Power grid data acquisition and analysis method and system based on artificial intelligence

Abstract

The invention discloses a power grid data acquisition and analysis method and system based on artificial intelligence, and relates to the technical field of power grid fault prediction. The method comprises the following steps: collecting power grid voltage data, carrying out preliminary fault diagnosis, and triggering a subsequent process if the fault is abnormal; collecting zero-sequence current data, constructing and preprocessing a zero-sequence matrix, obtaining a singular value order, decomposing the matrix, extracting a component matrix, carrying out transformation to obtain line component signals, and calculating the mean square similarity of any two line component signals; a line fault probability generation prediction report is obtained and uploaded to the cloud; component signals are extracted through singular value order decomposition, the similarity between lines is calculated, the fault probability is determined, the lines where faults possibly occur are located, a prediction report is generated for workers to analyze, and accurate prediction and rapid locating of the power grid line faults are achieved.

Description

Technical Field

[0001] This invention belongs to the field of power grid fault prediction technology, specifically relating to a power grid data acquisition and analysis method and system based on artificial intelligence. Background Technology

[0002] The power grid is a unified system integrating power generation, transmission, distribution, and consumption. Like the circulatory network of modern society, it comprises power plants, transmission lines, substations, and users. In daily life, the power grid safely and reliably delivers electricity from distant power plants to households, factories, schools, and hospitals, providing us with all the energy we need for lighting, cooling, heating, and power. It is not only the cornerstone of economic and social operation, safeguarding vital lifelines such as communication, transportation, and finance, but also an indispensable public infrastructure for modern life; its stable operation is directly related to national security and people's livelihoods.

[0003] The existing publication number: CN119644038A discloses a method for acquiring multidimensional feature data, performing a first preprocessing on the multidimensional feature data to obtain first multidimensional feature data, using a first neural network and a second neural network to extract spatial and temporal features from the multidimensional feature data respectively, constructing a prediction model based on the spatial and temporal features, defining a loss function for the prediction model, training the prediction model using an optimizer to obtain a first prediction model, and evaluating the first prediction model through evaluation indicators to achieve dynamic assessment and prediction of distribution network faults.

[0004] Although the above-mentioned technology processes the multidimensional feature data and constructs a prediction model based on the spatial and temporal features, the signal and noise separation effect is poor under the ferroresonant condition in the power grid system, resulting in distorted fault feature extraction. Furthermore, the fault features (such as exponential decay components and exponential decay sine components) change under different fault initial angles, making it easy to lose features and resulting in inaccurate prediction. Summary of the Invention

[0005] The purpose of this invention is to solve the problem that the fault-prone characteristics under ferroresonant conditions lead to inaccurate prediction of power grid line faults, and to propose a power grid data acquisition and analysis method and system based on artificial intelligence.

[0006] In a first aspect of this invention, a power grid data acquisition and analysis method based on artificial intelligence is first proposed, the method comprising:

[0007] Voltage data is collected from the power grid system. Preliminary fault diagnosis is performed based on the voltage data to obtain preliminary results. If the preliminary results are abnormal, subsequent detection procedures are triggered. The voltage data includes zero-sequence voltage and phase voltage amplitude.

[0008] Collect zero-sequence current data, construct a zero-sequence matrix based on the zero-sequence current data, and preprocess the zero-sequence matrix to obtain a new zero-sequence matrix; the new zero-sequence current data includes zero-sequence current values ​​and current sequences.

[0009] Obtain the singular value order, decompose the new zero-order matrix according to the singular value order and extract the component matrix, and obtain the component signal of the line by transforming the component matrix;

[0010] The mean square similarity of the component signals of any two lines is calculated, the line fault probability is calculated based on the mean square similarity, and a prediction report is generated and uploaded to the cloud based on the fault probability.

[0011] Optionally, a zero-sequence matrix is ​​constructed based on the zero-sequence current data, and a new zero-sequence matrix is ​​obtained by preprocessing the zero-sequence matrix, including:

[0012] The current sequence of each line is obtained by collecting zero-sequence current data through a total number of sampling points M. The number of rows m and columns n of the zero-sequence matrix are calculated based on the parity of the total number of sampling points M.

[0013] The current sequence is filled into an m×n zero-sequence matrix, and each column of the zero-sequence matrix is ​​normalized to obtain a new zero-sequence matrix.

[0014] Optionally, the new zero-order matrix is ​​decomposed according to the singular value order to obtain a singular value sequence and component matrices, including:

[0015] The correspondence between the three types of zero-sequence current components and singular values ​​is obtained from the parameter manual, resulting in five singular values, which form a singular value sequence. The correspondence is that the fundamental component generates two singular values ​​with the closest amplitude, the exponentially decaying sine component generates two singular values ​​with the closest amplitude, and the exponentially decaying component generates one singular value. The order of the singular values ​​is 5.

[0016] The new zero-order matrix is ​​transformed into a left auxiliary matrix and a right auxiliary matrix using matrix multiplication.

[0017] Perform eigenvalue decomposition on the left auxiliary matrix and the right auxiliary matrix respectively to obtain the left singular vector matrix and the right singular vector matrix;

[0018] Extracting the c-th column vector from the left singular vector matrix yields... Extracting the c-th column vector from the right singular vector matrix yields... ;

[0019] According to the above The above The component matrix is ​​obtained from the singular value sequence;

[0020]

[0021] in, The component matrix corresponding to the c-th singular value It is the c-th column vector of the left singular vector matrix. It is the transpose of the c-th column vector of the right singular vector matrix. Represented as the c-th singular value;

[0022] The component signals of the line are obtained by transforming the component matrix.

[0023] Optionally, the component signals of the line are obtained by transforming the component matrix, including:

[0024] The exponentially decaying sinusoidal component signal is obtained by performing an inverse transform on the component matrix corresponding to the second singular value of the exponentially decaying sinusoidal component. The process is as follows:

[0025]

[0026] in, Let i be the exponentially decaying sinusoidal component signal at the i-th sampling point. Let be the component matrix corresponding to the second singular value, where e is a natural number and i represents the sampling point index of the exponentially decaying sinusoidal component. Where S is the attenuation coefficient and S is the sampling period. It is the natural angular frequency of the equivalent circuit. It is the initial phase of the exponentially decaying sinusoidal component.

[0027] Optionally, the mean square similarity of the component signals of any two lines is calculated, and the line fault probability is calculated based on the mean square similarity, including:

[0028] The process of collecting the current values ​​at the i-th sampling point of the two lines and calculating the similarity is as follows:

[0029]

[0030]

[0031] in, Let be the zero-sequence current similarity coefficient between the k-th and j-th lines. Let M represent the mean square similarity of the k-th line, where M is the total number of sampling points for the exponentially decaying sinusoidal component. and For the current values ​​at the i-th sampling point of the k-th and j-th lines, and This is the average of the current values ​​of the k-th and j-th lines;

[0032] Based on the exponential decay component, the third singular value of each line is extracted from the singular value sequence of all lines to form a singular value set, and the maximum value in the singular value set is taken as the fault threshold; the fault threshold is a dynamic threshold.

[0033] The process of calculating the line fault probability using mean square similarity is as follows:

[0034]

[0035] in, Let be the failure probability of the k-th line. Let e ​​be the mean square similarity of the j-th route, and e be the natural constant.

[0036] The prediction result is obtained by comparing the failure probability with the failure threshold.

[0037] In a second aspect of this invention, an artificial intelligence-based power grid data acquisition and analysis system is proposed. The system includes a detection module, a preprocessing module, a decomposition module, and a diagnostic module, wherein:

[0038] The detection module is used to collect voltage data in the power grid system, perform preliminary fault diagnosis based on the voltage data to obtain preliminary results, and trigger subsequent detection processes if the preliminary results are abnormal; the voltage data includes zero-sequence voltage and phase voltage amplitude.

[0039] The preprocessing module is used to collect zero-sequence current data, construct a zero-sequence matrix based on the zero-sequence current data, and preprocess the zero-sequence matrix to obtain a new zero-sequence matrix; the new zero-sequence current data includes zero-sequence current values ​​and current sequences.

[0040] The decomposition module is used to obtain the singular value order, decompose the new zero-order matrix according to the singular value order and extract the component matrix, and obtain the component signal of the line by transforming the component matrix.

[0041] The diagnostic module is used to calculate the mean square similarity of the component signals of any two lines, calculate the line fault probability based on the mean square similarity, generate a prediction report based on the fault probability, and upload it to the cloud.

[0042] Optionally, the preprocessing includes a matrix construction module and a normalization module, wherein:

[0043] The matrix construction module is used to collect zero-sequence current data through a total number of sampling points M to obtain the current sequence of each line, and to calculate the number of rows m and columns n of the zero-sequence matrix based on the parity of the total number of sampling points M.

[0044] The normalization module is used to fill the current sequence into an m×n zero-sequence matrix, and normalize each column of the zero-sequence matrix to obtain a new zero-sequence matrix.

[0045] Optionally, the decomposition module includes a sequence construction module, a transformation module, a feature decomposition module, a matrix calculation module, and a transformation module, wherein:

[0046] The sequence construction module is used to obtain the correspondence between the three types of components of the zero-sequence current and the singular values ​​from the parameter manual to obtain five singular values ​​and form a singular value sequence. The correspondence is that the fundamental component generates two singular values ​​with the closest amplitude, the exponentially decaying sine component generates two singular values ​​with the closest amplitude, and the exponentially decaying component generates one singular value. The order of the singular values ​​is 5.

[0047] The transformation module is used to transform the new zero-order matrix into a left auxiliary matrix and a right auxiliary matrix according to matrix multiplication;

[0048] The eigenvalue decomposition module is used to perform eigenvalue decomposition on the left auxiliary matrix and the right auxiliary matrix respectively to obtain a left singular vector matrix and a right singular vector matrix;

[0049] The extraction module is used to extract the c-th column vector from the left singular vector matrix. Extracting the c-th column vector from the right singular vector matrix yields... ;

[0050] The matrix calculation module is used to calculate the matrix based on the matrix. The above The component matrix is ​​obtained from the singular value sequence;

[0051]

[0052] in, The component matrix corresponding to the c-th singular value It is the c-th column vector of the left singular vector matrix. It is the transpose of the c-th column vector of the right singular vector matrix. Represented as the c-th singular value;

[0053] The transformation module is used to obtain the component signals of the line by transforming the component matrix.

[0054] Optionally, the transformation module includes a component signal generation module, wherein:

[0055] The component signal generation module is used to perform an inverse transform on the component matrix corresponding to the two singular values ​​of the exponentially decaying sinusoidal component to obtain the exponentially decaying sinusoidal component signal. The process is as follows:

[0056]

[0057] in, Let i be the exponentially decaying sinusoidal component signal at the i-th sampling point. Let be the component matrix corresponding to the second singular value, where e is a natural number and i represents the sampling point index of the exponentially decaying sinusoidal component. Where S is the attenuation coefficient and S is the sampling period. It is the natural angular frequency of the equivalent circuit. It is the initial phase of the exponentially decaying sinusoidal component.

[0058] Optionally, the diagnostic module includes a similarity calculation module, a threshold generation module, a probability calculation module, and a prediction module, wherein:

[0059] The similarity calculation module is used to collect the current value of the i-th sampling point of the two lines and calculate the similarity process as follows:

[0060]

[0061]

[0062] in, Let be the zero-sequence current similarity coefficient between the k-th and j-th lines. Let M represent the mean square similarity of the k-th line, where M is the total number of sampling points for the exponentially decaying sinusoidal component. and For the current values ​​at the i-th sampling point of the k-th and j-th lines, and This is the average of the current values ​​of the k-th and j-th lines;

[0063] The threshold generation module is used to extract the third singular value of each line from the singular value sequence of all lines according to the exponential decay component to form a singular value set, and take the maximum value from the singular value set as the fault threshold; the fault threshold is a dynamic threshold.

[0064] The probability calculation module is used to calculate the fault probability of the line through mean square similarity as follows:

[0065]

[0066] in, Let be the failure probability of the k-th line. Let e ​​be the mean square similarity of the j-th route, and e be the natural constant.

[0067] The prediction module is used to obtain a prediction result by comparing the fault probability with the fault threshold.

[0068] The beneficial effects of this invention are:

[0069] This invention proposes an artificial intelligence-based method for power grid data acquisition and analysis. By collecting voltage data of power grid transmission lines, it preliminarily diagnoses anomalies and triggers subsequent detection. Then, it collects zero-sequence current data to construct and preprocess the zero-sequence matrix. It uses singular value decomposition to extract component signals, calculates the similarity between lines to determine the probability of faults, and locates the lines that may be faulty. Under the condition of ferroresonance, it can achieve accurate prediction and rapid location of power grid line faults. Attached Figure Description

[0070] The invention will now be further described with reference to the accompanying drawings.

[0071] Figure 1 A flowchart illustrating an artificial intelligence-based power grid data acquisition and analysis method provided in this embodiment of the invention;

[0072] Figure 2 This is a framework diagram of an artificial intelligence-based power grid data acquisition and analysis system provided in an embodiment of the present invention. Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and B can represent: A alone, A and B simultaneously, and B alone. Furthermore, descriptions involving "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" can explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0074] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0075] This invention provides an artificial intelligence-based method for power grid data acquisition and analysis. See also... Figure 1 , Figure 1 A flowchart illustrating an artificial intelligence-based power grid data acquisition and analysis method provided in this embodiment of the invention. The method includes the following steps:

[0076] S101: Collect voltage data in the power grid system, perform preliminary fault diagnosis based on the voltage data to obtain preliminary results, and trigger subsequent detection processes if the preliminary results are abnormal.

[0077] S102: Collect zero-sequence current data, construct a zero-sequence matrix based on the zero-sequence current data, and preprocess the zero-sequence matrix to obtain a new zero-sequence matrix;

[0078] S103, obtain the singular value order, decompose the new zero-sequence matrix according to the singular value order and extract the component matrix, and obtain the component signal of the line by transforming the component matrix;

[0079] S104 calculates the mean square similarity of the component signals of any two lines, calculates the line fault probability based on the mean square similarity, generates a prediction report based on the fault probability, and uploads it to the cloud.

[0080] The voltage data includes zero-sequence voltage and phase voltage amplitude.

[0081] The new zero-sequence current data includes zero-sequence current values ​​and current sequences.

[0082] This invention provides an artificial intelligence-based power grid data acquisition and analysis method. By collecting voltage data of power grid transmission lines, it initially diagnoses anomalies and triggers subsequent detection. Then, it collects zero-sequence current data to construct and preprocess a zero-sequence matrix. It uses singular value decomposition to extract component signals, calculates the similarity between lines to determine the probability of a fault, locates the line that may be faulty, and generates a prediction report for staff to analyze, thereby achieving accurate prediction and rapid location of power grid line faults.

[0083] In one implementation, preliminary fault diagnosis uses a lightweight threshold to first confirm whether a fault has occurred, avoiding misprocessing of normal signals and wasted computing power in subsequent complex processes. The process of obtaining preliminary results based on voltage data includes:

[0084] Start the system voltage real-time monitoring program to continuously collect the real-time value of the zero-sequence voltage of the transmission line;

[0085] If the real-time value of the zero-sequence voltage is greater than the abnormal threshold, it is determined that there is an abnormality in the line and a grounding fault may occur, triggering the subsequent detection process; the abnormal threshold is equal to 0.2 times the phase voltage amplitude, which is set based on historical experimental data;

[0086] If the real-time value of the zero-sequence voltage is not greater than the abnormal threshold, the operation of continuous monitoring will be returned, and the subsequent detection process will not be started.

[0087] In one implementation, singular values ​​are essentially eigenvalues ​​that quantify the contribution of each independent signal component to the original current signal after the zero-sequence current signal matrix is ​​decomposed.

[0088] In one implementation, the zero-sequence voltage and zero-sequence current values ​​are collected within the current hour and their changes are analyzed. The probability of a transmission line failure within the next four hours is then calculated and predicted.

[0089] In one embodiment, a zero-sequence matrix is ​​constructed based on zero-sequence current data, and a new zero-sequence matrix is ​​obtained by preprocessing the zero-sequence matrix, including:

[0090] The current sequence of each line is obtained by collecting zero-sequence current data through a total number of sampling points M. The number of rows m and columns n of the zero-sequence matrix are calculated based on the parity of the total number of sampling points M.

[0091] The current sequence is filled into an m×n zero-sequence matrix, and each column of the zero-sequence matrix is ​​normalized to obtain a new zero-sequence matrix.

[0092] In one implementation, the current sequence is: The m×n zero-sequence matrix h is constructed based on the current sequence as follows:

[0093]

[0094] If M is even If M is an odd number, The determination of the size of m and n based on the parity of M is to ensure that each sampling point of the zero-sequence current is included in the matrix without loss or redundancy, and to fully preserve the transient characteristics. This makes the row and column dimensions of the zero-sequence matrix as balanced as possible, ensures the stability of the matrix rank structure, and lays the foundation for accurate separation of signal components by subsequent singular value decomposition.

[0095] The new zero-order matrix obtained by normalizing the zero-order matrix includes:

[0096]

[0097] d(a) is the normalized deviation of the a-th column of the zero-order matrix (a=[1,2,3,...,n]).

[0098] In one implementation, the size of the zero-sequence matrix is ​​determined by the parity of the total number of sampling points M, the matrix is ​​constructed by filling the current sequence, and the new zero-sequence matrix is ​​obtained by normalization. This makes the matrix construction more flexible, and the normalization enhances the comparability of the data and improves the accuracy of subsequent analysis.

[0099] In one embodiment, decomposing the new zero-order matrix according to the singular value order and extracting the singular value sequence and component matrices includes:

[0100] The correspondence between the three types of zero-sequence current components and singular values ​​is obtained from the parameter manual, resulting in five singular values, which form a singular value sequence. The correspondence is as follows: the fundamental component generates two singular values ​​with the closest amplitude, the exponentially decaying sine component generates two singular values ​​with the closest amplitude, and the exponentially decaying component generates one singular value. The order of the singular values ​​is 5.

[0101] Transform the new zero-order matrix into a left auxiliary matrix and a right auxiliary matrix using matrix multiplication;

[0102] Perform eigenvalue decomposition on the left auxiliary matrix and the right auxiliary matrix respectively to obtain the left singular vector matrix and the right singular vector matrix;

[0103] Extracting the c-th column vector from the left singular vector matrix yields... Extracting the c-th column vector from the right singular vector matrix yields... ;

[0104] according to , The component matrix is ​​obtained from the singular value sequence;

[0105]

[0106] in, The component matrix corresponding to the c-th singular value It is the c-th column vector of the left singular vector matrix. It is the transpose of the c-th column vector of the right singular vector matrix. Represented as the c-th singular value;

[0107] The component signals of the line are obtained by transforming the component matrix.

[0108] In one implementation, the new zero-order matrix H is transformed into a left auxiliary matrix and a right auxiliary matrix, and eigenvalue decomposition is performed to obtain the left singular vector matrix and the right singular vector matrix, including:

[0109] Let m be the left auxiliary matrix, with dimensions m×m. Let n be the right auxiliary matrix, with dimensions n×n. It is the transpose of H;

[0110]

[0111]

[0112] in, It is the transpose of the left singular vector matrix. It is the transpose of the left singular vector matrix. It is an m×m diagonal matrix with diagonal elements of 1. eigenvalues, It is an n×n diagonal matrix with diagonal elements of 1. eigenvalues;

[0113] Singular value sequences are The relationship between the new zero-order matrix H and the component matrices is as follows:

[0114]

[0115] in, It is a left singular vector matrix. Let H be the transpose of the right singular vector matrix, and r represents the rank of the new zero-order matrix H being equal to the singular value order, where r equals 5.

[0116] In one implementation, singular value decomposition is used to process zero-sequence current data to accurately extract various fault feature signals, thereby improving fault identification accuracy. Matrix decomposition is used to efficiently extract key features, enhancing the sensitivity and accuracy of power grid fault detection.

[0117] In one embodiment, obtaining the component signals of the line by transforming the component matrix includes:

[0118] The exponentially decaying sinusoidal component signal is obtained by performing an inverse transform on the component matrix corresponding to the second singular value of the exponentially decaying sinusoidal component. The process is as follows:

[0119]

[0120] in, Let i be the exponentially decaying sinusoidal component signal at the i-th sampling point. Let be the component matrix corresponding to the second singular value, where e is a natural number and i represents the sampling point index of the exponentially decaying sinusoidal component. Where S is the attenuation coefficient and S is the sampling period. It is the natural angular frequency of the equivalent circuit. It is the initial phase of the exponentially decaying sinusoidal component.

[0121] In one implementation, S was set through experiments based on historically collected data. The exponentially decaying sine component was adopted because it has unique advantages such as differential stability, precise separation, and quantizability. The exponentially decaying component and the fundamental component cannot meet the requirements. The component matrix corresponding to the two singular values ​​of the exponentially decaying sine component is used.

[0122] In one implementation, the exponentially decaying sinusoidal component signal of each sampling point is accurately calculated using a specific inverse transform formula, which effectively reconstructs the component signal, helps to accurately extract transient features in the power grid, provides an important basis for subsequent fault detection and location, and improves the accuracy of diagnosis.

[0123] In one embodiment, the mean square similarity of the component signals of any two lines is calculated, and the line fault probability is calculated based on the mean square similarity, including:

[0124] The process of collecting the current values ​​at the i-th sampling point of the two lines and calculating the similarity is as follows:

[0125]

[0126]

[0127] in, Let be the zero-sequence current similarity coefficient between the k-th and j-th lines. Let M represent the mean square similarity of the k-th line, where M is the total number of sampling points for the exponentially decaying sinusoidal component. and For the current values ​​at the i-th sampling point of the k-th and j-th lines, and This is the average of the current values ​​of the k-th and j-th lines;

[0128] Based on the exponential decay component, the third singular value of each line is extracted from the singular value sequence of all lines to form a singular value set, and the maximum value in the singular value set is taken as the fault threshold; the fault threshold is a dynamic threshold.

[0129] The process of calculating the line fault probability using mean square similarity is as follows:

[0130]

[0131] in, Let be the failure probability of the k-th line. Let e ​​be the mean square similarity of the j-th route, and e be the natural constant.

[0132] The prediction result is obtained by comparing the failure probability with the failure threshold.

[0133] In one implementation, the threshold is calculated using the singular value of the exponentially decaying component because this component has unique fault characteristics, single features, and scene adaptability. It perfectly matches the function of threshold screening fault features and triggering criterion switching. In essence, it can directly distinguish faulty and healthy lines by the magnitude of the singular value. The other two components are difficult to implement.

[0134] The average current value of the kth line is calculated as follows:

[0135]

[0136] If the probability of failure is greater than the fault threshold, it indicates that the circuit is likely to fail and personnel need to be dispatched for repair. If the probability of failure is not greater than the fault threshold, it indicates that the circuit is caused by accidental factors, such as a sudden rise or fall in voltage that may occur when a large machine is operating.

[0137] In one implementation, the probability of line faults is accurately assessed by calculating the mean square similarity of the line component signals and combining it with a dynamic fault threshold. This improves the accuracy of fault prediction, effectively captures the differences in fault characteristics between lines, enhances the sensitivity of fault identification, and allows the dynamic threshold to adapt to different operating states, thereby enhancing the system's adaptability and reliability and providing strong support for predicting line faults in the future.

[0138] Based on the same inventive concept, this invention also provides an artificial intelligence-based power grid data acquisition and analysis system. See also Figure 2 , Figure 2 A framework diagram of an artificial intelligence-based power grid data acquisition and analysis system provided in this embodiment of the invention includes a detection module, a preprocessing module, a decomposition module, and a diagnostic module, wherein:

[0139] The detection module is used to collect voltage data in the power grid system, perform preliminary fault diagnosis based on the voltage data to obtain preliminary results, and trigger subsequent detection processes if the preliminary results are abnormal; the voltage data includes zero-sequence voltage and phase voltage amplitude.

[0140] The preprocessing module is used to collect zero-sequence current data, construct a zero-sequence matrix based on the zero-sequence current data, and preprocess the zero-sequence matrix to obtain a new zero-sequence matrix; the new zero-sequence current data contains zero-sequence current values ​​and current sequences.

[0141] The decomposition module is used to obtain the singular value order, decompose the new zero-sequence matrix according to the singular value order and extract the component matrix, and obtain the component signal of the line by transforming the component matrix.

[0142] The diagnostic module is used to calculate the mean square similarity of the component signals of any two lines, calculate the line fault probability based on the mean square similarity, generate a prediction report based on the fault probability, and upload it to the cloud.

[0143] The present invention provides an artificial intelligence-based power grid data acquisition and analysis system. By collecting voltage data of power grid transmission lines, it initially diagnoses anomalies and triggers subsequent detection. Then, it collects zero-sequence current data to construct and preprocess a zero-sequence matrix. It uses singular value decomposition to extract component signals, calculates the similarity between lines to determine the probability of faults, locates the lines that may be faulty, and generates a prediction report for staff to analyze, thereby achieving accurate prediction and rapid location of power grid line faults.

[0144] In one embodiment, preprocessing includes a matrix construction module and a normalization module, wherein:

[0145] The matrix construction module is used to collect zero-sequence current data through a total number of sampling points M to obtain the current sequence of each line, and to calculate the number of rows m and columns n of the zero-sequence matrix based on the parity of the total number of sampling points M.

[0146] The normalization module is used to fill the current sequence into an m×n zero-sequence matrix, and normalize each column of the zero-sequence matrix to obtain a new zero-sequence matrix.

[0147] In one embodiment, a sequence module is constructed to obtain the correspondence between the three types of components of the zero-sequence current and the singular values ​​from the parameter manual to obtain five singular values ​​and form a singular value sequence. The correspondence is that the fundamental component generates two singular values ​​with the closest amplitude, the exponentially decaying sine component generates two singular values ​​with the closest amplitude, and the exponentially decaying component generates one singular value. The order of the singular values ​​is 5.

[0148] The transformation module is used to transform a new zero-order matrix into a left auxiliary matrix and a right auxiliary matrix based on matrix multiplication.

[0149] The eigenvalue decomposition module is used to perform eigenvalue decomposition on the left auxiliary matrix and the right auxiliary matrix to obtain the left singular vector matrix and the right singular vector matrix, respectively.

[0150] The extraction module is used to extract the c-th column vector from the left singular vector matrix. Extracting the c-th column vector from the right singular vector matrix yields... ;

[0151] The matrix calculation module is used to calculate the matrix value based on the matrix calculation results. , The component matrix is ​​obtained from the singular value sequence;

[0152]

[0153] in, The component matrix corresponding to the c-th singular value It is the c-th column vector of the left singular vector matrix. It is the transpose of the c-th column vector of the right singular vector matrix. Represented as the c-th singular value;

[0154] The transformation module is used to obtain the component signals of the line by transforming the component matrix.

[0155] In one embodiment, the transformation module includes a component signal generation module, wherein:

[0156] The component signal generation module is used to perform an inverse transform on the component matrix corresponding to the second singular value of the exponentially decaying sinusoidal component to obtain the exponentially decaying sinusoidal component signal. The process is as follows:

[0157]

[0158] in, Let i be the exponentially decaying sinusoidal component signal at the i-th sampling point. Let be the component matrix corresponding to the second singular value, where e is a natural number and i represents the sampling point index of the exponentially decaying sinusoidal component. Where S is the attenuation coefficient and S is the sampling period. It is the natural angular frequency of the equivalent circuit. It is the initial phase of the exponentially decaying sinusoidal component.

[0159] In one embodiment, the diagnostic module includes a similarity calculation module, a threshold generation module, a probability calculation module, and a prediction module, wherein:

[0160] The similarity calculation module is used to collect the current value of the i-th sampling point of two lines and calculate the similarity. The process is as follows:

[0161]

[0162]

[0163] in, Let be the zero-sequence current similarity coefficient between the k-th and j-th lines. Let M represent the mean square similarity of the k-th line, where M is the total number of sampling points for the exponentially decaying sinusoidal component. and For the current values ​​at the i-th sampling point of the k-th and j-th lines, and This is the average of the current values ​​of the k-th and j-th lines;

[0164] The threshold generation module is used to extract the third singular value of each line from the singular value sequence of all lines based on the exponential decay component to form a singular value set, and take the maximum value from the singular value set as the fault threshold; the fault threshold is a dynamic threshold.

[0165] The probability calculation module is used to calculate the fault probability of the line through mean square similarity. The process is as follows:

[0166]

[0167] in, Let be the failure probability of the k-th line. Let e ​​be the mean square similarity of the j-th route, and e be the natural constant.

[0168] The prediction module is used to obtain prediction results by comparing the failure probability with the failure threshold.

[0169] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A power grid data acquisition and analysis method based on artificial intelligence, characterized in that, The method includes: Voltage data is collected from the power grid system. Preliminary fault diagnosis is performed based on the voltage data to obtain preliminary results. If the preliminary results are abnormal, subsequent detection procedures are triggered. The voltage data includes zero-sequence voltage and phase voltage amplitude. Collect zero-sequence current data, construct a zero-sequence matrix based on the zero-sequence current data, and preprocess the zero-sequence matrix to obtain a new zero-sequence matrix; the new zero-sequence current data includes zero-sequence current values ​​and current sequences. Obtain the singular value order, decompose the new zero-order matrix according to the singular value order and extract the component matrix, and obtain the component signal of the line by transforming the component matrix; The mean square similarity of the component signals of any two lines is calculated, the line fault probability is calculated based on the mean square similarity, and a prediction report is generated and uploaded to the cloud based on the fault probability.

2. The method for acquiring and analyzing power grid data based on artificial intelligence according to claim 1, characterized in that, A zero-sequence matrix is ​​constructed based on the zero-sequence current data. The zero-sequence matrix is ​​then preprocessed to obtain a new zero-sequence matrix, including: The current sequence of each line is obtained by collecting zero-sequence current data through a total number of sampling points M. The number of rows m and columns n of the zero-sequence matrix are calculated based on the parity of the total number of sampling points M. The current sequence is filled into an m×n zero-sequence matrix, and each column of the zero-sequence matrix is ​​normalized to obtain a new zero-sequence matrix.

3. The method for power grid data acquisition and analysis based on artificial intelligence according to claim 1, characterized in that, The new zero-order matrix is ​​decomposed based on the singular value order to obtain a singular value sequence and component matrices, including: The correspondence between the three types of zero-sequence current components and singular values ​​is obtained from the parameter manual, resulting in five singular values, which form a singular value sequence. The correspondence is that the fundamental component generates two singular values ​​with the closest amplitude, the exponentially decaying sine component generates two singular values ​​with the closest amplitude, and the exponentially decaying component generates one singular value. The order of the singular values ​​is 5. The new zero-order matrix is ​​transformed into a left auxiliary matrix and a right auxiliary matrix using matrix multiplication. Perform eigenvalue decomposition on the left auxiliary matrix and the right auxiliary matrix respectively to obtain the left singular vector matrix and the right singular vector matrix; Extracting the c-th column vector from the left singular vector matrix yields... Extracting the c-th column vector from the right singular vector matrix yields... ; According to the above The above The component matrix is ​​obtained from the singular value sequence: ； in, The component matrix corresponding to the c-th singular value It is the c-th column vector of the left singular vector matrix. It is the transpose of the c-th column vector of the right singular vector matrix. Represented as the c-th singular value; The component signals of the line are obtained by transforming the component matrix.

4. The method for acquiring and analyzing power grid data based on artificial intelligence according to claim 3, characterized in that, The component signals of the line are obtained by transforming the component matrix, including: The exponentially decaying sinusoidal component signal is obtained by performing an inverse transform on the component matrix corresponding to the second singular value of the exponentially decaying sinusoidal component. The process is as follows: ； in, Let i be the exponentially decaying sinusoidal component signal at the i-th sampling point. Let be the component matrix corresponding to the second singular value, where e is a natural number and i represents the sampling point index of the exponentially decaying sinusoidal component. Where S is the attenuation coefficient and S is the sampling period. It is the natural angular frequency of the equivalent circuit. It is the initial phase of the exponentially decaying sinusoidal component.

5. The method for power grid data acquisition and analysis based on artificial intelligence according to claim 1, characterized in that, Calculate the mean square similarity of the component signals of any two lines, and calculate the line fault probability based on the mean square similarity, including: The process of collecting the current values ​​at the i-th sampling point of the two lines and calculating the similarity is as follows: ； ； in, Let be the zero-sequence current similarity coefficient between the k-th and j-th lines. Let M represent the mean square similarity of the k-th line, where M is the total number of sampling points for the exponentially decaying sinusoidal component. and For the current values ​​at the i-th sampling point of the k-th and j-th lines, and This is the average of the current values ​​of the k-th and j-th lines; Based on the exponential decay component, the third singular value of each line is extracted from the singular value sequence of all lines to form a singular value set, and the maximum value in the singular value set is taken as the fault threshold; the fault threshold is a dynamic threshold. The process of calculating the line fault probability using mean square similarity is as follows: ； in, Let be the failure probability of the k-th line. Let e ​​be the mean square similarity of the j-th route, and e be the natural constant. The prediction result is obtained by comparing the failure probability with the failure threshold.

6. A power grid data acquisition and analysis system based on artificial intelligence, characterized in that, The system includes a detection module, a preprocessing module, a decomposition module, and a diagnostic module, wherein: The detection module is used to collect voltage data in the power grid system, perform preliminary fault diagnosis based on the voltage data to obtain preliminary results, and trigger subsequent detection processes if the preliminary results are abnormal; the voltage data includes zero-sequence voltage and phase voltage amplitude. The preprocessing module is used to collect zero-sequence current data, construct a zero-sequence matrix based on the zero-sequence current data, and preprocess the zero-sequence matrix to obtain a new zero-sequence matrix; the new zero-sequence current data includes zero-sequence current values ​​and current sequences. The decomposition module is used to obtain the singular value order, decompose the new zero-order matrix according to the singular value order and extract the component matrix, and obtain the component signal of the line by transforming the component matrix. The diagnostic module is used to calculate the mean square similarity of the component signals of any two lines, calculate the line fault probability based on the mean square similarity, generate a prediction report based on the fault probability, and upload it to the cloud.

7. The power grid data acquisition and analysis system based on artificial intelligence according to claim 6, characterized in that, The preprocessing includes a matrix construction module and a normalization module, wherein: The matrix construction module is used to collect zero-sequence current data through a total number of sampling points M to obtain the current sequence of each line, and to calculate the number of rows m and columns n of the zero-sequence matrix based on the parity of the total number of sampling points M. The normalization module is used to fill the current sequence into an m×n zero-sequence matrix, and normalize each column of the zero-sequence matrix to obtain a new zero-sequence matrix.

8. The power grid data acquisition and analysis system based on artificial intelligence according to claim 6, characterized in that, The decomposition module includes a sequence construction module, a transformation module, a feature decomposition module, a matrix calculation module, and a transformation module, wherein: The sequence construction module is used to obtain the correspondence between the three types of components of the zero-sequence current and the singular values ​​from the parameter manual to obtain five singular values ​​and form a singular value sequence. The correspondence is that the fundamental component generates two singular values ​​with the closest amplitude, the exponentially decaying sine component generates two singular values ​​with the closest amplitude, and the exponentially decaying component generates one singular value. The order of the singular values ​​is 5. The transformation module is used to transform the new zero-order matrix into a left auxiliary matrix and a right auxiliary matrix according to matrix multiplication; The eigenvalue decomposition module is used to perform eigenvalue decomposition on the left auxiliary matrix and the right auxiliary matrix respectively to obtain a left singular vector matrix and a right singular vector matrix; The extraction module is used to extract the c-th column vector from the left singular vector matrix. Extracting the c-th column vector from the right singular vector matrix yields... ; The matrix calculation module is used to calculate the matrix based on the matrix. The above The component matrix is ​​obtained from the singular value sequence: ； in, The component matrix corresponding to the c-th singular value It is the c-th column vector of the left singular vector matrix. It is the transpose of the c-th column vector of the right singular vector matrix. Represented as the c-th singular value; The transformation module is used to obtain the component signals of the line by transforming the component matrix.

9. The power grid data acquisition and analysis system based on artificial intelligence according to claim 8, characterized in that, The transformation module includes a component signal generation module, wherein: The component signal generation module is used to perform an inverse transform on the component matrix corresponding to the second singular value of the exponentially decaying sinusoidal component to obtain the exponentially decaying sinusoidal component signal. The process is as follows: ； in, Let i be the exponentially decaying sinusoidal component signal at the i-th sampling point. Let be the component matrix corresponding to the second singular value, where e is a natural number and i represents the sampling point index of the exponentially decaying sinusoidal component. Where S is the attenuation coefficient and S is the sampling period. It is the natural angular frequency of the equivalent circuit. It is the initial phase of the exponentially decaying sinusoidal component.

10. The power grid data acquisition and analysis system based on artificial intelligence according to claim 6, characterized in that, The diagnostic module includes a similarity calculation module, a threshold generation module, a probability calculation module, and a prediction module, wherein: The similarity calculation module is used to collect the current value of the i-th sampling point of the two lines and calculate the similarity process as follows: ； ； in, Let be the zero-sequence current similarity coefficient between the k-th and j-th lines. Let M represent the mean square similarity of the k-th line, where M is the total number of sampling points for the exponentially decaying sinusoidal component. and For the current values ​​at the i-th sampling point of the k-th and j-th lines, and This is the average of the current values ​​of the k-th and j-th lines; The threshold generation module is used to extract the third singular value of each line from the singular value sequence of all lines according to the exponential decay component to form a singular value set, and take the maximum value from the singular value set as the fault threshold; the fault threshold is a dynamic threshold. The probability calculation module is used to calculate the fault probability of the line through mean square similarity as follows: ； in, Let be the failure probability of the k-th line. Let e ​​be the mean square similarity of the j-th route, and e be the natural constant. The prediction module is used to obtain a prediction result by comparing the fault probability with the fault threshold.

Images