Abnormal electric equipment access alarm method and system and storage medium

Abstract

The invention provides an abnormal electric equipment access alarm method and system and a storage medium, and the method comprises the steps: obtaining initial power utilization data in a circuit, and determining a current tail end power grid state of a tail end power grid based on the initial power utilization data; cleaning the initial power consumption data based on the current terminal power grid state to obtain actual power consumption data; processing the actual power consumption data by using an incremental clustering algorithm to obtain a processing result; and based on the processing result, obtaining an identification result of whether the abnormal electric equipment is accessed, and based on the identification result, carrying out corresponding early warning. According to the invention, the access identification accuracy of the abnormal electric equipment can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of power system monitoring, and in particular to a method, system and storage medium for identifying and processing abnormal power equipment access. Background Art

[0002] With the rapid development of intelligence and automation, the power system has also entered the era of intelligence and automation. While bringing a larger scale to the power system, it also makes the power system more vulnerable to attacks. The conventional means of attack is to connect abnormal equipment at the end of the power grid. Therefore, how to identify the access of abnormal equipment in the end power grid has become an important issue in the operation and management of the power system at this stage.

[0003] At present, a fixed threshold is usually used to identify whether there are abnormal devices in the terminal power grid. Although this can identify abnormal devices, the accuracy of identification is poor, resulting in a false alarm rate of up to 30-40%. Summary of the invention

[0004] In order to improve the accuracy of identifying abnormal power-consuming device access, the embodiments of the present application provide a method, system and storage medium for identifying abnormal power-consuming device access.

[0005] In a first aspect, this embodiment provides a method for abnormal power consumption device access alarm, the method comprising: Acquire initial power consumption data in the circuit, and determine a current terminal power grid state of the terminal power grid based on the initial power consumption data; Cleaning the initial power consumption data based on the current terminal power grid state to obtain actual power consumption data; Using an incremental clustering algorithm to process the actual power consumption data to obtain a processing result; Based on the processing result, an identification result of whether there is any abnormal power-consuming device connected is obtained, and a corresponding early warning is issued based on the identification result.

[0006] In some embodiments, determining the current terminal power grid state of the terminal power grid based on the initial power consumption data includes: Obtain the most recently obtained historical terminal power grid state of the terminal power grid, determine whether the historical terminal power grid state is a steady state, and if so, obtain steady state parameters, and determine the current terminal power grid state of the terminal power grid based on the steady state parameters and initial power consumption data; If not, transient information indicating that the terminal power grid is in a transient state is obtained, and a current terminal power grid state of the terminal power grid is determined based on the transient information.

[0007] In some of these embodiments, the steady-state parameters include a first steady-state parameter or a second steady-state parameter. The first steady-state parameter includes a historical steady-state standard deviation, historical actual power consumption data, and a first adjustable coefficient. The second steady-state parameter includes a moving window mean, a steady-state mean, a steady-state standard deviation, and a second adjustable coefficient. Determining the current end-grid state of the end-grid based on the steady-state parameters and the initial power consumption data includes: If the steady-state parameter includes the first steady-state parameter, obtain the power consumption data difference between the initial power consumption data and the historical actual power consumption data, multiply the historical steady-state standard deviation by the first adjustable coefficient to obtain a first reference threshold, and determine whether the power consumption data difference is greater than the first reference threshold. If it is not greater, the current end-grid state of the end-grid is a steady state; If it is greater, the current end-grid state of the end-grid is a transient state; If the steady-state parameter includes the second steady-state parameter, obtain the mean difference between the moving window mean and the steady-state mean, multiply the steady-state standard deviation by the second adjustable coefficient to obtain a second reference threshold, and determine whether the mean difference is greater than the second reference threshold. If it is not greater, the current end-grid state of the end-grid is a steady state; If it is not greater, the current end-grid state of the end-grid is a transient state.

[0008] In some of these embodiments, cleaning the initial power consumption data based on the current end-grid state to obtain actual power consumption data includes: If the current end-grid state is a transient state, obtain a transient cleaning power consumption data group within a transient cleaning window in the past time direction starting from the initial power consumption data, where the transient cleaning power consumption data group includes the initial power consumption data; Determine the transient cleaning mean and the transient cleaning standard value of the transient cleaning power consumption data group, determine a transient threshold based on the transient cleaning mean and the transient cleaning standard value, and determine whether the initial power consumption data exceeds the transient threshold. If it exceeds, determine the transient cleaning mean as the actual power consumption data; If it does not exceed, determine the initial power consumption data as the actual power consumption data; If the current end-grid state is a steady state, obtain a steady-state cleaning power consumption data group within a steady-state cleaning window in the past time direction starting from the initial power consumption data, where the steady-state cleaning power consumption data group includes the initial power consumption data; Determine the steady-state cleaning mean and the steady-state cleaning standard value of the steady-state cleaning power consumption data group, determine a steady-state threshold based on the steady-state cleaning mean and the steady-state cleaning standard value, and determine whether the initial power consumption data exceeds the steady-state threshold. If it does not exceed, determine the initial power consumption data as the actual power consumption data; If it exceeds, update the current end - grid state to a transient state, and determine the actual power consumption data in the manner that the current end - grid state is a transient state.

[0009] In some embodiments, the using the incremental clustering algorithm to process the actual power consumption data to obtain a processing result includes: Taking the actual power consumption data as a starting point, obtain a clustered power consumption data group within a clustering window in the past direction, and perform feature extraction on the clustered power consumption data group to obtain a feature vector; Set the initialization of the incremental clustering algorithm; Calculate the distance between the feature vector and each existing cluster center; Based on the distance, determine whether the feature vector belongs to an existing cluster. If it belongs, update the cluster center of the existing cluster and output a first processing result; If it does not belong, put the feature vector into a buffer, and judge whether the number of vectors stored in the buffer exceeds a first preset number. If it does not exceed, continue to obtain the initial power consumption data in the circuit and output a second processing result; If it exceeds, judge whether the vectors in the buffer are similar. If they are similar, generate a new cluster to obtain a third processing result; If they are not similar, judge whether the number of vectors exceeds a second preset number. If it exceeds, generate a new cluster to obtain a fourth processing result; If it does not exceed, continue to obtain the initial power consumption data in the circuit and output a fifth processing result.

[0010] In some embodiments, the judging whether the vectors in the buffer are similar includes: Obtain the vector distances between pairwise vectors in the buffer, and judge whether all the vector distances are less than a preset vector distance. If so, the vectors in the buffer are similar; Otherwise, the vectors in the buffer are not similar; Or, Obtain the average value of the distances, and judge whether the average value of the distances is less than a preset average value. If so, the vectors in the buffer are similar; Otherwise, the vectors in the buffer are not similar.

[0011] In some embodiments, the obtaining the recognition result of whether there is an abnormal power - consuming device connected based on the processing result includes: If the processing result contains a new cluster, obtain the recognition result that there is an abnormal power - consuming device connected, and determine the access risk level based on the new cluster and the feature vector; If the processing result does not contain a new cluster, obtain the recognition result that there is no abnormal power - consuming device connected.

[0012] In some of these embodiments, the method further includes: After generating a new cluster, the vector used to generate the new cluster is also cleared from the buffer.

[0013] In a second aspect, the present embodiment provides a system for warning of abnormal power consumption device access. The system includes: The method further includes: After generating a new cluster, the vector used to generate the new cluster is also cleared from the buffer.

[0014] In a third aspect, the present embodiment provides a computer-readable storage medium, on which a computer program that can run on a processor is stored. When the computer program is executed by the processor, it implements a method for warning of abnormal power consumption device access as described in the first aspect.

[0015] By adopting the above method, the present application obtains the initial power consumption data in the circuit, determines the current state of the terminal power grid based on the initial power consumption data. Cleans the initial power consumption data based on the current state of the terminal power grid to obtain the actual power consumption data. Uses an incremental clustering algorithm to process the actual power consumption data to obtain a processing result. Based on the processing result, an identification result of whether there is an abnormal power consumption device access is obtained, and corresponding warnings are made based on the identification result. In this way, after preprocessing the obtained data, more accurate data can be obtained, and then the incremental clustering is further used to process the actual power consumption data, and it can be quickly and accurately determined whether there is an abnormal power consumption device access. In the case of determining that there is an abnormal power consumption device access, corresponding warning processing is performed; in the case of determining that there is no abnormal power consumption device access, device update processing is performed, thereby improving the accuracy and efficiency of identifying abnormal power consumption device access. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a block diagram of a method for warning of abnormal power consumption device access provided by an embodiment of the present application.

[0017] Figure 2 is a block diagram of a method for determining the current state of the terminal power grid based on the initial power consumption data provided by the present application.

[0018] Figure 3 is a schematic diagram of the incremental clustering processing flow provided by the present application.

[0019] Figure 4 is a schematic diagram of the process for warning of abnormal power consumption device access provided by the present application.

[0020] Figure 5 is a schematic diagram of the system connection for warning of abnormal power consumption device access provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To more clearly understand the purpose, technical solution, and advantages of this application, the following describes and explains this application in conjunction with the accompanying drawings and embodiments. However, those of ordinary skill in the art should understand that this application can be implemented without these details. For those of ordinary skill in the art, it is obvious that various changes can be made to the disclosed embodiments of this application, and without departing from the principles and scope of this application, the general principles defined in this application can be applied to other embodiments and application scenarios. Therefore, this application is not limited to the shown embodiments, but conforms to the broadest scope consistent with the scope claimed in this application.

[0022] The following further describes the embodiments of this application in detail with reference to the accompanying drawings of the specification.

[0023] Figure 1 It is a block diagram of a method for warning about the access of abnormal power-consuming devices provided by an embodiment of this application. As Figure 1 shown, a method for warning about the access of abnormal power-consuming devices includes the following steps: Step S100, obtain the initial power consumption data in the circuit, and determine the current terminal grid state of the terminal grid based on the initial power consumption data.

[0024] Step S200, clean the initial power consumption data based on the current terminal grid state to obtain the actual power consumption data.

[0025] This application is applicable to scenarios such as smart grids, industrial power consumption monitoring, and household power safety. Here, the description of the solution is from the perspective of the processing end. When the processing end has a task of identifying and processing the access of power-consuming devices, it will first obtain the initial power consumption data in the circuit. Among them, when the processing end has a task of identifying and processing the access of power-consuming devices can be determined according to the actual situation, and here no further limitation is made on how to determine the way for the processing end to have a task of identifying and processing the access of power-consuming devices.

[0026] Here, taking scenarios such as smart grids as an example for illustration, smart sensors are installed at the end of the power grid, and the processing end has a communication connection with the smart sensors. The processing end can obtain the initial power consumption data in the circuit through the smart sensors. Among them, the initial power consumption data refers to the power consumption data collected by the smart sensors. Since the collected power consumption data may be inaccurate, it is necessary to perform a certain processing on the obtained initial power consumption data, that is, cleaning, to obtain accurate power consumption data, that is, actual power consumption data.

[0027] The terminal grid is divided into two states, namely the steady state and the transient state. When the terminal grid is in different states, different processing methods are adopted for the collected initial power consumption data. Therefore, it is necessary to first determine the current terminal grid state of the terminal grid before cleaning the initial power consumption data. Figure 2It is a block diagram of a method for determining the current end - grid state of an end - grid based on initial power consumption data provided by this application. As Figure 2 shown, determining the current end - grid state of an end - grid based on initial power consumption data includes the following steps: Step S101: Obtain the historical end - grid state of the end - grid obtained most recently, and determine whether the historical end - grid state is a steady - state. If so, obtain the steady - state parameters, and determine the current end - grid state of the end - grid based on the steady - state parameters and the initial power consumption data.

[0028] Step S102: If not, obtain the transient information when the end - grid is in a transient state, and determine the current end - grid state of the end - grid based on the transient information.

[0029] The above - mentioned historical end - grid state specifically refers to the state corresponding to the most recent time of the end - grid from the current moment, and the historical end - grid state can be obtained by checking historical information. Different states correspond to different keywords. It can be determined whether there is a keyword corresponding to the steady - state in the historical end - grid state. If there is, it indicates that the historical end - grid state is a steady - state. If not, it indicates that the historical end - grid state is a transient state. The historical end - grid state is either a steady - state or a transient state.

[0030] When the historical end - grid state is a steady - state, it indicates that the end - grid was in a steady - state before the current moment. It is necessary to further obtain the steady - state parameters corresponding to the current moment, and then determine the current end - grid state of the end - grid at the current moment based on the steady - state parameters and the initial power consumption data. Among them, two methods can be used to determine the current end - grid state of the end - grid at the current moment. One is the instantaneous judgment method, and the other is the continuous judgment method. Among them, the steady - state parameters include the first steady - state parameter or the second steady - state parameter. The first steady - state parameter corresponds to the instantaneous judgment method, and the second steady - state parameter corresponds to the continuous judgment method. The first steady - state parameter includes the historical steady - state standard deviation, the historical actual power consumption data, and the first adjustable coefficient. The second problem parameter includes the moving - window mean, the steady - state mean, the steady - state standard deviation, and the second adjustable coefficient. Among them, determining the current end - grid state of the end - grid based on the steady - state parameters and the initial power consumption data includes the following steps: Step S101 - 1: If the steady - state parameters include the first steady - state parameter, obtain the power consumption data difference between the initial power consumption data and the historical actual power consumption data, multiply the historical steady - state standard deviation by the first adjustable coefficient to obtain the first reference threshold, and determine whether the power consumption data difference is greater than the first reference threshold. If not, the current end - grid state of the end - grid is a steady - state.

[0031] Step S101 - 2: If it is greater, the current end - grid state of the end - grid is a transient state.

[0032] Step S101-3, if the steady-state parameter includes the second steady-state parameter, obtain the mean difference between the moving window mean and the steady-state mean, multiply the steady-state standard deviation by the second adjustable coefficient to obtain the second reference threshold, and determine whether the mean difference is greater than the second reference threshold. If not, the current terminal grid state of the terminal grid is the steady state.

[0033] Step S101-4, if it is greater, the current terminal grid state of the terminal grid is the transient state.

[0034] For determining the current terminal grid state of the terminal grid at the current moment by using the instantaneous judgment method, the historical steady-state standard deviation refers to the standard deviation corresponding to the historical terminal grid state, and the historical actual power consumption data refers to the actual power consumption data corresponding to the historical terminal grid state. The historical steady-state standard deviation and the historical actual power consumption data are both stored at the processing end, and the historical steady-state standard deviation and the historical actual power consumption data can be obtained by viewing the information stored at the processing end. The first adjustable coefficient is determined by the staff according to the actual situation and is pre-stored at the processing end, and the first adjustable coefficient can be obtained by viewing the information stored at the processing end.

[0035] The judgment condition of the instantaneous judgment method is whether the change amount of adjacent current sampling points exceeds the threshold. The calculation formula is =| - |> , where is the current change amount of two adjacent sampling points, that is, the power consumption data difference; is the current value of the current at the current sampling moment, that is, the initial power consumption data; is the current value of the current at the previous sampling moment, that is, the actual power consumption data corresponding to the historical terminal grid state, is the historical steady-state standard deviation, is the first adjustable coefficient, which can be determined according to the actual situation. In this application, it is preferably selected as 3. is the first reference threshold. By substituting the relevant data into the above calculation formula to determine whether the calculation formula holds, that is, whether the power consumption data difference is greater than the first reference threshold. If not, the current terminal grid state of the terminal grid is still the steady state. If it is greater, the current terminal grid state of the terminal grid is the transient state.

[0036] For determining the current terminal grid state of the terminal grid at the current moment by using a continuous judgment method, the moving window mean refers to the average value of the actual power consumption data or initial power consumption data collected within a continuous time region with the current moment as the latest time. Among them, only the current moment corresponds to the initial power consumption data, and the other moments within the moving window correspond to the actual power consumption data. The steady-state mean refers to the average value obtained by replacing the earliest power consumption data among the multiple power consumption data corresponding to the historical terminal grid state with the initial power consumption data. The steady-state standard deviation refers to the standard value obtained by replacing the earliest power consumption data among the multiple power consumption data corresponding to the historical terminal grid state with the initial power consumption data. The second adjustable coefficient is determined by the staff according to the actual situation and is pre-stored in the processing end. The second adjustable coefficient can be obtained by viewing the information stored in the processing end.

[0037] The judgment condition of the continuous judgment method is whether the mean deviation between the moving window mean and the steady-state mean exceeds the threshold. The calculation formula is > , where is the mean deviation between the moving window mean and the steady-state mean, is the moving window mean, is the steady-state mean, is the steady-state standard deviation, is the second adjustable coefficient, which can be determined according to the actual situation. In this application, is preferably selected as 2.5. By substituting the relevant data into the above calculation formula to judge whether the calculation formula holds, that is, whether the mean deviation is greater than the second reference threshold. If not, the current terminal grid state of the terminal grid remains the steady-state. If it is greater, the current terminal grid state of the terminal grid is the transient state. In this way, when the historical terminal grid state is the steady-state, by using multiple methods to determine the current terminal grid state of the terminal grid at the current moment, the efficiency of determining the current terminal grid state of the terminal grid at the current moment can be improved.

[0038] When the grid state at the end is a transient state at the end of history, it indicates that the end grid has been in a transient state before the current moment. It is necessary to further obtain the transient information of the end grid in the transient state, and determine the current end grid state of the end grid based on the transient information. Among them, the transient information includes the time that the end grid has continuously been in the transient state this time and whether there are new anomalies during this period. By comparing the time that the end grid has continuously been in the transient state this time with the preset time, if the time that the end grid has continuously been in the transient state this time is less than the preset time, the current end grid state of the end grid is still the transient state; if the time that the end grid has continuously been in the transient state this time is not less than the preset time and there are no new anomalies, the current end grid state of the end grid changes to the steady state. Among them, a new anomaly refers to a situation where the difference in electricity consumption data is greater than the first reference threshold or the mean deviation is greater than the second reference threshold during the time that the end grid has continuously been in the transient state this time.

[0039] After determining the current end grid state of the end grid, the initial electricity consumption data is cleaned based on the current end grid state to obtain the actual electricity consumption data. Among them, cleaning the initial electricity consumption data based on the current end grid state to obtain the actual electricity consumption data includes the following steps: Step S201, if the current end grid state is a transient state, obtain the transient cleaning electricity consumption data group within the transient cleaning window in the past time direction starting from the initial electricity consumption data. Among them, the transient cleaning electricity consumption data group contains the initial electricity consumption data.

[0040] Step S202, determine the transient cleaning mean and the transient cleaning standard value of the transient cleaning electricity consumption data group, determine the transient threshold based on the transient cleaning mean and the transient cleaning standard value, and judge whether the initial electricity consumption data exceeds the transient threshold. If it exceeds, determine the transient cleaning mean as the actual electricity consumption data.

[0041] Step S203, if it does not exceed, determine the initial electricity consumption data as the actual electricity consumption data.

[0042] Step S204, if the current end grid state is a steady state, obtain the steady state cleaning electricity consumption data group within the steady state cleaning window in the past time direction starting from the initial electricity consumption data. Among them, the steady state cleaning electricity consumption data group contains the initial electricity consumption data.

[0043] Step S205, determine the problem cleaning mean and the steady state cleaning standard value of the steady state request data group, determine the steady state threshold based on the steady state cleaning mean and the steady state cleaning standard value, and judge whether the initial electricity consumption data exceeds the steady state threshold. If it does not exceed, determine the initial electricity consumption data as the actual electricity consumption data.

[0044] Step S206, if it exceeds, update the current end grid state to the transient state, and determine the actual electricity consumption data in the same way as when the current grid state is the transient state.

[0045] When the current terminal power grid state is in a transient state, the cleaning operation for the initial power consumption data is to start a transient cleaning window for dynamic processing. By examining the actual power consumption data falling within this transient cleaning window at historical moments and the initial power consumption data at the current moment, a transient cleaning power consumption data group within the transient cleaning window is obtained by retrieving data in the past time direction starting from the initial power consumption data. This transient cleaning power consumption data group includes the initial power consumption data and the actual power consumption data corresponding to other historical moments. Then, the average value of the power consumption data within the transient cleaning power consumption data group is calculated to obtain the transient cleaning average value, and the power consumption data within the transient cleaning power consumption data group is substituted into the standard deviation formula to obtain the transient cleaning standard value. Subsequently, the transient cleaning average value The range obtained from the transient cleaning standard value is determined as the transient threshold.

[0046] If the initial power consumption data obtained in step S100 above falls within this transient threshold, it indicates that the initial power consumption data is accurate. At this time, the actual power consumption data after cleaning the initial power consumption data is the initial power consumption data. If the initial power consumption data obtained in step S100 above does not fall within this transient threshold, it indicates that the initial power consumption data is inaccurate. At this time, the actual power consumption data after cleaning the initial power consumption data is the transient cleaning average value. After cleaning the initial power consumption data, more accurate power consumption data is obtained. After all, in the case of inaccurate initial power consumption data, using the average value can, to a certain extent, eliminate the detection error and improve the accuracy of the data, thereby improving the accuracy of subsequent identification of abnormal power-consuming equipment access.

[0047] When the current terminal power grid state is in a steady state, the cleaning operation for the initial power consumption data is to start a steady-state cleaning window for dynamic processing. By examining the actual power consumption data falling within this steady-state cleaning window at historical moments and the initial power consumption data at the current moment, a steady-state cleaning power consumption data group within the steady-state cleaning window is obtained by retrieving data in the past time direction starting from the initial power consumption data. This steady-state cleaning power consumption data group includes the initial power consumption data and the actual power consumption data corresponding to other historical moments. Then, the average value of the power consumption data within the steady-state cleaning power consumption data group is calculated to obtain the steady-state cleaning average value, and the power consumption data within the steady-state cleaning power consumption data group is substituted into the standard deviation formula to obtain the steady-state cleaning standard value. Subsequently, the steady-state cleaning average value The range obtained from the transient cleaning standard value is determined as the steady-state threshold.

[0048] If the initial power consumption data obtained in the above step S100 falls within the steady-state threshold, it indicates that the initial power consumption data is accurate. At this time, the actual power consumption data after cleaning the initial power consumption data is the initial power consumption data. If the initial power consumption data obtained in the above step S100 does not fall within the steady-state threshold, it indicates that the terminal state is abnormal. At this time, update the current terminal grid state from the steady state to the transient state, and use the steps in the above step S201-step S203 to clean and process the initial power consumption data to obtain the actual power consumption data after cleaning the initial power consumption data, which will not be elaborated here. In this way, when cleaning and processing the initial power consumption data, the current terminal grid state of the terminal grid can also be corrected, so as to adjust the operation of cleaning and processing the initial power consumption data, which can improve the accuracy of the actual power consumption data obtained after cleaning and processing the initial power consumption data, and further improve the accuracy of identifying the access of abnormal power consumption equipment in the follow-up.

[0049] Step S300, use the incremental clustering algorithm to process the actual power consumption data to obtain the processing result.

[0050] Step S400, based on the processing result, obtain the identification result of whether there is abnormal power consumption equipment access, and issue corresponding warnings based on the identification result.

[0051] After obtaining the actual power consumption data, the identification process for abnormal power consumption equipment access is carried out, that is, use the incremental clustering algorithm to process the actual power consumption data to obtain the processing result. Among them, using the incremental clustering algorithm to process the actual power consumption data to obtain the processing result includes the following steps: Step S301, starting from the actual power consumption data, obtain the clustering power consumption data group within the clustering window in the past direction, and extract features from the clustering power consumption data group to obtain the feature vector.

[0052] Step S302, set the initialization of the incremental clustering algorithm.

[0053] Step S303, calculate the distance between the feature vector and the center of each existing cluster.

[0054] Step S304, based on the distance, determine whether the feature vector belongs to an existing cluster. If it belongs, update the cluster center of the existing cluster and output the first processing result.

[0055] Step S305, if it does not belong, put the feature vector into the buffer area, and judge whether the number of vectors stored in the buffer area exceeds the first preset number. If it does not exceed, continue to obtain the initial power consumption data in the circuit and output the second processing result.

[0056] Step S306, if it exceeds, judge whether the vectors in the buffer area are similar. If they are similar, generate a new cluster to obtain the third processing result.

[0057] In step S307, if they are not similar, determine whether the number of vectors exceeds a second preset number. If it exceeds, generate a new cluster to obtain a fourth processing result.

[0058] In step S308, if it does not exceed, continue to obtain the initial power consumption data in the circuit and output a fifth processing result.

[0059] This application uses an incremental clustering algorithm, that is, the incremental DBSCAN (Incremental DBSCAN) algorithm combined with a clustering window to implement incremental clustering analysis. This clustering window is a sliding window. The actual power consumption data obtained in the above step S200 is divided according to a fixed clustering window, and a feature vector is extracted from each clustering window. There is overlap between the clustering windows. For example, the clustering window is 500 ms and it slides once every 50 ms to achieve real-time clustering, and at the same time determine whether it belongs to a known device cluster or form a new device cluster result.

[0060] Specifically, for the first step of feature extraction, the clustered power consumption data group refers to the actual power consumption data collected within a continuous time region with the current moment as the latest time. Among them, the size of the continuous time region is the size of the clustering window. Then, feature vector extraction is performed to obtain a feature vector , where represents the effective value of the current, represents the waveform sharpness, represents the waveform distortion degree, represents the current harmonic distortion rate, represents the maximum current change rate, represents the zero-crossing distortion. Different feature vectors represent different device types.

[0061] Subsequently, the obtained feature vectors are subjected to clustering analysis using a clustering algorithm. The clustering algorithm calculates the distance between it and the existing clustering centers. If the distance is within the threshold, update the center of the cluster; otherwise, put it into the buffer. When the buffer accumulates a sufficient number of similar new feature vectors, a new cluster is formed. When the number of clusters exceeds the upper limit, merge the two closest clusters. For each feature vector, the clustering algorithm will output a processing result, either belonging to an existing cluster, indicating a known device; or belonging to a new cluster, indicating a new device.

[0062] The specific approach is as follows. First, initialize the incremental clustering algorithm, set the distance threshold, buffer, set of cluster centers, and maximum number of clusters.

[0063] Then calculate the distance between the new feature vector and the existing clusters. For each new feature vector, calculate its distance from all existing cluster centers, and the Euler distance calculation formula can be used for calculation.

[0064] Next, it is determined whether the feature vector belongs to an existing cluster. That is, the minimum distance between the feature vector and the center of the existing cluster is found, and it is judged whether this minimum distance is less than the distance threshold set in the above initialization. If it is less, it indicates that the feature vector belongs to the existing cluster, that is, it belongs to the existing cluster corresponding to the minimum distance, and the cluster center of this existing cluster is updated, and a first processing result indicating that no new cluster is included is output.

[0065] If it is not less, it indicates that the feature vector does not belong to the existing cluster, and this feature vector is put into the buffer. Next, it is checked whether the number of vectors placed in the buffer exceeds a first preset number. Here, the first preset number is set according to actual needs. Preferably, the first preset number is set to the value five in this application. If the number of vectors does not exceed the first preset vector, the initial power consumption data in the circuit is continuously obtained subsequently, and a second processing result indicating that no new cluster is included is output.

[0066] If the number of vectors exceeds the first preset vector, it is subsequently judged whether these vectors in the buffer are similar. If they are similar, in order to relieve the pressure of storing vectors in the buffer and also to be able to quickly identify the access of abnormal power-consuming devices, these vectors in the buffer are used to generate a new cluster, and a new cluster center is created with the mean value of these vectors in the buffer, and a third processing result indicating that a new cluster is included is output. In addition, after generating the new cluster, the vectors used to generate the new cluster are cleared from the buffer.

[0067] At the same time, it is also judged whether the number of vectors in the buffer exceeds a second preset number. If the number of vectors exceeds the second preset number, the distances between all cluster centers can be calculated pairwise, the two clusters with the smallest distance are found, they are merged to generate a new cluster, and the original clusters are deleted at the same time, and a fourth processing result indicating that a new cluster is included is generated. If the number of vectors does not exceed the second preset number, the initial power consumption data in the circuit is continuously obtained, and a fifth processing result indicating that a new cluster is included is output.

[0068] Among them, judging whether the vectors in the buffer are similar includes the following steps: Step S30, obtain the vector distances between pairwise vectors in the buffer, and judge whether the vector distances are all less than the preset vector distance. If so, the vectors in the buffer are similar.

[0069] Step S31, otherwise, the vectors in the buffer are not similar.

[0070] Step S32, obtain the average value of the vector distances, and judge whether the average value is less than the preset mean value. If so, the vectors in the buffer are similar.

[0071] Step S33, otherwise, the vectors in the buffer are not similar.

[0072] Specifically, the Euler distance formula can be used to calculate the distances between them pairwise. If all the distances are less than a preset vector distance, or their average distance is less than a preset mean value, they are considered similar; otherwise, they are considered dissimilar.

[0073] Figure 3 is a schematic diagram of the incremental clustering processing flow provided by this application. As Figure 3 shown, first, a new feature vector is obtained, then the distances between this feature vector and all cluster centers are calculated. Next, based on the distance values, the nearest cluster to this feature vector is found, and then it is determined whether the distance between the feature vector and this nearest cluster is less than the initialized distance threshold. If it is less, the cluster center of the existing cluster is updated. If it is not less, this feature vector is added to the buffer, and it is judged whether the number of vectors in the buffer meets the first preset quantity. If it does not meet, the initial power consumption data in the circuit is continuously obtained; if it meets, it is judged whether the vectors in the buffer are similar. If they are not similar, the initial power consumption data in the circuit is also continuously obtained; if they are similar, a new cluster is created. Subsequently, it is judged whether the number of clusters in the buffer exceeds the second preset quantity. If it exceeds, the two closest clusters are merged. If it does not exceed, the initial power consumption data in the circuit is continuously obtained. In this way, by using the accurate power consumption data obtained above and further using incremental clustering to process the actual power consumption data, the processing result can be obtained quickly and accurately.

[0074] After obtaining the processing result, based on the processing result, an identification result of whether there is an abnormal power consumption device connected is obtained. Among them, obtaining an identification result of whether there is an abnormal power consumption device connected based on the processing result includes the following steps: Step S401, if the processing result contains a new cluster, an identification result of having an abnormal power consumption device connected is obtained, and the access risk level is determined based on the new cluster and the feature vector.

[0075] Step S402, if the processing result does not contain a new cluster, an identification result of having no abnormal power consumption device connected is obtained.

[0076] For the identification result that there is an abnormal power consumption device connected indicated by the processing result containing a new cluster, the access risk level of the abnormal device connection can be further determined based on the new cluster and the feature vector. Specifically as follows, If , then the risk level is level one, low risk, and notifications and log warnings are made.

[0077] If > 15% or abnormal or > 10%, then the risk level is level two, medium risk, and audible and visual warnings and data log warnings are made.

[0078] If > 25% or > 100A / ms, then the risk level is level three, high risk, and remote power-off and emergency notification warning are carried out.

[0079] Figure 4 is the schematic flow diagram of the warning for abnormal power-consuming equipment access provided by this application. As Figure 4 shown, first, preprocess the obtained data, that is, preprocess the data. Then extract features from the preprocessed data to obtain feature vectors. Next, perform incremental clustering processing on the feature vectors to detect whether new clusters are generated. If so, perform risk assessment, warning decision-making, and warning triggering in sequence. If not, perform device matching and update the device status in sequence. In this way, by preprocessing the obtained data, more accurate data can be obtained. Then, further use incremental clustering to process the actual power consumption data, and it can be quickly and accurately determined whether there is abnormal power-consuming equipment access. In the case of determining that there is abnormal power-consuming equipment access, corresponding warning processing is carried out; in the case of determining that there is no abnormal power-consuming equipment access, device update processing is carried out, so as to improve the accuracy and efficiency of identifying abnormal power-consuming equipment access.

[0080] Figure 5 is the schematic connection diagram of a system for warning abnormal power-consuming equipment access provided by this application. As Figure 5 shown, a system for warning abnormal power-consuming equipment access includes: a status confirmation module, a data cleaning module, and a warning processing module.

[0081] Among them, the status confirmation module is used to obtain the initial power consumption data in the circuit and determine the current end-grid status of the end grid based on the initial power consumption data. The data cleaning module is used to clean the initial power consumption data based on the current end-grid status to obtain the actual power consumption data. The warning processing module is used to process the actual power consumption data using an incremental clustering algorithm to obtain a processing result; based on the processing result, obtain an identification result of whether there is abnormal power-consuming equipment access, and perform corresponding warnings based on the identification result.

[0082] The other functions performed by the above status confirmation module, data cleaning module, and warning processing module, as well as the technical details of each function, are the same as or similar to the corresponding features in the method for warning abnormal power-consuming equipment access described above, so they will not be repeated here.

[0083] The embodiment of this application also provides a computer storage medium, on which a computer program is stored. When it runs on a computer, it enables the computer to execute the steps in the method for warning abnormal power-consuming equipment access described above.

[0084] It should be understood that although the steps in the flowchart of the accompanying drawings are shown sequentially in the direction of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless specifically stated herein, there is no strict order restriction for the execution of these steps, and they can be executed in other orders.

[0085] The above are only some embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for warning about the access of abnormal power-consuming equipment, characterized in that, The method includes: Obtaining initial power consumption data in a circuit, and determining the current end-grid state of the end grid based on the initial power consumption data; Cleaning the initial power consumption data based on the current end-grid state to obtain actual power consumption data; Processing the actual power consumption data using an incremental clustering algorithm to obtain a processing result; Obtaining an identification result of whether there is an abnormal power-consuming device connected based on the processing result, and performing corresponding early warnings based on the identification result; Wherein, the determining the current end-grid state of the end grid based on the initial power consumption data includes: Obtaining the historical end-grid state of the end grid obtained most recently, determining whether the historical end-grid state is a steady state. If so, obtaining steady-state parameters, and determining the current end-grid state of the end grid based on the steady-state parameters and the initial power consumption data; If not, obtaining transient information when the end grid is in a transient state, and determining the current end-grid state of the end grid based on the transient information.

2. The method according to claim 1, wherein The steady-state parameters include a first steady-state parameter or a second steady-state parameter. The first steady-state parameter includes a historical steady-state standard deviation, historical actual power consumption data, and a first adjustable coefficient. The second steady-state parameter includes a moving window mean, a steady-state mean, a steady-state standard deviation, and a second adjustable coefficient. The determining the current end-grid state of the end grid based on the steady-state parameters and the initial power consumption data includes: If the steady-state parameter includes the first steady-state parameter, obtaining the power consumption data difference between the initial power consumption data and the historical actual power consumption data, multiplying the historical steady-state standard deviation by the first adjustable coefficient to obtain a first reference threshold, and determining whether the power consumption data difference is greater than the first reference threshold. If not, the current end-grid state of the end grid is a steady state; If it is greater, the current end-grid state of the end grid is a transient state; If the steady-state parameter includes the second steady-state parameter, obtaining the mean difference between the moving window mean and the steady-state mean, multiplying the steady-state standard deviation by the second adjustable coefficient to obtain a second reference threshold, and determining whether the mean difference is greater than the second reference threshold. If not, the current end-grid state of the end grid is a steady state; If it is not greater, the current end-grid state of the end grid is a transient state.

3. The method according to claim 2, wherein The cleaning the initial power consumption data based on the current end-grid state to obtain actual power consumption data includes: If the current end-grid state is a transient state, obtaining a transient cleaning power consumption data group within a transient cleaning window in the past time direction starting from the initial power consumption data, where the transient cleaning power consumption data group includes the initial power consumption data; Determining the transient cleaning mean and the transient cleaning standard value of the transient cleaning power consumption data group, determining a transient threshold based on the transient cleaning mean and the transient cleaning standard value, and determining whether the initial power consumption data exceeds the transient threshold. If it exceeds, determining the transient cleaning mean as the actual power consumption data; If it does not exceed, determining the initial power consumption data as the actual power consumption data; If the current end - grid state is a steady - state, obtain a steady - state cleaning power - consumption data set within a steady - state cleaning window in the past - time direction starting from the initial power - consumption data, where the steady - state cleaning power - consumption data set includes the initial power - consumption data; Determine the steady - state cleaning mean and the steady - state cleaning standard value of the steady - state cleaning power - consumption data set, determine a steady - state threshold based on the steady - state cleaning mean and the steady - state cleaning standard value, and judge whether the initial power - consumption data exceeds the steady - state threshold. If it does not exceed, determine the initial power - consumption data as the actual power - consumption data; If it exceeds, update the current end - grid state to a transient state, and determine the actual power - consumption data in the manner that the current end - grid state is a transient state.

4. The method according to claim 1, wherein The process of using the incremental clustering algorithm to process the actual power - consumption data to obtain a processing result includes: Obtain a clustering power - consumption data set within a clustering window in the past - time direction starting from the actual power - consumption data, and perform feature extraction on the clustering power - consumption data set to obtain a feature vector; Set the initialization of the incremental clustering algorithm; Calculate the distance between the feature vector and each existing cluster center; Based on the distance, judge whether the feature vector belongs to an existing cluster. If it belongs, update the cluster center of the existing cluster and output a first processing result; If it does not belong, put the feature vector into a buffer, judge whether the number of vectors stored in the buffer exceeds a first preset number. If it does not exceed, continue to obtain the initial power - consumption data in the circuit and output a second processing result; If it exceeds, judge whether the vectors in the buffer are similar. If they are similar, generate a new cluster to obtain a third processing result; If they are not similar, judge whether the number of vectors exceeds a second preset number. If it exceeds, generate a new cluster to obtain a fourth processing result; If it does not exceed, continue to obtain the initial power - consumption data in the circuit and output a fifth processing result.

5. The method according to claim 4, wherein The judgment of whether the vectors in the buffer are similar includes: Obtain the vector distances between pairwise vectors in the buffer, and judge whether all the vector distances are less than a preset vector distance. If so, the vectors in the buffer are similar; Otherwise, the vectors in the buffer are not similar; Or, Obtain the average value of the vector distances, and judge whether the average value is less than a preset average value. If so, the vectors in the buffer are similar; Otherwise, the vectors in the buffer are not similar.

6. The method according to claim 4, wherein The process of obtaining an identification result of whether there is an abnormal power - consumption device connected based on the processing result includes: If the processing result contains a new cluster, obtain an identification result that there is an abnormal power - consumption device connected, and determine the access risk level based on the new cluster and the feature vector; If the processing result does not contain a new cluster, obtain an identification result that there is no abnormal power - consumption device connected.

7. The method according to claim 4, characterized in that The method further includes: After generating a new cluster, also clear the vectors used to generate the new cluster from the buffer.

8. A system for warning of abnormal power-consuming equipment access, characterized in that The system includes: a state confirmation module, a data cleaning module, and an alarm processing module; where, The state confirmation module is used to obtain the initial power - consumption data in the circuit and determine the current end - grid state of the end - grid based on the initial power - consumption data; The data cleaning module is used to clean the initial power consumption data based on the current end - grid state to obtain the actual power consumption data; The alarm processing module is used to process the actual power consumption data using an incremental clustering algorithm to obtain a processing result; based on the processing result, an identification result of whether there is an abnormal power - consuming device connected is obtained, and corresponding warnings are issued based on the identification result; Among them, determining the current end - grid state of the end - grid based on the initial power consumption data includes: Obtaining the historical end - grid state of the end - grid obtained most recently, determining whether the historical end - grid state is a steady - state, if so, obtaining steady - state parameters, and determining the current end - grid state of the end - grid based on the steady - state parameters and the initial power consumption data; If not, obtaining transient information when the end - grid is in a transient state, and determining the current end - grid state of the end - grid based on the transient information.

9. A computer-readable storage medium having stored thereon a computer program that can run on a processor, characterized in that, When the computer program is executed by the processor, it implements a method for warning of abnormal power - consuming device access as described in any one of claims 1 to 7.

Images