Power failure information monitoring method and system based on real-time position

Abstract

The invention discloses a power failure information monitoring method and system based on a real-time position, and relates to the technical field of power failure information monitoring, and the method comprises the steps: collecting the data of an intelligent electric meter, calculating the power flow of a line, carrying out the sparse regression through an LASSO algorithm, and recognizing a key line which is strongly related to a power failure event; window sliding processing is carried out based on the key line, and an event matrix is generated according to statistical characteristics; the TF-IDF value of each feature is calculated based on the time frequency of the statistical features in the event matrix. According to the method, power flow data and voltage phase angle data are used as feature matrixes, the sparsity of LASSO regression is beneficial to eliminating non-critical line data with large noise or interference, the complexity and redundancy of a model are reduced, feature importance of a critical line in different time windows is enhanced through a TF-IDF weighting matrix, and the accuracy of the LASSO regression is improved. According to the method, the SVM model is helped to recognize the power failure event more accurately, and after SVM preliminary classification, isolated misjudgment can be corrected through MDC auxiliary recognition and clustering analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of power outage information monitoring, and in particular to a power outage information monitoring method and system based on real-time location. Background Art

[0002] With the increasing complexity of power systems and demand loads, grid monitoring and fault detection technologies are also developing rapidly. The popularization of smart meters and data acquisition technologies has brought new development opportunities for power monitoring, especially the application of real-time monitoring and big data analysis in the power sector, which has enabled power outage monitoring to shift from a single device level to a system level.

[0003] However, existing data collection and fault detection solutions have limitations in the accuracy of data analysis and processing. They easily ignore the dynamic characteristics of power outage events, cannot accurately identify lines that are highly correlated with power outage events, and are insufficiently responsive to changes in event patterns, resulting in limitations in their performance in the event of sudden power outages. In particular, they are relatively weak in the detection, positioning and alarm of complex power outages occurring in specific areas. They lack in-depth mining of smart meter data, especially in the identification of key lines and the extraction of fault features. They are often unable to accurately identify high-risk lines for power outages, resulting in a waste of resources and affecting the efficiency of response. Summary of the invention

[0004] In view of the problems existing in the above-mentioned existing real-time location-based power outage information monitoring method and system, the present invention is proposed.

[0005] The purpose of the present invention is to provide a power outage information monitoring method and system based on real-time location. The problem to be solved is that it is impossible to accurately identify lines that are highly related to power outage events, and the response to changes in event patterns is insufficient, especially in the detection, positioning and alarm of complex power outage events occurring in specific areas. There is a lack of deep mining of smart meter data, especially in the identification of key lines and the extraction of fault features. It is often impossible to accurately identify high-risk lines for power outages, resulting in a waste of resources and affecting the efficiency of the response.

[0006] The technical solution of the present invention is: a method for monitoring power outage information based on real-time location, comprising:

[0007] Collect smart meter data, calculate line power flow, perform sparse regression using the LASSO algorithm, and identify key lines that are strongly correlated with power outages;

[0008] Perform window sliding processing based on key lines and generate event matrix based on statistical characteristics;

[0009] Based on the time frequency of statistical features in the event matrix, the TF-IDF value of each feature is calculated, and the power outage label is predicted through the SVM (support vector machine) model. The fault feature vector is extracted based on the event matrix, and the cosine similarity is calculated. Clustering is performed based on the minimum diagnosable clustering MDC (Minimum Detectable Cluster), and correction judgment is made on the lines whose labels are predicted by the SVM model;

[0010] Based on the MDC clustering output, a location data set is constructed to perform location clustering and hierarchical real-time alarm.

[0011] As a preferred solution of the real-time location-based power outage information monitoring method of the present invention, wherein: the collecting of smart meter data includes:

[0012] Collect data based on smart meters, and collect data collected by environmental sensors and fault information reported by users;

[0013] The collected data will be standardized and cleaned, including removing outliers and noise.

[0014] As a preferred solution of the real-time location-based power outage information monitoring method of the present invention, wherein: the line power flow is calculated, sparse regression is performed through the LASSO algorithm, and key lines strongly related to the power outage event are identified, including:

[0015] Determine the voltage phase angle data of the line and the node, and calculate the power flow based on the voltage at both ends of the line, which is expressed as:

[0016] Pij=ViVjYijcos(θi-θj);

[0017] Where Pij represents the power flow of the voltage at nodes i and j at both ends of the line, Vi and Vj represent the i-th node and j-th node at both ends of the line respectively, Yij represents the admittance of the i-th node and j-th node as the two-end line, θi and θj represent the voltage phase angles of nodes i and j respectively;

[0018] The voltage data and power flow data are used as the feature matrix X, and whether the power outage event occurs is used as the element of the target vector y. 1 is set to indicate a power outage and 0 is set to indicate a normal state. The LASSO algorithm is used for sparse regression to identify the lines that are strongly correlated with the power outage event, which is expressed as:

[0019]

[0020] Where β represents the regression coefficient, each β represents the sensitivity of the line between nodes i and j to the power outage event, λ represents the regularization parameter, X represents the feature matrix, and y represents the target vector, indicating the power outage event status;

[0021] According to the β obtained by LASSO regression, the sum of the mean and standard deviation of the absolute value of the regression coefficient based on historical data is used as the sensitivity threshold. If β is greater than or equal to the sensitivity threshold, the line between the two end nodes represented by β is judged as a critical line.

[0022] As a preferred solution of the real-time location-based power outage information monitoring method of the present invention, wherein: the window sliding processing is performed based on the key line, and the event matrix is ​​generated according to the statistical characteristics, including:

[0023] Perform time window sliding processing based on the voltage and current data of the critical lines;

[0024] The mean, standard deviation, extreme value and slope within the window are calculated as statistical features, and the frequency of occurrence of statistical features of each key line is calculated as the time frequency;

[0025] The statistical characteristics and time frequency of each key line in each time window are summarized to generate an event matrix.

[0026] As a preferred solution of the power outage information monitoring method based on real-time location described in the present invention, wherein: the time frequency of the statistical features in the event matrix is ​​based on, the TF-IDF value of each feature is calculated, the power outage label is predicted through the SVM model, the fault feature vector is extracted based on the event matrix, and the cosine similarity is calculated, and clustering is performed based on the minimum diagnosable clustering MDC, and the line whose label is predicted by the SVM model is corrected and determined, including,

[0027] The time frequency of each statistical feature in the event matrix is ​​counted, and the TF-IDF value of each feature is calculated, which is expressed as:

[0028]

[0029] Where TF(a,d) represents the TF-IDF value of statistical feature a in window d, tf(a,d) represents the frequency of occurrence of statistical feature a in window d, N represents the total number of time windows, and df(a) represents the total number of statistical feature a in the time window;

[0030] Based on the calculation of TF(a,d), weighted processing is performed according to the event matrix to construct a weighted matrix;

[0031] Based on historical data, a label is assigned to each time window according to the power outage event record as a training set, with 1 representing a power outage event and 0 representing a non-power outage event;

[0032] The radial basis function kernel RBF is selected as the kernel function of SVM, and the SVM model is trained based on the training set. The training is performed by minimizing the classification error and maximizing the interval objective function, and the optimal parameters are determined by cross-validation.

[0033] Use the trained SVM model to predict labels for the input weight matrix;

[0034] Based on the feature combination of each key line in the event matrix within the time window, the fault feature vector of the line in the time window is extracted;

[0035] Calculate the cosine similarity between the fault feature vectors of any two lines, and use the sum of the mean and standard deviation of the historical data as the similarity threshold. If the cosine similarity is greater than or equal to the similarity threshold, it is judged that the states of the lines with corresponding similarity in the same time window are highly similar.

[0036] Clustering is performed based on the minimum diagnosable cluster MDC, traversing all line combinations of fault feature vectors, and generating multiple MDCs covering the state combinations of all lines in the same time window;

[0037] Correct the line that has been predicted by the SVM model and is judged to be in a power outage state. If other lines in the MDC cluster where the line is located are not in a power outage state, re-evaluate the state of the line.

[0038] If most of the lines in the MDC cluster where the line is located are judged to be in a power outage state by SVM, the confidence of the power outage event of the entire cluster is increased and marked as a high-confidence power outage state;

[0039] Output the final power outage status matrix, which contains the power outage status and confidence information of each line.

[0040] As a preferred solution of the power outage information monitoring method based on real-time location of the present invention, wherein: the location data set is constructed according to the MDC clustering output, and location clustering is performed, including:

[0041] The final power outage status matrix output by MDC clustering is combined with the geographic location information of the smart meter to construct a location data set. Each row of the data set contains the meter ID, power outage status, confidence, and geographic location.

[0042] The density-based clustering algorithm DBSCAN is selected for location clustering, and the radius parameter of the DBSCAN algorithm is determined by calculating the average neighbor distance in the geographic location data;

[0043] Based on the radius parameter, the average number of neighbors of the smart meter within the radius parameter is calculated, and the average number of neighbors of all smart meter locations is calculated to obtain the average number of neighbors of each meter within the radius parameter range;

[0044] Based on the average number of neighbors, the k-neighborhood distance of each smart meter within the radius parameter is calculated, and the k-neighborhood distances are sorted from small to large and a k-distance graph is drawn. Based on the obvious inflection point of the k-distance graph, the number of meters covered by the k-neighborhood distance corresponding to the inflection point is taken as the minimum sample number MinPts;

[0045] The DBSCAN clustering algorithm is applied. If the number of neighbors within the radius parameter of the meter location is greater than or equal to the minimum number of samples, it is marked as a core point, and a cluster is formed based on the radius parameter and the minimum number of samples with the core point as the starting point.

[0046] The DBSCAN clustering results are output to form multiple cluster groups, each of which represents a potential power outage area.

[0047] As a preferred solution of the real-time location-based power outage information monitoring method of the present invention, the hierarchical real-time alarm includes:

[0048] The sum of the average number and standard deviation of historical cluster groups is used as the alarm threshold;

[0049] If the number of meters in the potential power outage area is greater than or equal to the alarm threshold, it is judged as a serious power outage accident. If the number of meters in the potential power outage area is less than the alarm threshold, it is judged as a minor power outage accident.

[0050] Based on the determination of a serious power outage, a serious alarm is issued, the emergency response team is notified, repair personnel are dispatched quickly, and regional power dispatch is coordinated;

[0051] Based on the judgment of a minor power outage, a maintenance team was dispatched to conduct on-site confirmation and repair.

[0052] Another object of the present invention is to provide a system for monitoring power outage information based on real-time location, which comprises:

[0053] Data acquisition module, collecting real-time data of smart meters;

[0054] The critical line identification module calculates the power flow of each line and uses the LASSO algorithm for sparse regression to identify the critical lines that are strongly correlated with power outage events;

[0055] The event matrix generation module performs time window sliding processing on the key lines to generate the event matrix;

[0056] The power outage label prediction module calculates the TF-IDF value of each feature based on the time frequency of the statistical features in the event matrix, and uses the SVM model to predict the power outage label for the weighted matrix;

[0057] Minimum diagnosable clustering module, based on the similarity of fault feature vectors, corrects the SVM prediction results through the MDC clustering method;

[0058] The location clustering module uses the DBSCAN clustering algorithm to perform cluster analysis on the location data set and identify potential power outage areas;

[0059] The graded real-time alarm module triggers minor or major alarms based on the scale and severity of the power outage area.

[0060] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned power outage information monitoring method based on real-time location when executing the computer program.

[0061] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned power outage information monitoring method based on real-time location.

[0062] The beneficial effects of the present invention are as follows: by using power flow data and voltage phase angle data as feature matrices, the sparsity of LASSO regression helps to exclude non-critical line data with large noise or interference, reduce the complexity and redundancy of the model, and through the TF-IDF weighted matrix, strengthen the feature importance of critical lines in different time windows, help the SVM model to more accurately identify power outage events, and after the initial SVM classification, through MDC auxiliary identification and cluster analysis, isolated misjudgments can be corrected, and the confidence of the overall power outage event of the cluster can be improved, thereby providing decision makers with more reliable power outage information and improving the stability of power outage judgment. Through the results of DBSCAN clustering, the core area and boundary area of ​​the power outage area can be intuitively identified, so as to quickly lock the fault center. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0064] Figure 1 It is a flow chart of the method for monitoring power outage information based on real-time location of the present invention.

[0065] Figure 2 It is a structural schematic diagram of a power outage information monitoring system based on real-time location of the present invention. DETAILED DESCRIPTION

[0066] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0067] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0068] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0069] Example 1, reference Figure 1 , which is the first embodiment of the present invention, and which provides a method for monitoring power outage information based on real-time location. The method for monitoring power outage information based on real-time location includes:

[0070] S1, collects smart meter data, calculates line power flow, and performs sparse regression through the LASSO algorithm to identify key lines that are strongly correlated with power outage events;

[0071] Preferably, collecting smart meter data includes:

[0072] Collect data based on smart meters, and collect data collected by environmental sensors and fault information reported by users;

[0073] The collected data will be standardized and cleaned, including removing outliers and noise.

[0074] Data standardization can be used to standardize data from different sources (such as electricity meters, environmental sensors, and user reports) to ensure that the comparison and combination of different data sources in subsequent analysis are more accurate. This can reduce calculation errors caused by inconsistent data formats and improve the overall consistency and reliability of data. Data cleaning can improve the sensitivity and accuracy of real fault signals by removing outliers and noise and filtering out invalid data.

[0075] Furthermore, the line power flow is calculated, and sparse regression is performed using the LASSO algorithm to identify key lines that are strongly correlated with power outage events, including:

[0076] Determine the voltage phase angle data of the line and the node, and calculate the power flow based on the voltage at both ends of the line, which is expressed as:

[0077] Pij=ViVjYijcos(θi-θj);

[0078] Where Pij represents the power flow of the voltage at nodes i and j at both ends of the line, Vi and Vj represent the i-th node and j-th node at both ends of the line respectively, Yij represents the admittance of the i-th node and j-th node as the two-end line, θi and θj represent the voltage phase angles of nodes i and j respectively;

[0079] The voltage data and power flow data are used as the feature matrix X, and whether the power outage event occurs is used as the element of the target vector y. 1 is set to indicate a power outage and 0 is set to indicate a normal state. The LASSO algorithm is used for sparse regression to identify the lines that are strongly correlated with the power outage event, which is expressed as:

[0080]

[0081] Where β represents the regression coefficient, each β represents the sensitivity of the line between nodes i and j to the power outage event, λ represents the regularization parameter, which is used to control sparsity and make the model easier to identify the key lines, X represents the feature matrix, and y represents the target vector, indicating the power outage event status;

[0082] Use the training set to train the parameters of the LASSO algorithm, use the mean square error to calculate the error between the sample value and the training value, and take the value with the smallest calculated error as the optimal regularization parameter;

[0083] According to the β obtained by LASSO regression, the sum of the mean and standard deviation of the absolute value of the regression coefficient based on historical data is used as the sensitivity threshold. If β is greater than or equal to the sensitivity threshold, the line between the two end nodes represented by β is judged as a critical line;

[0084] The identified critical lines can be used to optimize PMU (Phasor Measurement Unit) deployment and fault detection response strategies, allowing for priority monitoring of lines that have a significant impact on power outages and achieving efficient allocation of resources.

[0085] After obtaining the key lines through the sparse regression of the LASSO algorithm, resources can be concentrated to prioritize monitoring of lines that are sensitive to power outages, reduce data interference from irrelevant lines, and improve the accuracy of power outage event identification, especially when resources are limited. The accuracy and efficiency of fault detection are significantly improved. After identifying the key lines, PMU and other monitoring equipment can be prioritized according to the sensitivity of these lines to achieve real-time monitoring of key nodes and lines, avoid monitoring blind spots caused by insufficient monitoring points, optimize deployment to make monitoring resource configuration more targeted, reduce the cost of PMU, and improve the coverage and effectiveness of monitoring. Using power flow data and voltage phase angle data as feature matrices, the sparsity of LASSO regression helps to exclude non-critical line data with large noise or interference, and reduce The complexity and redundancy of the model improve the stability and robustness of the identification of abnormal events, thereby reducing the misjudgment and false alarm rates. Based on the identification of key lines, fault prediction and location can be performed more efficiently, and the probability of power outages can be focused on lines that are sensitive to power outages. By focusing on key lines, the complexity and computational complexity of the prediction model are significantly reduced, and potential power outage risks can be responded to more quickly. The sensitivity analysis results provide quantitative fault impact indicators to guide operation and maintenance personnel to reasonably allocate maintenance resources and reduce resource waste after power outages. With the continuous updating of historical data, LASSO regression can dynamically update the regression coefficients and re-identify key lines, which is convenient for quickly adjusting fault detection strategies when the grid structure or load characteristics change, and realizing adaptive monitoring and flexible adjustment.

[0086] S2, performs window sliding processing based on key lines and generates an event matrix based on statistical features;

[0087] Preferably, window sliding processing is performed based on key lines, and an event matrix is ​​generated according to statistical characteristics, including:

[0088] Perform time window sliding processing based on the voltage and current data of the critical lines;

[0089] The mean, standard deviation, extreme value and slope within the window are calculated as statistical features, and the frequency of occurrence of statistical features of each key line is calculated as the time frequency;

[0090] The statistical characteristics and time frequency of each key line in each time window are summarized to generate an event matrix, where the rows represent different time windows and the columns represent the statistical characteristics (mean, standard deviation, extreme value, slope) and time frequency of each key line. The matrix contains the feature combination of each line in each time window.

[0091] Through time window sliding processing, statistical features such as mean, standard deviation, extreme value and slope of key line voltage and current are extracted in real time, which can capture small fluctuations and sudden changes in line status and improve sensitivity to faults or abnormal conditions. The combination of statistical features (such as mean, standard deviation, extreme value and slope) can effectively characterize the overall operation trend and instantaneous fluctuations of the line, thereby providing comprehensive feature support for power outage prediction and anomaly detection. The event matrix provides a full picture of the key lines at different times by summarizing the feature combinations in different time windows, and can more intuitively associate line characteristics with fault modes.

[0092] S3, based on the time frequency of the statistical features in the event matrix, calculate the TF-IDF value of each feature, predict the power outage label through the SVM model, extract the fault feature vector based on the event matrix, calculate the cosine similarity, and perform clustering based on the minimum diagnosable clustering MDC, and make corrections and judgments on the lines whose labels are predicted by the SVM model;

[0093] Preferably, based on the time frequency of the statistical features in the event matrix, the TF-IDF value of each feature is calculated, the power outage label is predicted through the SVM model, the fault feature vector is extracted based on the event matrix, and the cosine similarity is calculated, and clustering is performed based on the minimum diagnosable cluster MDC, and the line whose label is predicted by the SVM model is corrected and determined, including:

[0094] The time frequency of each statistical feature in the event matrix is ​​counted, and the TF-IDF value of each feature is calculated, which is expressed as:

[0095]

[0096] Where TF(a,d) represents the TF-IDF value of statistical feature a in window d, tf(a,d) represents the frequency of occurrence of statistical feature a in window d, N represents the total number of time windows, and df(a) represents the total number of statistical feature a in the time window;

[0097] Based on the calculation of TF(a,d), weighted processing is performed according to the event matrix to construct a weighted matrix. Each column represents the weighted value of the feature of a key line in a certain time window, and each row represents the feature combination of all lines in a time window.

[0098] Based on historical data, a label is assigned to each time window according to the power outage event record as a training set, with 1 representing a power outage event and 0 representing a non-power outage event;

[0099] The radial basis function kernel RBF is selected as the kernel function of SVM, and the SVM model is trained based on the training set. The training is performed by minimizing the classification error and maximizing the interval objective function, and the optimal parameters are determined by cross-validation.

[0100] Use the trained SVM model to predict labels for the input weight matrix;

[0101] Based on the feature combination of each key line in the event matrix within the time window, the fault feature vector of the line in the time window is extracted;

[0102] Calculate the cosine similarity between the fault feature vectors of any two lines, and use the sum of the mean and standard deviation of the historical data as the similarity threshold. If the cosine similarity is greater than or equal to the similarity threshold, it is judged that the states of the lines with corresponding similarity in the same time window are highly similar.

[0103] Clustering is performed based on the minimum diagnosable cluster MDC, traversing all line combinations of fault feature vectors, and generating multiple MDCs covering the state combinations of all lines in the same time window;

[0104] Correct the line that has been predicted by the SVM model and is judged to be in a power outage state. If other lines in the MDC cluster where the line is located are not in a power outage state, re-evaluate the state of the line.

[0105] If most of the lines in the MDC cluster where the line is located are judged to be in a power outage state by SVM, the confidence of the power outage event of the entire cluster is increased and marked as a high-confidence power outage state;

[0106] Output the final power outage status matrix, which contains the power outage status and confidence information of each line.

[0107] The TF-IDF weighted matrix strengthens the feature importance of key lines in different time windows, helping the SVM model to more accurately identify power outage events. The weighted matrix effectively captures the frequency and importance of statistical features, reduces the interference of irrelevant features on classification, and improves the accuracy of power outage detection. The SVM model using the radial basis function kernel (RBF) optimizes the classification performance by minimizing the classification error and maximizing the interval objective function. It can adapt to nonlinear and complex power outage patterns, help to more comprehensively capture the pattern changes of power outage events, and improve the ability to respond to diversified power outage events. After the initial SVM classification, isolated misjudgments can be corrected through MDC auxiliary identification and cluster analysis. For example, if a line is judged to be in a power outage state, but other lines in its MDC cluster are in normal state, the line state can be re-evaluated, thereby reducing the probability of misjudgment and improving the accuracy of power outage detection. MDC clustering By setting the cosine similarity threshold, the similarity judgment of the fault feature vector is ensured, and the highly similar line combination in the power outage event can be identified. When most lines in the same MDC are judged to be in a power outage state, the confidence of the power outage event of the cluster as a whole is improved, thereby providing decision makers with more reliable power outage information and improving the stability of power outage judgment. The TF-IDF weighted feature matrix is ​​optimized in terms of feature importance. By removing irrelevant features and emphasizing important features, the interference of redundant data is reduced, making the training and reasoning of the SVM model more efficient. At the same time, MDC clustering is used to further reduce repeated judgments and reduce the overall computational overhead, ensuring that the accuracy of power outage detection is improved without increasing the computational burden. Through the joint application of SVM and MDC, the final power outage state matrix output contains power outage state and confidence information, which not only provides power outage judgment results, but also provides the credibility of power outage events on different lines.

[0108] S4, constructing a location data set based on the MDC clustering output, performing location clustering, and performing hierarchical real-time alarm;

[0109] Preferably, a location data set is constructed based on the MDC clustering output, and location clustering is performed, including:

[0110] The final power outage status matrix output by MDC clustering is combined with the geographic location information (such as longitude and latitude) of the smart meter to construct a location data set. Each row of the data set contains the meter ID, power outage status, confidence, and geographic location.

[0111] The density-based clustering algorithm DBSCAN is selected for location clustering, and the radius parameter of the DBSCAN algorithm is determined by calculating the average neighbor distance in the geographic location data;

[0112] Based on the radius parameter, the average number of neighbors of the smart meter within the radius parameter is calculated, and the average number of neighbors of all smart meter locations is calculated to obtain the average number of neighbors of each meter within the radius parameter range;

[0113] Based on the average number of neighbors, the k-neighborhood distance of each smart meter within the radius parameter is calculated, and the k-neighborhood distances are sorted from small to large and a k-distance graph is drawn. Based on the obvious inflection point of the k-distance graph, the number of meters covered by the k-neighborhood distance corresponding to the inflection point is taken as the minimum sample number MinPts;

[0114] The DBSCAN clustering algorithm is applied. If the number of neighbors within the radius parameter of the meter location is greater than or equal to the minimum number of samples, it is marked as a core point, and a cluster is formed based on the radius parameter and the minimum number of samples with the core point as the starting point.

[0115] The DBSCAN clustering results are output to form multiple cluster groups, each of which represents a potential power outage area.

[0116] By combining the geographic location and power outage status information of smart meters, the geographic scope of the power outage area can be identified with high precision. The DBSCAN clustering algorithm aggregates power outage events according to the spatial density of the meters to form an accurate power outage area, making the power outage location more accurate and convenient for rapid response and troubleshooting. The average neighboring distance and k-neighboring distance map based on the geographic location data are used to automatically determine the radius parameter and the minimum number of samples of DBSCAN, which helps to reduce the complexity of manual parameter adjustment and improve the degree of automation of clustering. Through the results of DBSCAN clustering, the core area and boundary area of ​​the power outage area can be intuitively identified, thereby quickly locking the fault center, helping operation and maintenance personnel to identify the main affected areas more quickly, improving the efficiency of fault location, and supporting optimized maintenance and resource allocation.

[0117] Furthermore, graded real-time alarms are carried out, including:

[0118] The sum of the average number and standard deviation of historical cluster groups is used as the alarm threshold;

[0119] If the number of meters in the potential power outage area is greater than or equal to the alarm threshold, it is judged as a serious power outage accident. If the number of meters in the potential power outage area is less than the alarm threshold, it is judged as a minor power outage accident.

[0120] Based on the determination of a serious power outage, a serious alarm is issued, the emergency response team is notified, repair personnel are dispatched quickly, and regional power dispatch is coordinated;

[0121] Based on the judgment of a minor power outage, a maintenance team was dispatched to conduct on-site confirmation and repair.

[0122] Through the hierarchical judgment of minor and severe power outages, appropriate response configuration can be made for power outages of different degrees. In the case of minor power outages, only maintenance teams are dispatched to reduce the use of emergency resources; in the case of severe power outages, the emergency response team is directly mobilized to ensure that serious accidents are handled quickly, optimizing the efficiency of resource allocation. The alarm threshold is set based on the number of cluster groups and standard deviation of historical data, which can more scientifically identify power outages that are truly serious or potentially threatening, reducing the false alarm rate. Through the hierarchical judgment mechanism, unnecessary emergency resource calls can be avoided in minor power outages, reducing the frequency and cost of emergency responses.

[0123] Example 2, reference Figure 2 , which is the second embodiment of the present invention, and which is different from the previous embodiment, provides a system for a power outage information monitoring method based on real-time location, comprising:

[0124] Data acquisition module, collecting real-time data of smart meters;

[0125] The critical line identification module calculates the power flow of each line and uses the LASSO algorithm for sparse regression to identify the critical lines that are strongly correlated with power outage events;

[0126] The event matrix generation module performs time window sliding processing on the key lines to generate the event matrix;

[0127] The power outage label prediction module calculates the TF-IDF value of each feature based on the time frequency of the statistical features in the event matrix, and uses the SVM model to predict the power outage label for the weighted matrix;

[0128] Minimum diagnosable clustering module, based on the similarity of fault feature vectors, corrects the SVM prediction results through the MDC clustering method;

[0129] The location clustering module uses the DBSCAN clustering algorithm to perform cluster analysis on the location data set and identify potential power outage areas;

[0130] The graded real-time alarm module triggers minor or major alarms based on the scale and severity of the power outage area.

[0131] Example 3

[0132] This is the third embodiment of the present invention. This embodiment is different from the previous two embodiments. In order to verify the beneficial effects of the method of the present invention, the method of the present invention is compared with the traditional method through experiments for demonstration and proof.

[0133] The method of the present invention and the traditional method are respectively implemented on 4 lines, and the experimental results are formed into the following table:

[0134]

[0135] It can be observed from the table that the method of the present invention has significant advantages in analyzing line status, successfully detects potential power outage status of these lines, and significantly improves the detection sensitivity of power outage events.

[0136] Example 4

[0137] This is the fourth embodiment of the present invention. This embodiment is different from the first three embodiments. In order to verify the beneficial effects of the method of the present invention, the specific operation process of the present invention will be demonstrated through experiments.

[0138] The voltage and current data of four lines in the area are collected by smart meters and recorded once a minute for a total of 5 minutes (i.e. a time window). The data example is as follows:

[0139]

[0140] Normalize the data, calculate the line power flow and generate the feature matrix X. The data example is as follows:

[0141]

[0142] Identify critical lines through LASSO regression and set the power outage state target vector y, y = [1, 0, 1, 0] (assuming that lines 1 and 3 are out of power, and lines 2 and 4 are normal);

[0143] Through the LASSO objective function, the regression coefficients of the calculated routes are β1 = 0.8, β2 = 0.2, β3 = 0.9, and β4 = 0.1;

[0144] Based on the sensitivity threshold Line 1 and Line 3 are identified as critical lines;

[0145] Generate an event matrix based on the key lines and calculate TF-IDF, expressed as:

[0146]

[0147] Power outage label prediction based on SVM model;

[0148] Extract the feature combination of the power outage line from the event matrix and calculate the cosine similarity. For example, the feature vectors of line 1 and line 3 in window 1 are:

[0149] A = [0.25, 0.15, 0.10, 0.12];

[0150] B = [0.24, 0.12, 0.16, 0.11];

[0151] Calculate cosine similarity:

[0152]

[0153] Determine that line states are similar based on a similarity threshold;

[0154] Clustering is performed based on the fault feature vector to generate multiple MDCs covering the line status combination in the same window. If the status of other lines in the MDC is not a power outage, the fault line status is re-evaluated; if most lines are a power outage, the overall confidence is increased and the power outage status matrix is ​​output:

[0155] Line number Power outage status Confidence Remark 1 1 0.95 High confidence power outage status 2 0 0.60 No blackout, low confidence 3 1 0.92 High confidence power outage status 4 0 0.58 No blackout, low confidence

[0156] This power outage status matrix is ​​the final monitoring output and can be directly used for decision support of graded alarms and on-site disposal.

[0157] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0158] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0159] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0160] It should be understood that the various parts of the present invention can be implemented with hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented with software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented with hardware, as in another embodiment, any one of the following technologies known in the art or their combination can be used to implement: a discrete logic circuit with a logic gate circuit for implementing a logical function for a data signal, a dedicated integrated circuit with a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those of ordinary skill in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for monitoring power outage information based on real-time location, characterized in that: include, Collect smart meter data, calculate line power flow, perform sparse regression using the LASSO algorithm, and identify key lines that are strongly correlated with power outages; Perform window sliding processing based on key lines and generate event matrix based on statistical characteristics; Based on the time frequency of the statistical features in the event matrix, the TF-IDF value of each feature is calculated, and the power outage label is predicted through the SVM model. The fault feature vector is extracted based on the event matrix, and the cosine similarity is calculated. Clustering is performed based on the minimum diagnosable clustering MDC, and correction judgment is made on the lines whose labels are predicted by the SVM model; Based on the MDC clustering output, a location data set is constructed to perform location clustering and hierarchical real-time alarm.

2. The method for monitoring power outage information based on real-time location according to claim 1, characterized in that: The collecting of smart meter data includes: Collect data based on smart meters, and collect data collected by environmental sensors and fault information reported by users; The collected data will be standardized and cleaned, including removing outliers and noise.

3. The method for monitoring power outage information based on real-time location according to claim 1, characterized in that: The calculation of line power flow, sparse regression through LASSO algorithm, and identification of key lines strongly related to power outage events include: Determine the voltage phase angle data of the line and the node, and calculate the power flow based on the voltage at both ends of the line, which is expressed as: Pij=ViVjYijcos(θi-θj); Where Pij represents the power flow of the voltage at nodes i and j at both ends of the line, Vi and Vj represent the i-th node and j-th node at both ends of the line respectively, Yij represents the admittance of the i-th node and j-th node as the two-end line, θi and θj represent the voltage phase angles of nodes i and j respectively; The voltage data and power flow data are used as the feature matrix X, and whether the power outage event occurs is used as the element of the target vector y. 1 is set to indicate a power outage and 0 is set to indicate a normal state. The LASSO algorithm is used for sparse regression to identify the lines that are strongly correlated with the power outage event, which is expressed as: Where β represents the regression coefficient, each β represents the sensitivity of the line between nodes i and j to the power outage event, λ represents the regularization parameter, X represents the feature matrix, and y represents the target vector, indicating the power outage event status; According to the β obtained by LASSO regression, the sum of the mean and standard deviation of the absolute value of the regression coefficient based on historical data is used as the sensitivity threshold. If β is greater than or equal to the sensitivity threshold, the line between the two end nodes represented by β is judged as a critical line.

4. The method for monitoring power outage information based on real-time location according to claim 1, characterized in that: The window sliding processing is performed based on the key line, and the event matrix is ​​generated according to the statistical characteristics, including: Perform time window sliding processing based on the voltage and current data of the critical lines; The mean, standard deviation, extreme value and slope within the window are calculated as statistical features, and the frequency of occurrence of statistical features of each key line is calculated as the time frequency; The statistical characteristics and time frequency of each key line in each time window are summarized to generate an event matrix.

5. The method for monitoring power outage information based on real-time location according to claim 1, characterized in that: The method is based on the time frequency of the statistical features in the event matrix, calculates the TF-IDF value of each feature, predicts the power outage label through the SVM model, extracts the fault feature vector based on the event matrix, calculates the cosine similarity, and performs clustering based on the minimum diagnosable clustering MDC, and corrects and determines the line for which the label is predicted by the SVM model, including: The time frequency of each statistical feature in the event matrix is ​​counted, and the TF-IDF value of each feature is calculated, which is expressed as: Where TF(a,d) represents the TF-IDF value of statistical feature a in window d, tf(a,d) represents the frequency of occurrence of statistical feature a in window d, N represents the total number of time windows, and df(a) represents the total number of statistical feature a in the time window; Based on the calculation of TF(a,d), weighted processing is performed according to the event matrix to construct a weighted matrix; Based on historical data, a label is assigned to each time window according to the power outage event record as a training set, with 1 representing a power outage event and 0 representing a non-power outage event; The radial basis function kernel RBF is selected as the kernel function of SVM, and the SVM model is trained based on the training set. The training is performed by minimizing the classification error and maximizing the interval objective function, and the optimal parameters are determined by cross-validation. Use the trained SVM model to predict labels for the input weight matrix; Based on the feature combination of each key line in the event matrix within the time window, the fault feature vector of the line in the time window is extracted; Calculate the cosine similarity between the fault feature vectors of any two lines, and use the sum of the mean and standard deviation of the historical data as the similarity threshold. If the cosine similarity is greater than or equal to the similarity threshold, it is judged that the states of the lines with corresponding similarity in the same time window are highly similar. Clustering is performed based on the minimum diagnosable cluster MDC, traversing all line combinations of fault feature vectors, and generating multiple MDCs covering the state combinations of all lines in the same time window; Correct the line that has been predicted by the SVM model and is judged to be in a power outage state. If other lines in the MDC cluster where the line is located are not in a power outage state, re-evaluate the state of the line. If most of the lines in the MDC cluster where the line is located are judged to be in a power outage state by SVM, the confidence of the power outage event of the entire cluster is increased and marked as a high-confidence power outage state; Output the final power outage status matrix, which contains the power outage status and confidence information of each line.

6. The method for monitoring power outage information based on real-time location according to claim 5, characterized in that: The method of constructing a location data set based on the MDC clustering output and performing location clustering includes: The final power outage status matrix output by MDC clustering is combined with the geographic location information of the smart meter to construct a location data set. Each row of the data set contains the meter ID, power outage status, confidence, and geographic location. The density-based clustering algorithm DBSCAN is selected for location clustering, and the radius parameter of the DBSCAN algorithm is determined by calculating the average neighbor distance in the geographic location data; Based on the radius parameter, the average number of neighbors of the smart meter within the radius parameter is calculated, and the average number of neighbors of all smart meter locations is calculated to obtain the average number of neighbors of each meter within the radius parameter range; Based on the average number of neighbors, the k-neighborhood distance of each smart meter within the radius parameter is calculated, and the k-neighborhood distances are sorted from small to large and a k-distance graph is drawn. Based on the obvious inflection point of the k-distance graph, the number of meters covered by the k-neighborhood distance corresponding to the inflection point is taken as the minimum sample number MinPts; The DBSCAN clustering algorithm is applied. If the number of neighbors within the radius parameter of the meter location is greater than or equal to the minimum number of samples, it is marked as a core point, and a cluster is formed based on the radius parameter and the minimum number of samples with the core point as the starting point. The DBSCAN clustering results are output to form multiple cluster groups, each of which represents a potential power outage area.

7. The method for monitoring power outage information based on real-time location according to claim 1, characterized in that: The hierarchical real-time alarm includes: The sum of the average number and standard deviation of historical cluster groups is used as the alarm threshold; If the number of meters in the potential power outage area is greater than or equal to the alarm threshold, it is judged as a serious power outage accident. If the number of meters in the potential power outage area is less than the alarm threshold, it is judged as a minor power outage accident. Based on the determination of a serious power outage, a serious alarm is issued, the emergency response team is notified, repair personnel are dispatched quickly, and regional power dispatch is coordinated; Based on the judgment of a minor power outage, a maintenance team was dispatched to conduct on-site confirmation and repair.

8. A system for the real-time location-based power outage information monitoring method according to any one of claims 1 to 7, characterized in that: include, Data acquisition module, collecting real-time data of smart meters; The critical line identification module calculates the power flow of each line and uses the LASSO algorithm for sparse regression to identify the critical lines that are strongly correlated with power outage events; The event matrix generation module performs time window sliding processing on the key lines to generate the event matrix; The power outage label prediction module calculates the TF-IDF value of each feature based on the time frequency of the statistical features in the event matrix, and uses the SVM model to predict the power outage label for the weighted matrix; Minimum diagnosable clustering module, based on the similarity of fault feature vectors, corrects the SVM prediction results through the MDC clustering method; The location clustering module uses the DBSCAN clustering algorithm to perform cluster analysis on the location data set and identify potential power outage areas; The graded real-time alarm module triggers minor or major alarms based on the scale and severity of the power outage area.

9. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of the power outage information monitoring method based on real-time location described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the power outage information monitoring method based on real-time location described in any one of claims 1 to 7 are implemented.

Images