository according to the matching vector set to obtain a screening vector repository; Based on the reference terminal vector, the features of the matching vector set are supplemented according to the obtained screening vector repository to obtain a classification vector repository.

9. The method for identifying terminal behavior of a power system based on deep learning according to claim 1, characterized in that: The process of setting the terminal data threshold for form replacement includes: Based on the classified flow segment repository, terminal data thresholds are set according to terminal comprehensive data, and the terminal data thresholds are standardized and checked to obtain standardized threshold flow segments; Encode and segment the terminal feature flow segment based on the classified flow segment repository to obtain the terminal feature code element segment; The obtained standard threshold flow segment is formatted and a threshold reference matrix is ​​obtained.

10. A method for identifying power system terminal behavior based on deep learning according to claim 9, characterized in that: The process of behavioral screening of the classified flow segment repository against the threshold reference matrix includes: Based on the classification flow segment repository, analog judgment is performed on the terminal feature code element segment according to the threshold reference matrix to obtain a reference judgment result; According to the obtained reference judgment result, the terminal feature flow segment is forward matched to obtain abnormal terminal data, and according to the obtained abnormal terminal data, behavior analogy is performed to obtain abnormal data behavior.

Images