Power terminal log fault analysis and prediction method and system based on deep learning

Abstract

The present invention provides a method and system for analyzing and predicting power terminal log faults based on deep learning. When a power terminal self-checks and finds abnormal operating status, the power terminal sends an abnormal operating status instruction through a corresponding gateway; a remote control center performs a legitimacy verification based on the power terminal identification number and the power terminal key in the abnormal operating status instruction. When the verification is successful, the remote control center feeds back the log information reporting instruction to the power terminal through the corresponding gateway, and does not process further when the verification fails; the power terminal transmits the power terminal log information at this time to the remote control center through the corresponding gateway; the remote control center preprocesses the power terminal log information, extracts N feature data from the preprocessed power terminal log information, and inputs the N feature data corresponding to the power terminal into a trained deep learning algorithm model to perform power fault prediction to output and display the specific power fault type of the power terminal and a first-level warning.

Description

Technical Field

[0001] The present invention relates to the technical field of power terminal log fault analysis, and in particular to a power terminal log fault analysis and prediction method and system based on deep learning. Background Art

[0002] With the construction and development of smart grids, the application of power system automation is becoming increasingly widespread. Existing smart grids have initially automated the collection of power information, and metering systems can now utilize concentrators for remote meter reading. However, the potential for problems at power terminals is complex and diverse, making it a significant challenge for power companies to promptly identify and troubleshoot issues within the vast amount of power logs.

[0003] Currently, some power terminal log fault analysis relies primarily on manual screening. When a fault occurs at an on-site power terminal, staff must visit the terminal to copy the relevant log information. A team of experienced power experts then promptly and effectively investigates the cause of the problem to ensure the normal operation of the power system. However, this method is time-consuming and inefficient. Manual inspections cannot guarantee that anomalies will be detected within the specified timeframe, and sometimes oversight can lead to serious consequences. Alternatively, power terminals transmit log information in real time to a remote control center. The remote control center processes each transmitted log to identify any faults. This implementation requires processing large amounts of data, resulting in slow processing and memory consumption. Summary of the Invention

[0004] In response to the problems and shortcomings of the prior art, the present invention provides a method and system for analyzing and predicting power terminal log faults based on deep learning.

[0005] The present invention solves the above technical problems through the following technical solutions:

[0006] The present invention provides a method for analyzing and predicting power terminal log faults based on deep learning, which is characterized in that it includes the following steps:

[0007] S1. Each power terminal in the set area self-checks to see if there is any abnormal operating behavior. If one of the power terminals self-checks and finds that there is any abnormal operating behavior, the process proceeds to step S2. If the power terminal self-checks and finds that there is no abnormal operating behavior, the process repeats step S1.

[0008] S2. The power terminal sends an abnormal operation status instruction to the remote control center through the corresponding gateway. The abnormal operation status instruction includes an abnormal timestamp, an abnormal operation status identifier, a power terminal identification number, a power terminal key, and a gateway identification number. The correspondence between the power terminal and the gateway is a one-to-one correspondence or a many-to-one correspondence;

[0009] S3, the remote control center performs a legitimacy check based on the power terminal identification number and the power terminal key in the abnormal operation status instruction, and proceeds to step S4 if the check succeeds, or proceeds to step S5 if the check fails;

[0010] S4. The remote control center feeds back a log information reporting instruction to the power terminal through the corresponding gateway, and enters step S6. The log information reporting instruction includes the abnormal timestamp, the log information reporting identifier, and the power terminal identification number;

[0011] S5. The remote control center does not perform any further processing;

[0012] S6. The power terminal transmits the power terminal log information formed by the abnormal timestamp node to the remote control center through the corresponding gateway;

[0013] S7, the remote control center preprocesses the power terminal log information to obtain preprocessed power terminal log information;

[0014] S8. The remote control center extracts key data from the pre-processed power terminal log information according to a preset extraction rule to obtain N feature data, where N is a positive integer;

[0015] S9. The remote control center inputs N feature data corresponding to the power terminal into the corresponding trained deep learning algorithm model to perform power fault prediction to output the specific power fault type of the power terminal, which is a line fault, transformer fault, bus fault or other fault;

[0016] S10. The remote control center displays the specific power fault type of the power terminal and issues a first-level alarm prompt;

[0017] S11. The remote control center counts and displays the number of times each type of power fault occurs at the power terminal within a set time period. When the number of a certain type of power fault at the power terminal reaches the set number, a secondary alarm is issued and relevant responsible personnel are notified.

[0018] Preferably, the following steps are included before step S1:

[0019] When a power terminal among the power terminals self-checks and finds abnormal operation status, the power terminal log information formed by the abnormal timestamp node is directly transmitted to the remote control center through the corresponding gateway;

[0020] The remote control center preprocesses the power terminal log information to obtain preprocessed power terminal log information;

[0021] The remote control center extracts key data from the pre-processed power terminal log information according to preset extraction rules to obtain N feature data;

[0022] The remote control center displays N feature data for the staff to mark the power fault type of the power terminal based on the N feature data. The N feature data and the corresponding power fault type are grouped together, and multiple sets of training samples are obtained based on this operation;

[0023] The remote control center inputs multiple sets of training samples into the deep learning algorithm model for training to obtain a trained deep learning algorithm model, where each N feature data is used as the model input and each power fault type is used as the model output.

[0024] The present invention also provides a power terminal log fault analysis and prediction system based on deep learning, which is characterized in that it includes multiple power terminals, multiple gateways and a remote control center, and the corresponding relationship between the power terminals and the gateways is a one-to-one correspondence or a many-to-one correspondence;

[0025] Each power terminal in the set area is used to self-check whether it has abnormal operating status. If a power terminal in the power terminal self-checks and finds abnormal operating status, the power terminal sends an abnormal operating status instruction to the remote control center through the corresponding gateway. If the power terminal self-checks and finds no abnormal operating status, the operation is repeated. The abnormal operating status instruction includes an abnormal timestamp, an abnormal operating status identifier, a power terminal identification number, a power terminal key, and a gateway identification number.

[0026] The remote control center is used to perform a legitimacy check based on the power terminal identification number and the power terminal key in the operation status abnormality instruction. When the verification is successful, the remote control center feeds back a log information reporting instruction to the power terminal through the corresponding gateway. When the verification fails, the remote control center does not perform further processing. The log information reporting instruction includes an abnormal timestamp, a log information reporting identifier, and the power terminal identification number;

[0027] The power terminal is used to transmit the power terminal log information formed by the abnormal timestamp node to the remote control center through the corresponding gateway;

[0028] The remote control center is used to preprocess the power terminal log information to obtain preprocessed power terminal log information;

[0029] The remote control center is used to extract key data from the pre-processed power terminal log information according to preset extraction rules to obtain N feature data, where N is a positive integer;

[0030] The remote control center is used to input N feature data corresponding to the power terminal into the corresponding trained deep learning algorithm model to perform power fault prediction and output the specific power fault type of the power terminal, which is a line fault, transformer fault, bus fault or other fault;

[0031] The remote control center is used to display the specific power fault type of the power terminal and issue a first-level alarm prompt;

[0032] The remote control center is used to count and display the number of times each type of power fault occurs at the power terminal within a set time period. When the number of a certain type of power fault at the power terminal reaches the set number, a secondary alarm prompt is issued and the relevant responsible personnel are notified.

[0033] Preferably, when a certain power terminal among the power terminals is used for self-checking and abnormal operation behavior occurs, the power terminal log information formed by the abnormal timestamp node is directly transmitted to the remote control center through the corresponding gateway;

[0034] The remote control center is used to preprocess the power terminal log information to obtain preprocessed power terminal log information;

[0035] The remote control center is used to extract key data from the pre-processed power terminal log information according to preset extraction rules to obtain N feature data;

[0036] The remote control center is used to display N feature data for the staff to mark the power fault type of the power terminal based on the N feature data. The N feature data and the corresponding power fault type are grouped together, and multiple sets of training samples are obtained based on this operation;

[0037] The remote control center is used to input multiple sets of training samples into the deep learning algorithm model for training to obtain a trained deep learning algorithm model, where each N feature data is used as the model input and each power fault type is used as the model output.

[0038] The positive and progressive effects of the present invention:

[0039] The present invention transmits the power terminal log information to the remote control center only when the power terminal self-checks and finds abnormal operating status behavior. The remote control center pre-processes the power terminal log information and then extracts N feature data from it. The N feature data corresponding to the power terminal are input into the corresponding trained deep learning algorithm model to predict power faults and output the specific power fault type of the power terminal. The prediction method of the present invention has high accuracy, fast processing speed, fast real-time performance, and does not occupy a large amount of memory. Moreover, the present invention can also count the number of times each power fault type occurs in the power terminal within a certain time period. When the number of a certain power fault type in the power terminal reaches the set number, a secondary alarm prompt is issued and the relevant responsible personnel are notified by SMS or email, which is timely and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a structural block diagram of a power terminal log fault analysis and prediction system according to a preferred embodiment of the present invention.

[0041] Figure 2 This is a flow chart of fault training according to a preferred embodiment of the present invention.

[0042] Figure 3 This is a fault prediction flow chart of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0043] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention. On the contrary, the embodiments of the present invention include all variations, modifications, and equivalents that fall within the spirit and scope of the appended claims.

[0044] In the description of the present invention, it should be understood that “plurality” means two or more.

[0045] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0046] The following describes a method and system for analyzing and predicting faults in power terminal logs based on deep learning according to an embodiment of the present invention with reference to the accompanying drawings.

[0047] like Figure 1 As shown, a power terminal log fault analysis and prediction system based on deep learning includes multiple power terminals, multiple gateways and a remote control center. The power terminals exchange information with the remote control center through the gateways. The correspondence between the power terminals and the gateways is a one-to-one correspondence or a many-to-one correspondence. Figure 1 A one-to-one relationship is shown in .

[0048] like Figure 2 and Figure 3 As shown, a power terminal log fault analysis and prediction method based on deep learning includes the following steps:

[0049] See Figure 2 , fault training process:

[0050] Step 101: When a certain power terminal in a set area self-checks and has abnormal operation status, it directly transmits the power terminal log information formed by the abnormal timestamp node to the remote control center through the corresponding gateway. This is the same for each power terminal.

[0051] Step 102: The remote control center preprocesses the power terminal log information to obtain preprocessed power terminal log information. The preprocessing operation includes: deleting meaningless punctuation marks, converting non-standard characters into standard characters, etc.

[0052] Step 103: The remote control center extracts key data from the pre-processed power terminal log information according to preset extraction rules to obtain N feature data.

[0053] In step 104, the remote control center displays N feature data for the staff to mark the power fault type of the power terminal based on the N feature data. This step is based on manual marking, and the power fault type corresponding to the N feature data is marked. These N feature data and the corresponding power fault type are grouped together. Based on this operation, multiple groups of training samples can be obtained, and the power fault type is line fault, transformer fault, bus fault or other fault.

[0054] Step 105: Input multiple sets of training samples into the deep learning algorithm model for training to obtain a trained deep learning algorithm model, wherein each N feature data is used as an input of the deep learning algorithm model, and each power fault type is used as an output of the deep learning algorithm model.

[0055] Based on this fault training process, a trained deep learning algorithm model can be obtained. The next step is to use this trained deep learning algorithm model for prediction.

[0056] See Figure 3 , fault prediction process:

[0057] Step 201: Set each power terminal in the area to self-check whether it has abnormal operating status. If one of the power terminals self-checks and finds abnormal operating status, go to step 202. If the power terminal self-checks and finds no abnormal operating status, repeat step 201. Each power terminal is the same.

[0058] Step 202: The power terminal sends an abnormal operation status instruction to the remote control center through the corresponding gateway. The abnormal operation status instruction includes an abnormal timestamp, an abnormal operation status identifier, a power terminal identification number, a power terminal key, and a gateway identification number. The correspondence between the power terminal and the gateway is a one-to-one correspondence or a many-to-one correspondence.

[0059] Step 203: The remote control center performs a validity check based on the power terminal identification number and the power terminal key in the abnormal operation status instruction, and proceeds to step 204 when the check succeeds, and proceeds to step 205 when the check fails.

[0060] In step 204 , the remote control center feeds back a log information reporting instruction to the power terminal through the corresponding gateway, and then proceeds to step 206 . The log information reporting instruction includes an abnormal timestamp, a log information reporting identifier, and a power terminal identification number.

[0061] Step 205: The remote control center does not perform any further processing.

[0062] Step 206: The power terminal transmits the power terminal log information formed by the abnormal timestamp node to the remote control center through the corresponding gateway.

[0063] Step 207: The remote control center preprocesses the power terminal log information to obtain preprocessed power terminal log information. The preprocessing operation includes: deleting meaningless punctuation marks, converting non-standard characters into standard characters, etc.

[0064] Step 208: The remote control center extracts key data from the pre-processed power terminal log information according to preset extraction rules to obtain N feature data, where N is a positive integer.

[0065] Step 209: The remote control center inputs the N feature data corresponding to the power terminal into the corresponding trained deep learning algorithm model to perform power fault prediction to output the specific power fault type of the power terminal, which is line fault, transformer fault, bus fault or other fault.

[0066] Step 210: The remote control center displays the specific power fault type of the power terminal and issues a first-level alarm prompt.

[0067] Step 211: The remote control center counts and displays the number of times each type of power fault occurs at the power terminal within a set time period (e.g., one week). When the number of a certain type of power fault at the power terminal reaches a set number (e.g., 5 times), a secondary alarm is issued and the relevant responsible personnel are notified by SMS or email.

[0068] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A power terminal log fault analysis and prediction method based on deep learning, characterized in that: It includes the following steps: S1. Each power terminal in the set area self-checks to see if there is any abnormal operating behavior. If one of the power terminals self-checks and finds that there is any abnormal operating behavior, the process proceeds to step S2. If the power terminal self-checks and finds that there is no abnormal operating behavior, the process repeats step S1. S2. The power terminal sends an abnormal operation status instruction to the remote control center through the corresponding gateway. The abnormal operation status instruction includes an abnormal timestamp, an abnormal operation status identifier, a power terminal identification number, a power terminal key, and a gateway identification number. The correspondence between the power terminal and the gateway is a one-to-one correspondence or a many-to-one correspondence; S3, the remote control center performs a legitimacy check based on the power terminal identification number and the power terminal key in the abnormal operation status instruction, and proceeds to step S4 if the check succeeds, or proceeds to step S5 if the check fails; S4. The remote control center feeds back a log information reporting instruction to the power terminal through the corresponding gateway, and enters step S6. The log information reporting instruction includes the abnormal timestamp, the log information reporting identifier, and the power terminal identification number; S5. The remote control center does not perform any further processing; S6. The power terminal transmits the power terminal log information formed by the abnormal timestamp node to the remote control center through the corresponding gateway; S7, the remote control center preprocesses the power terminal log information to obtain preprocessed power terminal log information; S8. The remote control center extracts key data from the pre-processed power terminal log information according to a preset extraction rule to obtain N feature data, where N is a positive integer; S9. The remote control center inputs N feature data corresponding to the power terminal into the corresponding trained deep learning algorithm model to perform power fault prediction to output the specific power fault type of the power terminal, which is a line fault, transformer fault, bus fault or other fault; S10. The remote control center displays the specific power fault type of the power terminal and issues a first-level alarm prompt; S11. The remote control center counts and displays the number of times each type of power fault occurs at the power terminal within a set time period. When the number of a certain type of power fault at the power terminal reaches the set number, a secondary alarm is issued and relevant responsible personnel are notified.

2. The power terminal log fault analysis and prediction method according to claim 1 is characterized in that: Before step S1, the following steps are included: When a power terminal among the power terminals self-checks and finds abnormal operation status, the power terminal log information formed by the abnormal timestamp node is directly transmitted to the remote control center through the corresponding gateway; The remote control center preprocesses the power terminal log information to obtain preprocessed power terminal log information; The remote control center extracts key data from the pre-processed power terminal log information according to preset extraction rules to obtain N feature data; The remote control center displays N feature data for the staff to mark the power fault type of the power terminal based on the N feature data. The N feature data and the corresponding power fault type are grouped together, and multiple sets of training samples are obtained based on this operation; The remote control center inputs multiple sets of training samples into the deep learning algorithm model for training to obtain a trained deep learning algorithm model, where each N feature data is used as the model input and each power fault type is used as the model output.

3. A power terminal log fault analysis and prediction system based on deep learning, characterized in that: It includes multiple power terminals, multiple gateways and a remote control center, and the corresponding relationship between the power terminals and the gateways is a one-to-one correspondence or a many-to-one correspondence; Each power terminal in the set area is used to self-check whether it has abnormal operating status. If a power terminal in the power terminal self-checks and finds abnormal operating status, the power terminal sends an abnormal operating status instruction to the remote control center through the corresponding gateway. If the power terminal self-checks and finds no abnormal operating status, the operation is repeated. The abnormal operating status instruction includes an abnormal timestamp, an abnormal operating status identifier, a power terminal identification number, a power terminal key, and a gateway identification number. The remote control center is used to perform a legitimacy check based on the power terminal identification number and the power terminal key in the operation status abnormality instruction. When the verification is successful, the remote control center feeds back a log information reporting instruction to the power terminal through the corresponding gateway. When the verification fails, the remote control center does not perform further processing. The log information reporting instruction includes an abnormal timestamp, a log information reporting identifier, and the power terminal identification number; The power terminal is used to transmit the power terminal log information formed by the abnormal timestamp node to the remote control center through the corresponding gateway; The remote control center is used to preprocess the power terminal log information to obtain preprocessed power terminal log information; The remote control center is used to extract key data from the pre-processed power terminal log information according to preset extraction rules to obtain N feature data, where N is a positive integer; The remote control center is used to input N feature data corresponding to the power terminal into the corresponding trained deep learning algorithm model to perform power fault prediction and output the specific power fault type of the power terminal, which is a line fault, transformer fault, bus fault or other fault; The remote control center is used to display the specific power fault type of the power terminal and issue a first-level alarm prompt; The remote control center is used to count and display the number of times each type of power fault occurs at the power terminal within a set time period. When the number of a certain type of power fault at the power terminal reaches the set number, a secondary alarm prompt is issued and the relevant responsible personnel are notified.

4. The power terminal log fault analysis and prediction system according to claim 3 is characterized in that: When a power terminal among the power terminals is used for self-checking and abnormal operation behavior occurs, the power terminal log information formed by the abnormal timestamp node is directly transmitted to the remote control center through the corresponding gateway; The remote control center is used to preprocess the power terminal log information to obtain preprocessed power terminal log information; The remote control center is used to extract key data from the pre-processed power terminal log information according to preset extraction rules to obtain N feature data; The remote control center is used to display N feature data for the staff to mark the power fault type of the power terminal based on the N feature data. The N feature data and the corresponding power fault type are grouped together, and multiple sets of training samples are obtained based on this operation; The remote control center is used to input multiple sets of training samples into the deep learning algorithm model for training to obtain a trained deep learning algorithm model, where each N feature data is used as the model input and each power fault type is used as the model output.

Images