Relay protection equipment constant value modification behavior identification method, device and equipment

Abstract

The invention relates to a method, a device and equipment for identifying a constant value modification behavior of relay protection equipment. The method comprises the following steps: acquiring an alarm signal of a secondary monitoring master station of a power system; an alarm signal is used as input, a trained signal identification model is called, whether the alarm signal is a constant value modification signal sent by the relay protection equipment or not is identified, a signal identification result is obtained, and the trained signal identification model is obtained through training on the basis of a historical alarm signal carrying a constant value modification signal category label; and under the condition that the signal identification result represents that the alarm signal comprises a plurality of constant value modification signals, identifying a constant value modification behavior category corresponding to each constant value modification signal. By adopting the method, the efficiency of identifying the constant value modification behavior of the relay protection equipment can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of relay protection for electric power systems, and in particular to a method, device and equipment for identifying the setting modification behavior of relay protection equipment. Background Art

[0002] Relay protection is the first line of defense to ensure the safe and stable operation of the power system. In the process of ensuring the safe and stable operation of the power system, whether the relay protection device can operate reliably is closely related to whether its setting value is appropriate. The setting value of the protection device is determined by the system operation mode. With the continuous access of random, intermittent, and fluctuating power sources such as wind power and photovoltaic power, the power system highlights the characteristics of a high proportion of new energy, and the uncertainty faced by the system operation is becoming more and more obvious. Under strong uncertainty, the operation mode of the high-proportion new energy power system is changeable. Affected by this, the number of modifications to the setting values ​​of protection devices such as AC lines has increased significantly compared with the past.

[0003] During the setting value modification process, the setting value setting results must be correctly sent to the corresponding protection device within the planned time. If the setting value modification is wrong or the setting value modification operation is not completed within the specified time, once a fault occurs, the relay protection device will not be able to isolate the fault in time, and may even cause a larger power outage. At present, the power control center mainly relies on the remote telephone communication between the secondary monitoring personnel and the on-site operation and maintenance personnel to realize the perception of the setting value modification process, and then ensure that the protection setting value is sent on time and accurately, but this method consumes too much attention of the secondary monitoring personnel.

[0004] Therefore, a more effective method for identifying constant value modification behavior is urgently needed to assist the secondary monitoring personnel of the power control center to quickly and accurately remotely perceive the on-site constant value modification process and ensure the safe and stable operation of the power system. Summary of the invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for identifying the constant value modification behavior of relay protection equipment, which can improve the efficiency of identifying the constant value modification behavior of relay protection equipment in response to the above-mentioned technical problems.

[0006] In a first aspect, the present application provides a method for identifying setting modification behavior of a relay protection device, comprising:

[0007] Obtain the alarm signal from the secondary monitoring master station of the power system;

[0008] Taking the alarm signal as input, calling the trained signal recognition model, identifying whether the alarm signal is a fixed value modification signal sent by the relay protection device, and obtaining the signal recognition result, the trained signal recognition model is trained based on the historical alarm signal carrying the fixed value modification signal category label;

[0009] When the signal identification result indicates that the alarm signal includes a plurality of fixed value modification signals, the fixed value modification behavior category corresponding to each fixed value modification signal is identified.

[0010] In a second aspect, the present application also provides a device for identifying setting modification behavior of a relay protection device, comprising:

[0011] A signal acquisition module is used to obtain the alarm signal of the secondary monitoring master station of the power system;

[0012] A signal category recognition module is used to take the alarm signal as input, call the trained signal recognition model, identify whether the alarm signal is a fixed value modification signal sent by the relay protection device, and obtain a signal recognition result. The trained signal recognition model is trained based on historical alarm signals carrying fixed value modification signal category labels;

[0013] The fixed value modification behavior category identification module is used to identify the fixed value modification behavior category corresponding to each fixed value modification signal when the signal identification result indicates that the alarm signal contains multiple fixed value modification signals.

[0014] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps in any one of the above-mentioned embodiments of the method for identifying the constant value modification behavior of a relay protection device are implemented.

[0015] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in any one of the above-mentioned embodiments of the method for identifying the setting modification behavior of a relay protection device.

[0016] In a fifth aspect, the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps in any one of the above-mentioned embodiments of the method for identifying the constant value modification behavior of a relay protection device.

[0017] The above-mentioned method, device, computer device, computer-readable storage medium and computer program product for identifying the fixed value modification behavior of relay protection equipment can obtain a signal recognition model by pre-training based on the alarm signal carrying the fixed value modification signal category label. In the actual fixed value modification identification process, the massive alarm signals sent to the secondary monitoring main station of the power system can be input into the trained signal recognition model, so as to efficiently and accurately identify the fixed value modification signal caused by the fixed value modification behavior. When it is identified that the alarm signal contains multiple fixed value modification signals, the fixed value modification behavior category corresponding to each fixed value modification signal is further identified, thereby realizing the remote identification of the fixed value modification behavior of the relay protection. The whole process does not need to rely on the practice of manual telephone contact perception or manual query, which improves the efficiency of the fixed value modification behavior identification. Furthermore, the identification result can assist the power control personnel to ensure that the fixed value is modified on time and accurately, thereby improving the safety and stability of the power system operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 An application environment diagram of a method for identifying a setting modification behavior of a relay protection device in one embodiment;

[0020] Figure 2 A schematic diagram of a flow chart of a method for identifying a setting modification behavior of a relay protection device in one embodiment;

[0021] Figure 3 A schematic flow chart of a method for identifying a setting modification behavior of a relay protection device in another embodiment;

[0022] Figure 4 A flowchart of a method for identifying a setting modification behavior of a relay protection device in another embodiment;

[0023] Figure 5 A flowchart of a method for identifying a setting modification behavior of a relay protection device in another embodiment;

[0024] Figure 6 A detailed flow chart of a method for identifying a setting modification behavior of a relay protection device in one embodiment;

[0025] Figure 7 A structural block diagram of a device for identifying a setting modification behavior of a relay protection device in one embodiment;

[0026] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0028] The method for identifying the setting modification behavior of a relay protection device provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The server 104 is deployed with a signal recognition model trained based on historical alarm signals carrying fixed value modification signal category labels.

[0029] Specifically, the operator may collect the alarm signal from the power system secondary monitoring main station through the terminal 102 and upload it to the server 104, and then send the relay protection device constant value modification behavior identification message to the server 104 through the terminal 102. The server 104 obtains the alarm signal, and then calls the signal recognition model to identify whether the alarm signal is a constant value modification signal sent by the relay protection device to obtain the signal recognition result. Thirdly, when the signal recognition result indicates that the alarm signal contains multiple constant value modification signals, the constant value modification behavior category corresponding to each constant value modification signal is identified.

[0030] The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, relay protection equipment, and terminals of a secondary monitoring master station of a power system. The server 104 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0031] In an exemplary embodiment, Figure 2 As shown in the figure, a method for identifying the setting modification behavior of relay protection equipment is provided. Figure 1 The server 104 in FIG. 1 is used as an example to illustrate the process, including the following S200 to S600. Among them:

[0032] S200, obtaining an alarm signal from a secondary monitoring master station of the power system.

[0033] Among them, the power system secondary monitoring master station is the main station system for the power control center to uniformly monitor and control each secondary equipment in the power system. It can collect, process, analyze and display the real-time operation information of the power grid secondary equipment to realize remote monitoring and control of the power grid secondary equipment and ensure the safe and stable operation of the power grid. The alarm signal is a power secondary alarm signal, which is generated by the power secondary equipment in the power system. The power secondary equipment is an auxiliary equipment for monitoring, measuring, controlling, regulating and protecting the primary equipment of the power system, and is not directly connected to the main circuit of electric energy. The content of the alarm signal includes but is not limited to the signal attributes and corresponding attribute values ​​such as the upload time, device name, plant station, signal name, alarm type and alarm level.

[0034] In the power system, when the power secondary equipment fails, operates abnormally, or requires maintenance or overhaul, some power secondary equipment will generate an alarm signal and send it to the power system secondary monitoring master station to assist the maintenance personnel in taking appropriate measures. Power secondary equipment includes but is not limited to relay protection equipment, automatic voltage regulating equipment, automatic monitoring system, and DC power supply equipment.

[0035] In practical applications, the alarm signals sent by the secondary power equipment can be obtained through the power system secondary monitoring master station.

[0036] S400, taking the alarm signal as input, calling the trained signal recognition model, identifying whether the alarm signal is a fixed value modification signal sent by the relay protection device, and obtaining the signal recognition result. The trained signal recognition model is trained based on historical alarm signals carrying fixed value modification signal category labels.

[0037] The signal identification result includes the signal category of the alarm signal, and the signal category includes a fixed value modification signal and a non-fixed value modification signal.

[0038] The modification of the setting value of relay protection equipment involves the safe and stable operation of the power system. The setting value (also called the setting value or set value) is the threshold and parameter that triggers the action of the relay protection equipment when a fault is detected. These setting values ​​need to be accurately set according to the actual operating conditions of the power grid, equipment characteristics and protection requirements. When the grid structure changes (such as adding new lines, transformers, etc.), the load characteristics change, or the protection requirements are updated, the setting value needs to be modified to ensure the correctness and reliability of the protection of the relay protection equipment.

[0039] The fixed value modification behavior can include multiple fixed value modification sub-behaviors such as safety measure execution, fixed value modification, and safety measure recovery. Among them, safety measure execution and safety measure recovery respectively represent the execution and recovery of the content of safety technical measures to achieve the locking and starting of the relay protection function of the relay protection equipment. Each fixed value modification sub-behavior will trigger multiple fixed value modification signals, and different fixed value modification sub-behaviors will trigger different fixed value modification signals.

[0040] When modifying the setting value of the relay protection device, the relay protection device will send multiple setting value modification signals to the power system secondary monitoring master station, and the power system secondary monitoring master station will receive alarm signals from multiple power secondary devices at the same time. In order to realize the remote identification of the setting value modification behavior of the relay protection device, it is necessary to first identify the setting value modification signal sent by the relay protection device from the alarm signal sent by the power secondary device. Identifying the setting value modification signal sent by the relay protection device can be achieved by training a deep learning model.

[0041] In practical applications, a signal recognition model can be pre-built and trained using historical alarm signals carrying fixed value modification signal category labels to obtain a trained signal recognition model. The pre-built signal recognition model can be based on a recurrent neural network, a Transformer model, or a multi-layer perceptron model. In specific implementation, the acquired alarm signal is input into the trained signal recognition model to identify whether the alarm signal is a fixed value modification signal sent by the relay protection device.

[0042] S600: When the signal identification result indicates that the alarm signal includes a plurality of fixed value modification signals, identify a fixed value modification behavior category corresponding to each fixed value modification signal.

[0043] Each fixed value modification behavior category represents a single fixed value modification behavior, which is used to distinguish the fixed value modification behaviors made to different relay protection devices, and the fixed value modification behaviors made to the same relay protection device at different times. Since the fixed value modification behavior will trigger multiple fixed value modification signals, the identification of the fixed value modification behavior is essentially to identify the fixed value modification behavior to which each fixed value modification signal belongs.

[0044] In practical applications, considering that the fixed value modification behaviors of different types of relay protection devices are different, the fixed value modification signals triggered are also different. By analyzing the signal characteristics of the fixed value modification signals sent by different types of relay protection devices, the fixed value modification behavior category identification rules corresponding to each relay protection device model can be set. In the case where the signal identification result indicates that the alarm signal contains multiple fixed value modification signals, the fixed value modification behavior category identification rules are used according to the relay protection device model to which the fixed value modification signal belongs to identify the fixed value modification behavior category corresponding to the fixed value modification signal.

[0045] In the above-mentioned method for identifying the fixed value modification behavior of relay protection equipment, by acquiring a large number of alarm signals sent to the secondary monitoring main station of the power system and inputting them into the signal recognition model trained based on the alarm signal carrying the fixed value modification signal category label, the fixed value modification signal caused by the fixed value modification behavior can be efficiently and accurately identified. When it is identified that the alarm signal contains multiple fixed value modification signals, the fixed value modification behavior category corresponding to each fixed value modification signal is further identified, thereby realizing the remote identification of the fixed value modification behavior of the relay protection. The whole process does not need to rely on manual telephone contact perception or manual query, which improves the efficiency of the fixed value modification behavior identification. Furthermore, the identification results can assist power control personnel to ensure that the fixed value is modified on time and accurately, thereby improving the safety and stability of the power system operation.

[0046] In an exemplary embodiment, S600 includes S620. Wherein:

[0047] S620, calling the constructed constant value modification behavior recognition model to identify the constant value modification behavior category corresponding to each constant value modification signal, the constant value modification behavior recognition model is obtained based on a set of constant value modification signals corresponding to multiple different constant value modification behavior categories and preset constant value modification behavior category recognition rules.

[0048] In practical applications, we can first obtain multiple fixed value modification signal sets corresponding to different fixed value modification behavior categories, clean and integrate multiple fixed value modification signal sets, and build a fixed value modification signal feature library. Based on the fixed value modification signal feature library, preset value modification behavior category recognition rules, and then build a knowledge-driven decision-making fixed value modification behavior recognition model based on the fixed value modification signal feature library and the preset fixed value modification behavior category recognition rules. When the signal recognition result indicates that the alarm signal contains multiple fixed value modification signals, call the fixed value modification behavior recognition model to determine the fixed value modification behavior category corresponding to each fixed value modification signal.

[0049] In this embodiment, by pre-setting the fixed value modification behavior category identification rules, a fixed value modification behavior identification model is constructed based on the fixed value modification signals corresponding to multiple different fixed value modification behavior categories and the fixed value modification behavior category identification rules. Therefore, when the signal identification result indicates that the alarm signal contains multiple fixed value modification signals, the constructed model is called to identify the fixed value modification behavior category corresponding to each fixed value modification signal, thereby improving the efficiency of identifying the fixed value modification behavior of the relay protection equipment.

[0050] In practical applications, the boundary signal of the fixed value modification signal corresponding to each fixed value modification behavior category is first identified, so as to obtain the category of the fixed value modification signal based on the category of the boundary signal. Figure 3As shown, S600 also includes S640 to S680. Among them:

[0051] S640, filtering out the first signal and the last signal corresponding to different fixed value modification behavior categories from the multiple fixed value modification signals.

[0052] Each fixed value modification behavior is composed of multiple fixed value modification sub-behaviors, and there is an operation priority between the multiple sub-behaviors. Exemplarily, the fixed value modification behavior includes security measure execution, fixed value modification, and security measure recovery, and the execution order satisfies security measure execution->fixed value modification->security measure recovery. Different sub-behaviors trigger multiple different fixed value modification signals respectively.

[0053] Among them, the first signal is the first fixed value modification signal sent by the relay protection device among multiple fixed value modification signals corresponding to each fixed value modification behavior category, and the tail signal is the last fixed value modification signal sent by the relay protection device among multiple fixed value modification signals corresponding to each fixed value modification behavior category.

[0054] In order to identify which fixed value modification signals are triggered by each fixed value modification behavior category, the boundaries of different fixed value modification behavior categories (i.e., the first signal and the last signal) can be obtained by analyzing the characteristics of the first signal and the last signal, so as to analyze the fixed value modification signals between the first signal and the last signal.

[0055] In practical applications, the first signal and the last signal corresponding to different fixed value modification behaviors are screened out from multiple fixed value modification signals. This can be done by analyzing the signal contents of the fixed value modification signals triggered by the first fixed value modification sub-behavior and the last fixed value modification sub-behavior in each fixed value modification behavior obtained historically, obtaining the keywords and keyword groups of the first signal and the last signal, and screening out different first signals and last signals based on the content of the alarm signal, as well as the keywords and keyword groups.

[0056] S660: Based on the first signal and the last signal corresponding to each fixed value modification behavior category, a fixed value modification signal set corresponding to the fixed value modification behavior category is constructed.

[0057] After screening out different first and last signals, the fixed value modification signal between the first and last signals belonging to the same fixed value modification behavior category can be identified based on the first and last signals. Therefore, a corresponding fixed value modification signal set is constructed for each fixed value modification behavior category.

[0058] S680, classify the fixed value modification signals under the same fixed value modification behavior category into the same fixed value modification signal set, and determine the fixed value modification behavior category corresponding to each fixed value modification signal.

[0059] Through the contents of the first signal and the last signal, the constant value modification signal between the first signal and the last signal can be determined, and the constant value modification behavior category of the middle constant value modification signal can be determined as the constant value modification behavior category of the first signal and the last signal, and the constant value modification signals belonging to the same constant value modification behavior category can be divided into the same constant value modification signal set.

[0060] In this embodiment, from multiple fixed value modification signals, the first signal and the last signal caused by different fixed value modification behavior categories are first determined, and then based on the analogy of the fixed value modification behaviors to which the multiple first signal and the last signal belong, each fixed value modification signal caused by each fixed value modification behavior category is determined, thereby realizing the identification of the fixed value modification behavior category of each fixed value modification signal, thereby improving the efficiency of the fixed value modification behavior identification.

[0061] In order to improve the accuracy of identifying the fixed value modification signal from a large number of alarm signals, a signal recognition model is constructed and trained. In an exemplary embodiment, Figure 4 As shown, before S200, it also includes S110 to S180. Among them:

[0062] S110, obtaining historical alarm signals from a secondary monitoring master station of the power system.

[0063] The historical alarm signal is an alarm signal sent by the power secondary equipment in a historical time period, which may be one month, half a year, or one year.

[0064] After obtaining the historical alarm signals, a signal screening mechanism can be established to filter the historical alarm signals to improve the efficiency of identifying the fixed value modification signals. Specifically, the alarm signals that may be fixed value modification signals can be preliminarily screened out based on the specific fields in the fixed value modification signal content. The specific fields are fields related to the fixed value modification behavior, and the specific field set can be pre-established based on experience.

[0065] Use specific fields as signal filtering conditions to filter alarm signals:

[0066]

[0067] In the formula, It is the alarm signal after signal screening; An alarm signal that passes the screening; It is a specific field in the alarm signal content; A pre-built set of specific fields.

[0068] In practical applications, the alarm signals sent by the power secondary equipment within a month are obtained from the power system secondary monitoring main station, and the alarm signals that may be fixed value modification signals are preliminarily screened out through the signal screening mechanism.

[0069] S120: Determine the signal category label of the historical alarm signal.

[0070] Among them, the signal category labels may include "fixed value modification signal" and "non-fixed value modification signal".

[0071] In practical applications, signal category labels are used to annotate the filtered historical alarm signals:

[0072]

[0073] In the formula, It is the marked warning signal; For warning signals; To determine whether the alarm signal is a fixed value modification signal label, Indicates that the alarm signal is a fixed value modification signal. Indicates that the alarm signal is a non-fixed value modification signal.

[0074] S130, performing feature selection and feature processing on the historical alarm signals after data annotation to obtain initial training samples.

[0075] First, feature selection is performed on the content of the historical alarm signal after data annotation to determine the main features in the content of the historical alarm signal. Among them, the main features can be determined according to the triggering principle of the fixed value modification signal sent by the relay protection device.

[0076] In actual applications, signal attributes such as the sending time, equipment name, plant station and signal name in the alarm signal content can be selected as the main features:

[0077]

[0078] In the formula, The main features of the signal identification model input; The upload time; is the device name; For the affiliated plant and station; is the signal name.

[0079] The feature processing of the historical alarm signal after data annotation may include a connection processing operation of the attributes of the historical alarm signal and a sliding signal window processing of the historical alarm signal after connection processing.

[0080] Since the attribute values ​​of alarm signals have weaker semantic continuity when they are directly associated than the natural semantic continuity of general sentences, the signal attributes in the historical alarm signals can be first connected to help the signal recognition model better capture the characteristics of the historical alarm signals. Specifically, based on the relationship between the signal attributes of the historical alarm signals and the input requirements of the pre-sub-model of the signal recognition model, the sentence segmentation mark "[SEP]" is used to segment each historical alarm signal into sentences and each signal attribute in the historical alarm signal into words to indicate the relative positions between the signal attributes.

[0081] For each historical alarm signal, perform connection processing of signal attributes:

[0082]

[0083] In the formula, It is the historical alarm signal after signal attribute connection processing; Indicates signal attribute connection processing operation; The main features used as input to the signal classification model.

[0084] The historical alarm signals after connection processing are processed by sliding signal window processing to obtain multiple historical alarm signal sequences, so that the signal recognition model can perform feature analysis on the historical alarm signal sequences. Specifically, the window length of the sliding signal window is determined to be a preset number of signals, and the historical alarm signals are moved and framed with the sliding signal window of the preset number of signals to obtain a historical alarm signal sequence of the preset number of signal lengths. The preset number of signals is determined by considering the correlation of multiple fixed value modification signals caused by a single fixed value modification behavior. Within the sliding time window of the preset number of signals, the time interval for uploading the first historical alarm signal and the last historical alarm signal of the historical alarm signal sequence shall not exceed the longest time of the fixed value modification behavior. The longest time of the fixed value modification behavior is set based on experience.

[0085] Use a sliding signal window with a preset number of signals to move and select historical alarm signals:

[0086]

[0087] In the formula, is the historical alarm signal sequence corresponding to the sliding signal window; For the Historical alarm signals; It is the most recent historical alarm signal within the sliding signal window; is the number of historical alarm signals in the sliding signal window.

[0088] After being processed by the sliding signal window, a plurality of historical alarm signal sequences are obtained, and these historical alarm signal sequences are determined as initial training samples.

[0089] S140, performing data enhancement on the initial training samples to obtain target training samples.

[0090] Among them, the target training sample is the input data of the signal recognition model.

[0091] In practical applications, the initial training samples can be enhanced based on data enhancement methods. Data enhancement methods may include synonym replacement, random insertion, and random deletion, etc. The corresponding enhanced training samples may be generated according to different data enhancement methods, and the target training samples may be obtained according to the initial training samples and the training samples enhanced by different data enhancement methods. Specifically, synonym replacement may be to replace the signal content of the historical alarm signal of the initial training sample; random insertion may be to randomly insert an alarm signal into the historical alarm signal sequence of the initial training sample; and random deletion may be to randomly delete the alarm signal from the historical alarm signal sequence of the initial training sample.

[0092] S150, connecting the first deep learning model and the second deep learning model in series to construct a signal recognition model.

[0093] Among them, the first deep learning model can be BERT (Bidirectional Encoder Representationsfrom Transformers, a bidirectional encoding representation algorithm based on the Transformer algorithm), and the second deep learning model can be Transformer (converter algorithm).

[0094] The signal recognition model in this embodiment is a model obtained by connecting BERT and Transformer in series. Specifically, BERT and the encoder of Transformer are connected in series to construct a data-driven signal recognition model. Among them, BERT is the front sub-model of the signal recognition model, and the encoder of Transformer is the rear sub-model of the signal recognition model.

[0095] S160, input the target training sample into the constructed signal recognition model, so that the signal recognition model learns the semantic features of each historical alarm signal in the target training sample and the timing information between different historical alarm signals, and determines the signal category of each historical alarm signal.

[0096] A single fixed value modification behavior triggers multiple fixed value modification signals, and the content of the fixed value modification signal is related to the fixed value modification behavior. For example, the signal names in the fixed value modification signal content may include "fixed value modification", "device lock", "device alarm", "channel abnormality", etc. A single fixed value modification behavior contains multiple fixed value modification sub-behaviors, and each fixed value modification sub-behavior may contain multiple operations, so that each fixed value modification sub-behavior triggers multiple fixed value modification signals. The execution interval of different fixed value modification sub-behaviors may be long (hours), so that the power system secondary monitoring master station may receive hundreds of other alarm signals in the fixed value modification signals triggered by the first fixed value modification sub-behavior and the last fixed value modification sub-behavior.

[0097] Therefore, the key to the signal recognition model to identify whether an alarm signal is a constant value modification signal lies in: first, for each alarm signal, the correlation information of each signal attribute of the alarm signal is mined as the first key information to analyze whether the alarm signal may be a constant value modification signal; second, on the basis of the above, the correlation information between different alarm signals is mined as the second key information. The correlation relationship between different alarm signals includes the correlation relationship between adjacent signals in the alarm signal sequence and the correlation relationship between long-distance alarm signals, thereby identifying whether the alarm signal is a constant value modification signal.

[0098] This application uses a signal recognition model based on BERT-Transformer Encoder fusion to mine the above key information, so as to determine whether the alarm signal is a fixed value modification signal. The specific description is as follows:

[0099] The embedding layer of the BERT model can map the input alarm signal sequence to obtain three sets of feature vectors. Among them, the first set of vectors is the token embedding vector that represents the semantics of the signal attributes of the alarm signal, the second set of vectors is the segment embedding vector that represents the global semantic information of the signal attributes of the alarm signal in the alarm signal sequence, and the third set is the position embedding vector that represents the position information of the signal attributes of the alarm signal. The sum of these three sets of vectors (as shown in the following formula) is used as the input feature vector of the pre-model (ie, BERT).

[0100]

[0101] In the formula, is the feature vector corresponding to the alarm signal; is the word vector corresponding to the warning signal; is the segment vector corresponding to the alarm signal; is the position vector corresponding to the warning signal.

[0102] Through the BERT model, the signal recognition model can fully obtain the character semantics of the signal attributes of each alarm signal in the input alarm signal sequence, that is, the first key information. The signal character semantics and temporal correlation information between different alarm signals in the alarm signal sequence are obtained, that is, the correlation between adjacent alarm signals in the second key information. The output result of the pre-sub-model is used as the input variable of the post-sub-model (Transformer encoder), and the multi-head attention mechanism of the Transformer encoder is used to further capture the correlation between long-distance alarm signals in the second key information, so that the signal recognition model can fully capture the semantic features of each historical alarm signal in the input alarm signal sequence, as well as the temporal information between different historical alarm signals, so as to more comprehensively understand the signal content of the input alarm signal and model its characteristics to identify the signal category of the alarm signal.

[0103] In practical applications, after constructing the signal recognition model, the target training samples are input into the signal recognition model so that the model can learn the semantic features of each historical alarm signal in the target training samples and the timing information between different historical alarm signals, and output the signal category of each historical alarm signal.

[0104] S170, determining a loss function value of a signal recognition model based on the signal category label of each historical alarm signal in the target training sample and the signal category of the output historical alarm signal.

[0105] A loss function is used to quantify the training effect of the signal recognition model. The loss function includes but is not limited to the cross entropy loss function, Hinge loss function, exponential loss function, and logarithmic loss function.

[0106] Exemplarily, a cross entropy loss function is used to quantify the training effect of the signal recognition model, and the cross entropy loss function value of the signal recognition model is determined:

[0107]

[0108] In the formula, is the cross entropy loss function; is the total number of historical alarm signals; It is the historical alarm signal sequence number; is the signal category label; identifying warning signals for the signal recognition model; Modify the probability value of the signal for a given value.

[0109] S180, iteratively updating the parameters of the signal recognition model based on the loss function value until a preset training end condition is reached to obtain a trained signal recognition model.

[0110] In practical applications, the signal recognition model can be back-propagated according to the loss function value, and the parameters of the signal recognition model can be updated by gradient descent to minimize the loss function value. The preset training end condition can be that the loss function value is continuously less than the preset loss threshold within the preset number of iterations, and the training of the signal recognition model is stopped to obtain the trained signal recognition model.

[0111] In this embodiment, target training samples are obtained by performing data annotation, feature selection and feature processing, and data enhancement on historical alarm signals to improve the diversity of training samples, which helps to improve the learning efficiency and generalization ability of the signal recognition model. The signal recognition model constructed based on the target training sample training captures the characteristics of the fixed value modification signal to achieve the recognition of the fixed value modification signal, thereby improving the accuracy of the fixed value modification signal recognition, which is conducive to improving the accuracy of the fixed value modification behavior recognition.

[0112] In order to improve the diversity of training samples, in an exemplary embodiment, Figure 5 As shown, S140 includes at least one of S142 to S146. Among them:

[0113] Since the fixed value modification signals account for a relatively low proportion in the massive alarm signals, and the multiple fixed value modification signals corresponding to each fixed value modification behavior generally have the characteristics of missing signals, repeated transmission and inconsistent orders, the alarm signal sequences in the initial training samples can be simulated based on signal category labels to achieve data enhancement.

[0114] S142, for each alarm signal sequence, select an alarm signal segment in the alarm signal sequence, generate an analog alarm signal matching the alarm signal in the alarm signal segment, and interpolate the analog alarm signal into the alarm signal sequence to obtain a first training sample.

[0115] The simulated alarm signal may be a signal having some signal properties different from those of the alarm signal.

[0116] When the secondary operation and maintenance personnel modify the setting information of the relay protection model, there may be one or more repeated setting modification operations, resulting in the same setting modification signal being repeatedly sent to the master station system. Based on the characteristics of repeated upload, the situation where the setting modification signal is repeatedly sent to the secondary monitoring master station of the power system is simulated, and data enhancement is performed on the initial training samples. Specifically, for each alarm signal sequence, an alarm signal segment is randomly selected to generate a simulated alarm signal that matches the alarm signal in the alarm signal segment. For example, a signal with the same signal attribute values ​​as the alarm signal in the alarm signal segment except for the attribute value of the upload time is generated as a simulated signal. According to the maximum number of repeated operations, it is randomly interpolated into the alarm signal sequence to obtain the first training sample. The maximum number of repeated operations can be determined based on experience.

[0117] S144, for each alarm signal sequence, reorder the alarm signals in the alarm signal sequence that are sent at the same time to obtain a second training sample.

[0118] In the process of executing the fixed value modification of relay protection, multiple fixed value modification signals may be triggered at the same time, and the order of sending the fixed value modification signals triggered at the same time is randomly arranged. Based on the characteristics of different orders, the random order of sending fixed value modification signals at the same sending time is simulated to enhance the data of the initial training samples. Specifically, for each alarm signal sequence, according to the sending time attribute in the alarm signal, a random function is used to randomly sort the alarm signals sent at the same time to obtain the second training sample.

[0119] S146, for each alarm signal sequence, assign a weight to each alarm signal in the alarm signal sequence, determine a missing analog signal set based on the weight of each alarm signal, remove the alarm signals in the missing analog signal set from the alarm signal sequence, and generate a third training sample.

[0120] During the process of modifying the setting value of the relay protection equipment, communication channel interruption and other situations may occur, resulting in the situation that some of the setting value modification signals caused by the setting value modification behavior are not sent to the secondary monitoring master station of the power system. Based on the characteristics of signal missing, the failure of sending the setting value modification signal to the secondary monitoring master station of the power system is simulated, and the data of the initial training samples is enhanced.

[0121] Specifically, for each alarm signal sequence, a weight value is randomly assigned to each alarm signal in the alarm signal sequence, and the weight value represents the probability that the alarm signal is a missing alarm signal. The generated weight value follows the Bernoulli distribution:

[0122]

[0123] In the formula, is the weight value; represents the Bernoulli distribution.

[0124] For each alarm signal sequence, a missing alarm signal number is randomly generated, and the missing alarm signal number represents the number of alarm signals that are randomly eliminated. Based on the size of the randomly assigned weight value, the eliminated alarm signals are determined to obtain a missing simulation signal set. For example, alarm signals with larger weight values ​​are determined as eliminated alarm signals until the number of eliminated alarm signals meets the randomly generated missing alarm signal number, and the missing simulation signal set corresponding to the alarm signal sequence is obtained.

[0125] Remove the alarm signals in the missing simulation signal set from the alarm signal sequence to obtain the alarm signal sequence after signal missing simulation:

[0126]

[0127]

[0128] In the formula, It is the index of the alarm signal retained after the signal window passes the simulation of the failure of sending up; is the weight set corresponding to the removed alarm signals; The number of missing alarm signals refers to the number of alarm signals that are randomly removed in the simulation. The number of missing alarm signals is set to be less than the total number of alarm signals in the signal window. ; It is the alarm signal retained after the signal window passes through the upload failure simulation.

[0129] Considering special cases, for a single fixed value modification behavior, there may be multiple fixed value modification signals that are successfully sent up or failed to be sent up at the same time. When simulating this situation, for each alarm signal sequence, after randomly assigning a weight value to each alarm signal in the sequence, determine the average weight value of the alarm signal in the sequence, replace the weight value of each alarm signal with the average weight value, and determine whether to remove the alarm signal based on the replaced average weight value. After the signal missing simulation, the third training sample is obtained.

[0130] In other embodiments, data enhancement is performed on the initial training sample, further comprising: for each alarm signal sequence, obtaining equipment inventory information corresponding to different relay protection equipment models, and replacing the attribute value of the target signal attribute of the alarm signal in the alarm signal sequence based on the equipment inventory information to obtain a fourth training sample. For example, the target signal attribute may include the device name and the plant to which it belongs.

[0131] In this embodiment, targeted data enhancement is performed on the initial training samples based on the characteristics of the upward transmission of multiple fixed value modification signals caused by the fixed value modification behavior, thereby improving the diversity of the training samples of the signal recognition model and reducing the possibility of imbalance between positive and negative samples in the training samples, which is beneficial to improving the generalization ability and robustness of the signal recognition model, and further improving the accuracy of identifying the fixed value modification signals based on the signal recognition model.

[0132] In order to identify the fixed value modification behavior category to which the fixed value modification signal belongs, a fixed value modification behavior identification model is constructed. In an exemplary embodiment, Figure 6 As shown, before S200, it also includes S182 to S188. Among them:

[0133] S182, obtaining a set of setting modification signals of multiple models of relay protection devices.

[0134] In practical applications, the logs of different types of relay protection devices can be obtained to analyze the fixed value modification behaviors of different types of relay protection devices and the fixed value modification signal sets sent by the relay protection devices. Each fixed value modification signal set includes a fixed value modification signal subset corresponding to each fixed value modification sub-behavior.

[0135] S184, determining the similarity of the signaling principles of any two relay protection devices of different models.

[0136] Among them, the similarity of the signaling principle represents the similarity of the triggering logic of the set value modification signal of the relay protection device.

[0137] For any two relay protection devices of different models, based on the set of fixed value modification signals of the relay protection device pair, the similarity of the sending principles of the relay protection device pair is determined by analyzing the similarity of the number of fixed value modification signals triggered by each fixed value modification sub-behavior in the set of fixed value modification signals, and the similarity of the content of the fixed value modification signals triggered by each fixed value modification sub-behavior in the set of fixed value modification signals.

[0138] The ratio similarity can be used to quantify the similarity of the number of setting modification signals triggered by the setting modification sub-behavior of the relay protection device pair:

[0139]

[0140] In the formula, The similarity of the number of setting modification signals triggered by setting modification sub-behaviors corresponding to different types of relay protection devices; , Respectively represent the relay protection model And relay protection model ; Modify the sub-behavior sequence number for the fixed value; Modify the total number of sub-behaviors for a fixed value; , Respectively represent the relay protection model And relay protection model Correspondingly, the fixed value modifies the sub-behavior The number of set value modification signals triggered.

[0141] The Jaccard coefficient can be used to quantify the similarity of the content of the setting modification signal triggered by the setting modification sub-behavior of the relay protection device pair:

[0142]

[0143] In the formula, The similarity of the contents of the setting modification signal triggered by the setting modification sub-behavior corresponding to different types of relay protection equipment; , Respectively represent the fixed value modification sub-behavior and set value to modify sub-behavior ; , Respectively represent the relay protection model And relay protection model Correspondingly, the fixed value modifies the sub-behavior The triggered setting modifies the signal content.

[0144] For any two relay protection devices of different models, after obtaining the similarity in number and content of the setting modification signals caused by each identical setting modification sub-behavior of the relay protection device pair, the similarity in the signaling principle between the relay protection device pairs is determined:

[0145]

[0146] In the formula, Relay protection model And relay protection model The constant value modifies the behavior of the signaling principle similarity.

[0147] S186, determining a plurality of relay protection device categories based on the signaling principle similarity and the signaling principle similarity threshold, wherein each relay protection device category includes a plurality of relay protection devices of different models.

[0148] In practical applications, a signaling principle similarity threshold is set, such as setting the signaling principle similarity threshold to 0.6 or 0.8. According to the signaling principle similarity and the signaling principle similarity threshold, graph clustering or hierarchical clustering methods are used to divide multiple models of relay protection devices into multiple relay protection device categories, which characterizes that the triggering logic of the set value modification signal of the relay protection devices under the same relay protection device category is similar.

[0149] S188, for each relay protection device category, based on the set of fixed value modification signals under the relay protection device category, construct a fixed value modification behavior identification sub-model corresponding to the relay protection device category, and the fixed value modification behavior identification model includes fixed value modification behavior identification sub-models corresponding to different relay protection device categories.

[0150] For each relay protection device category, the set value modification signal set under the relay protection device category is cleaned and integrated into a set value modification signal database. Exemplarily, for each set value modification sub-behavior in each set value modification set, the set value modification sub-sets corresponding to the set value modification sub-behavior are integrated to construct a set value modification signal database.

[0151] According to the characteristics of the fixed value modification signal corresponding to each fixed value modification sub-behavior, set the fixed value modification behavior category identification rules. Exemplarily, by analyzing the keywords and keyword groups in the fixed value modification signal content corresponding to each fixed value modification sub-behavior, the fixed value modification signal content is matched with the keywords, and the relay protection device to which the fixed value modification signal belongs is determined according to the matching result, thereby determining the fixed value modification behavior category corresponding to the fixed value modification signal. Based on the fixed value modification signal database and the fixed value modification behavior category identification rules, a fixed value modification behavior identification sub-model for knowledge-driven decision-making is constructed.

[0152] In this embodiment, by determining the similarity of the signaling principles of different models of relay protection devices, relay protection devices with similar principles of triggering fixed value modification signals are classified into one category. For each classified category, a fixed value modification behavior identification sub-model is constructed based on multiple fixed value modification signal sets under the category, so as to facilitate the identification of the fixed value modification behavior category for the fixed value modification signals sent by different devices, which is conducive to improving the accuracy of the fixed value modification behavior identification.

[0153] Each fixed value modification behavior category includes multiple fixed value modification sub-behaviors, and each fixed value modification signal set includes multiple fixed value modification signals related to the fixed value modification sub-behaviors. In an exemplary embodiment, S188 includes S1882 to S1888. Among them:

[0154] S1882, in each set of fixed value modification signals, the fixed value modification signal with the highest priority among the fixed value modification signals corresponding to the first fixed value modification sub-behavior is determined as the first signal.

[0155] Each fixed value modification sub-behavior contains multiple sub-operations, thereby triggering multiple fixed value modification signals. For example, for the execution of the fixed value modification sub-behavior of the security measure, multiple sub-operations are executed according to the security measure content. For example, "exit hard pressure plate XXX" and "exit soft pressure plate XXX", etc., different sub-operations have execution priorities, and the sub-operations with the same execution priority can be multiple or one.

[0156] For each set of fixed value modification signals, first determine the fixed value modification signal subset corresponding to the first fixed value modification sub-behavior, determine the execution priority based on the content of the fixed value modification signal (such as signal name) in the fixed value modification signal subset, and determine the fixed value modification signal with the highest priority as the first signal.

[0157] S1884, determining the fixed value modification signal with the highest priority among the fixed value modification signals corresponding to the last fixed value modification sub-behavior in each fixed value modification signal set as the tail signal.

[0158] Continuing from the above, for each set of fixed value modification signals, first determine the fixed value modification signal subset corresponding to the last fixed value modification sub-behavior, and determine the execution priority based on the content of the fixed value modification signal (such as signal name) in the fixed value modification signal subset. The lowest execution priority represents the highest priority, and the fixed value modification signal with the highest priority is determined as the tail signal.

[0159] S1886, collecting the first signals in the plurality of fixed value modification signal sets to obtain a first signal data set, and collecting the last signals in the plurality of fixed value modification signal sets to obtain a last signal data set.

[0160] Collect the first signals in multiple fixed value modification signal sets to obtain the first signal data set:

[0161]

[0162]

[0163] Where: The first signal data set; Modify the subbehavior for the first setting The corresponding first signal; Modify the total number of sub-behaviors for the first setting under the relay protection device category; Respectively, they are relay protection device models The corresponding fixed value modification sub-behavior The triggered setting modifies the content of the signal, Belong to the same relay protection equipment category.

[0164] Collect the tail signals in multiple fixed value modification signal sets to obtain the tail signal data set:

[0165]

[0166]

[0167] Where: is the tail signal data set; Modify the subbehavior for the last setting The corresponding tail signal; Modify the total number of sub-actions for the last setting under the relay protection device category; Respectively, they are relay protection device models Corresponding fixed value modification sub-behavior The triggered setting modifies the content of the signal, Belong to the same relay protection equipment category.

[0168] After the first signal data set and the last signal data set are obtained, the fixed value modification signals in the data set can be reordered according to the priorities of the fixed value modification signals in the data set. For the first signal data set, it is sorted in ascending order of priority, and for the last signal data set, it is sorted in descending order of priority.

[0169] S1888, based on the first signal data set, the last signal data set and the preset first signal analysis rules and the preset last signal analysis rules, a fixed value modification behavior recognition sub-model is obtained.

[0170] The preset first signal analysis rule can be to match the content of the fixed value modification signal with the content of the fixed value modification signal in the first signal data set, and determine whether the fixed value modification signal is the first signal based on the matching result; the tail signal analysis rule corresponds to the first signal analysis rule.

[0171] For each type of relay protection equipment, the first signal data set and the tail signal data set are used as the database, and the preset first signal analysis rules and tail signal analysis rules are used as the first signal and tail signal judgment decisions respectively, to construct a constant value modification behavior identification sub-model corresponding to the type of relay protection equipment and a knowledge-driven decision.

[0172] In this embodiment, firstly, for each relay protection device category, the first signal data set and the tail signal data set of the category are generated, and the first signal data set and the tail signal data set contain the characteristics of the first signal and the tail signal in the fixed value modification signal set caused by different fixed value modification behavior categories; secondly, the fixed value modification behavior identification sub-model is constructed in combination with the preset first signal analysis rules and the preset tail signal analysis rules. In this way, when identifying the fixed value modification behavior category of the fixed value modification signal, calling the corresponding fixed value modification behavior identification sub-model is conducive to improving the accuracy of identification.

[0173] For different fixed value modification signals, corresponding fixed value modification behavior identification sub-models are used to identify the fixed value modification behavior category. In an exemplary embodiment, S620 also includes S622 to S624. Among them:

[0174] S622, determining a fixed value modification behavior identification sub-model corresponding to each fixed value modification signal.

[0175] In actual applications, for each fixed value modification signal, the model of the relay protection device to which the fixed value modification signal belongs is determined according to the device name in the fixed value modification signal content, and the category of the relay protection device is determined according to the relay protection device model. Therefore, based on the category of the relay protection device, the corresponding fixed value modification behavior identification sub-model is determined.

[0176] S624, for each fixed value modification signal, calling the fixed value modification behavior identification sub-model corresponding to the fixed value modification signal to identify the fixed value modification behavior category corresponding to the fixed value modification signal.

[0177] After determining the fixed value modification behavior identification model corresponding to each fixed value modification signal, the fixed value modification signal can be first classified according to the corresponding fixed value modification behavior identification sub-model, and then sorted according to the sending time of the fixed value modification signal, and the fixed value modification behavior identification sub-model is called. The fixed value modification behavior identification sub-model is based on the first signal data set, the tail signal data set and the preset first signal analysis rules and the preset tail signal analysis rules. First, multiple first signals and tail signals in the input fixed value modification signal are identified, and different first signals correspond to different fixed value modification behavior categories, and the fixed value modification behavior category of the tail signal corresponding to the first signal is determined. Secondly, based on the signal content of the first signal and the tail signal, such as the sending time, the device name and the alarm name, the fixed value modification signal between the first signal and the tail signal is determined, and the fixed value modification behavior category of the intermediate fixed value modification signal is determined as the fixed value modification behavior category of the corresponding first signal.

[0178] In this embodiment, when identifying the fixed value modification behavior category of the fixed value modification signal, calling the fixed value modification behavior identification sub-model corresponding to the fixed value modification signal is beneficial to improving the accuracy of identification.

[0179] By using the preset first signal data set and tail signal data set, the first signal and tail signal belonging to different fixed value modification behavior categories are respectively identified from the fixed value modification signal. In an exemplary embodiment, S640 includes S642 to S644. Among them:

[0180] S642, for each fixed value modification signal, compare the fixed value modification signal with the fixed value modification signals in the preset first signal data set in positive order to determine whether the fixed value modification signal is the first signal.

[0181] In actual applications, according to the device name and alarm type in the content of the constant value modification signal, they are compared with the device name and alarm type of the constant value modification signal in the first signal data set in positive order, that is, starting from the first constant value modification signal in the first signal data set. When the device name and the alarm type match, the constant value modification signal is determined to be the first signal.

[0182] S644, for each fixed value modification signal, compare the fixed value modification signal with the fixed value modification signals in the preset tail signal data set in reverse order to determine whether the fixed value modification signal is a tail signal.

[0183] In actual applications, according to the device name and alarm type in the content of the fixed value modification signal, they are compared in reverse order, that is, starting from the last fixed value modification signal in the tail signal data set, with the device name and alarm type of the fixed value modification signal in the tail signal data set. When the device name and the alarm type match, the fixed value modification signal is determined to be a tail signal.

[0184] In the present embodiment, the first signal and the last signal belonging to different fixed value modification behavior categories are respectively identified from the fixed value modification signal through the preset first signal data set and the last signal data set, thereby improving the accuracy of determining the first signal and the last signal of different fixed value modification behavior categories. Therefore, based on the fixed value modification behavior categories corresponding to the first signal and the last signal, the fixed value modification behavior categories of other fixed value modification signals can be determined, which is beneficial to improving the efficiency of identifying the fixed value modification behavior.

[0185] In order to make a clearer description of the method for identifying the setting modification behavior of a relay protection device provided in the present application, a specific embodiment is described below. The specific embodiment includes the following steps:

[0186] S1, obtain the alarm signal of the secondary monitoring main station of the power system, take the alarm signal as input, call the trained signal recognition model, identify whether the alarm signal is a fixed value modification signal sent by the relay protection device, and obtain the signal recognition result.

[0187] S2. When the signal recognition result indicates that the alarm signal contains multiple constant value modification signals, determine the constant value modification behavior identification sub-model corresponding to each constant value modification signal. For each constant value modification signal, call the constant value modification behavior identification sub-model corresponding to the constant value modification signal. For each constant value modification signal, compare the constant value modification signal with the constant value modification signals in the preset first signal data set in positive order to determine whether the constant value modification signal is the first signal. For each constant value modification signal, compare the constant value modification signal with the constant value modification signals in the preset tail signal data set in reverse order to determine whether the constant value modification signal is the tail signal.

[0188] S3, based on the head signal and tail signal corresponding to each fixed value modification behavior category, construct a fixed value modification signal set corresponding to the fixed value modification behavior category, divide the fixed value modification signals under the same fixed value modification behavior category into the same fixed value modification signal set, and determine the fixed value modification behavior category corresponding to each fixed value modification signal.

[0189] In this embodiment, before obtaining the alarm signal of the power system secondary monitoring master station, a signal recognition model and a fixed value modification behavior category recognition model are constructed. The construction process of the signal recognition model and the fixed value modification behavior category recognition model is as follows:

[0190] S101, obtaining historical alarm signals from the secondary monitoring master station of the power system, determining signal category labels of the historical alarm signals, performing feature selection and feature processing on the historical alarm signals after data annotation, and obtaining initial training samples.

[0191] S102, performing data enhancement on the initial training samples.

[0192] The initial training sample includes multiple alarm signal sequences, and data enhancement may include the following methods: first, for each alarm signal sequence, select an alarm signal segment in the alarm signal sequence, generate an analog alarm signal that matches the alarm signal in the alarm signal segment, and interpolate the analog alarm signal into the alarm signal sequence to obtain a first training sample; second, for each alarm signal sequence, reorder the alarm signals sent at the same time in the alarm signal sequence to obtain a second training sample; third, for each alarm signal sequence, assign a weight to each alarm signal in the alarm signal sequence, determine the missing analog signal set based on the weight of each alarm signal, remove the alarm signals in the missing analog signal set from the alarm signal sequence, and generate a third training sample.

[0193] S103, connecting the first deep learning model and the second deep learning model in series to construct a signal recognition model, and inputting the target training sample into the constructed signal recognition model, so that the signal recognition model learns the semantic features of each historical alarm signal in the target training sample and the timing information between different historical alarm signals, and determines the signal category of each historical alarm signal.

[0194] S104, based on the signal category label of each historical alarm signal in the target training sample and the signal category of the output historical alarm signal, determine the loss function value of the signal recognition model, and based on the loss function value, iteratively update the parameters of the signal recognition model until the preset training end condition is reached to obtain the trained signal recognition model.

[0195] S105, obtaining a set of fixed value modification signals of multiple models of relay protection devices, determining the similarity of the signaling principles of any two different models of relay protection devices, and determining multiple relay protection device categories based on the signaling principle similarity and the signaling principle similarity threshold, wherein each relay protection device category includes multiple different models of relay protection devices.

[0196] S106, each fixed value modification behavior category includes multiple fixed value modification sub-behaviors, and each fixed value modification signal set includes multiple fixed value modification signals related to the fixed value modification sub-behaviors. For each relay protection device category, the fixed value modification signal with the highest priority among the fixed value modification signals corresponding to the first fixed value modification sub-behavior in each fixed value modification signal set is determined as the first signal, and the fixed value modification signal with the highest priority among the fixed value modification signals corresponding to the last fixed value modification sub-behavior in each fixed value modification signal set is determined as the last signal. The first signals in multiple fixed value modification signal sets are aggregated to obtain a first signal data set, and the tail signals in multiple fixed value modification signal sets are aggregated to obtain a tail signal data set. Based on the first signal data set, the tail signal data set, the preset first signal analysis rule, and the preset tail signal analysis rule, a constant value modification behavior identification submodel is obtained.

[0197] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0198] In an exemplary embodiment, Figure 7 As shown, a relay protection device setting modification behavior identification device 600 is provided, comprising: a signal acquisition module 610, a signal category identification module 620 and a setting modification behavior category identification module 630, wherein:

[0199] The signal acquisition module 610 is used to acquire the alarm signal of the secondary monitoring master station of the power system.

[0200] The signal category identification module 620 is used to take the alarm signal as input, call the trained signal identification model, identify whether the alarm signal is a fixed value modification signal sent by the relay protection device, and obtain a signal identification result. The trained signal identification model is trained based on historical alarm signals carrying fixed value modification signal category labels.

[0201] The fixed value modification behavior category identification module 630 is used to identify the fixed value modification behavior category corresponding to each fixed value modification signal when the signal identification result indicates that the alarm signal includes multiple fixed value modification signals.

[0202] In an exemplary embodiment, the constant value modification behavior category identification module 630 is also used to call the constructed constant value modification behavior identification model to identify the constant value modification behavior category corresponding to each constant value modification signal; the constant value modification behavior identification model is obtained based on a set of constant value modification signals corresponding to multiple different constant value modification behavior categories and preset constant value modification behavior category identification rules.

[0203] In an exemplary embodiment, the constant value modification behavior category identification module 630 is also used to filter out the first signal and the last signal corresponding to different constant value modification behavior categories from multiple constant value modification signals; construct a set of constant value modification signals corresponding to the fixed value modification behavior category based on the first signal and the last signal corresponding to each fixed value modification behavior category; divide the constant value modification signals under the same fixed value modification behavior category into the same set of constant value modification signals, and determine the constant value modification behavior category corresponding to each fixed value modification signal.

[0204] In an exemplary embodiment, the relay protection device setting modification behavior identification device 600 further includes a signal recognition model training module 640 .

[0205] The signal acquisition module 610 is also used to acquire historical alarm signals from the secondary monitoring master station of the power system.

[0206] The signal recognition model training module 640 is used to determine the signal category label of the historical alarm signal; perform feature selection and feature processing on the historical alarm signal after data annotation to obtain an initial training sample; perform data enhancement on the initial training sample to obtain a target training sample; connect the first deep learning model and the second deep learning model in series to construct a signal recognition model; input the target training sample into the constructed signal recognition model so that the signal recognition model learns the semantic features of each historical alarm signal in the target training sample and the timing information between different historical alarm signals to determine the signal category of each historical alarm signal; based on the signal category label of each historical alarm signal in the target training sample and the signal category of the output historical alarm signal, determine the loss function value of the signal recognition model; based on the loss function value, iteratively update the parameters of the signal recognition model until the preset training end condition is reached to obtain the trained signal recognition model.

[0207] In an exemplary embodiment, the signal recognition model training module 640 is also used to select an alarm signal segment in the alarm signal sequence for each alarm signal sequence, generate an analog alarm signal that matches the alarm signal in the alarm signal segment, and interpolate the analog alarm signal into the alarm signal sequence to obtain a first training sample; for each alarm signal sequence, reorder the alarm signals sent at the same time in the alarm signal sequence to obtain a second training sample; for each alarm signal sequence, assign a weight to each alarm signal in the alarm signal sequence, determine the missing analog signal set based on the weight of each alarm signal, and remove the alarm signals in the missing analog signal set from the alarm signal sequence to generate a third training sample.

[0208] In an exemplary embodiment, the relay protection device setting value modification behavior identification device 600 further includes a setting value modification behavior identification model construction module 650. Among them:

[0209] The signal acquisition module 610 is also used to acquire a set of fixed value modification signals of multiple models of relay protection devices.

[0210] The constant value modification behavior identification model construction module 650 is used to determine the similarity of the signaling principles of any two different models of relay protection devices; based on the signaling principle similarity and the signaling principle similarity threshold, multiple relay protection device categories are determined, wherein each relay protection device category includes multiple different models of relay protection devices; for each relay protection device category, based on the set of constant value modification signals under the relay protection device category, a constant value modification behavior identification sub-model corresponding to the relay protection device category is constructed, and the constant value modification behavior identification model includes constant value modification behavior identification sub-models corresponding to different relay protection device categories.

[0211] In an exemplary embodiment, the fixed value modification behavior recognition model construction module 650 is also used to determine the fixed value modification signal with the highest priority among the fixed value modification signals corresponding to the first fixed value modification sub-behavior in each fixed value modification signal set as the first signal; determine the fixed value modification signal with the highest priority among the fixed value modification signals corresponding to the last fixed value modification sub-behavior in each fixed value modification signal set as the last signal; aggregate the first signals in multiple fixed value modification signal sets to obtain a first signal data set, and aggregate the tail signals in multiple fixed value modification signal sets to obtain a tail signal data set; based on the first signal data set, the tail signal data set and the preset first signal analysis rule and the preset tail signal analysis rule, obtain the fixed value modification behavior recognition sub-model.

[0212] In an exemplary embodiment, the constant value modification behavior category identification module 630 is also used to determine the constant value modification behavior identification sub-model corresponding to each constant value modification signal; for each constant value modification signal, the constant value modification behavior identification sub-model corresponding to the fixed value modification signal is called to identify the constant value modification behavior category corresponding to the fixed value modification signal.

[0213] In an exemplary embodiment, the constant value modification behavior category identification module 630 is also used to, for each constant value modification signal, perform a positive comparison between the constant value modification signal and the constant value modification signal in a preset first signal data set to determine whether the constant value modification signal is the first signal; for each constant value modification signal, perform a reverse comparison between the constant value modification signal and the constant value modification signal in a preset tail signal data set to determine whether the constant value modification signal is the tail signal.

[0214] Each module in the above relay protection device setting modification behavior identification device 600 can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0215] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store alarm signals. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for identifying the setting modification behavior of a relay protection device is implemented.

[0216] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0217] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps in any one of the above-mentioned embodiments of the method for identifying the setting modification behavior of a relay protection device are implemented.

[0218] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in any one of the above-mentioned embodiments of the method for identifying the setting modification behavior of a relay protection device are implemented.

[0219] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps in any one of the above-mentioned embodiments of the method for identifying the setting modification behavior of a relay protection device.

[0220] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0221] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0222] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0223] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for identifying the setting modification behavior of a relay protection device, characterized in that: The method comprises: Obtain the alarm signal from the secondary monitoring master station of the power system; Taking the alarm signal as input, calling a trained signal recognition model, identifying whether the alarm signal is a fixed value modification signal sent by a relay protection device, and obtaining a signal recognition result, wherein the trained signal recognition model is trained based on historical alarm signals carrying a fixed value modification signal category label; In a case where the signal identification result indicates that the alarm signal includes a plurality of fixed value modification signals, a fixed value modification behavior category corresponding to each fixed value modification signal is identified.

2. The method according to claim 1, characterized in that The identifying the fixed value modification behavior category corresponding to each fixed value modification signal includes: Calling the constructed fixed value modification behavior identification model to identify the fixed value modification behavior category corresponding to each fixed value modification signal; The fixed value modification behavior identification model is obtained based on fixed value modification signal sets corresponding to different fixed value modification behavior categories and preset fixed value modification behavior category identification rules.

3. The method according to claim 1 or 2, characterized in that: The identifying the fixed value modification behavior category corresponding to each fixed value modification signal also includes: Filtering out the first signal and the last signal corresponding to different fixed value modification behavior categories from the plurality of fixed value modification signals; Based on the first signal and the last signal corresponding to each fixed value modification behavior category, construct a fixed value modification signal set corresponding to the fixed value modification behavior category; The fixed value modification signals under the same fixed value modification behavior category are divided into the same fixed value modification signal set, and the fixed value modification behavior category corresponding to each of the fixed value modification signals is determined.

4. The method according to claim 2, characterized in that: Before obtaining the alarm signal from the power system secondary monitoring master station, the method further includes: Obtaining historical alarm signals from the power system secondary monitoring master station; Determining a signal category label of the historical alarm signal; Perform feature selection and feature processing on the historical alarm signals after data annotation to obtain initial training samples; Performing data enhancement on the initial training samples to obtain target training samples; Connecting the first deep learning model and the second deep learning model in series to construct a signal recognition model; Inputting the target training sample into the constructed signal recognition model, so that the signal recognition model learns the semantic features of each historical alarm signal in the target training sample and the timing information between different historical alarm signals, and determines the signal category of each of the historical alarm signals; Determining a loss function value of the signal recognition model based on the signal category label of each historical alarm signal in the target training sample and the signal category of the output historical alarm signal; Based on the loss function value, the parameters of the signal recognition model are iteratively updated until a preset training end condition is reached to obtain a trained signal recognition model.

5. The method according to claim 4, characterized in that The initial training sample includes a plurality of warning signal sequences, and the data enhancement of the initial training sample includes at least one of the following methods: The first method is to select an alarm signal segment in the alarm signal sequence for each alarm signal sequence, generate an analog alarm signal matching the alarm signal in the alarm signal segment, and interpolate the analog alarm signal into the alarm signal sequence to obtain a first training sample; The second method is to reorder the alarm signals sent at the same time in each alarm signal sequence to obtain a second training sample; The third method is to assign a weight to each alarm signal in the alarm signal sequence for each alarm signal sequence, determine a missing analog signal set based on the weight of each alarm signal, remove the alarm signals in the missing analog signal set from the alarm signal sequence, and generate a third training sample.

6. The method according to claim 2, characterized in that Before obtaining the alarm signal from the power system secondary monitoring master station, the method further includes: Obtain the set value modification signal set of multiple models of relay protection equipment; Determine the similarity of the signaling principles of any two different types of relay protection devices; Based on the signaling principle similarity and the signaling principle similarity threshold, determining a plurality of relay protection device categories, wherein each of the relay protection device categories includes a plurality of relay protection devices of different models; For each of the relay protection device categories, based on the set of constant value modification signals under the relay protection device category, a constant value modification behavior identification sub-model corresponding to the relay protection device category is constructed, and the constant value modification behavior identification model includes constant value modification behavior identification sub-models corresponding to different relay protection device categories.

7. The method according to claim 6, characterized in that Each of the fixed value modification behavior categories includes a plurality of fixed value modification sub-behaviors, and each of the fixed value modification signal sets includes fixed value modification signals associated with the plurality of fixed value modification sub-behaviors; The constructing of a setting modification behavior identification sub-model corresponding to the relay protection device category based on the setting modification signal set under the relay protection device category includes: Determine the fixed value modification signal with the highest priority among the fixed value modification signals corresponding to the first fixed value modification sub-behavior in each of the fixed value modification signal sets as the first signal; Determine the fixed value modification signal with the highest priority among the fixed value modification signals corresponding to the last fixed value modification sub-behavior in each of the fixed value modification signal sets as the tail signal; Collecting the first signals in the plurality of fixed value modification signal sets to obtain a first signal data set, and collecting the last signals in the plurality of fixed value modification signal sets to obtain a last signal data set; Based on the first signal data set, the last signal data set, the preset first signal analysis rule, and the preset last signal analysis rule, the fixed value modification behavior identification sub-model is obtained.

8. The method according to claim 6 or 7, characterized in that: The identifying the fixed value modification behavior category corresponding to each fixed value modification signal also includes: Determine a fixed value modification behavior identification sub-model corresponding to each fixed value modification signal; For each fixed value modification signal, a fixed value modification behavior identification sub-model corresponding to the fixed value modification signal is called to identify the fixed value modification behavior category corresponding to the fixed value modification signal.

9. The method according to claim 3, characterized in that: The step of selecting the first signal and the last signal corresponding to different fixed value modification behavior categories from the plurality of fixed value modification signals comprises: For each fixed value modification signal, the fixed value modification signal is compared with the fixed value modification signal in the preset first signal data set in positive sequence to determine whether the fixed value modification signal is the first signal; For each fixed value modification signal, the fixed value modification signal is compared in reverse order with the fixed value modification signals in a preset tail signal data set to determine whether the fixed value modification signal is a tail signal.

10. A device for identifying setting value modification behavior of relay protection equipment, characterized in that: The device comprises: A signal acquisition module is used to obtain the alarm signal of the secondary monitoring master station of the power system; A signal category identification module, used to take the alarm signal as input, call a trained signal identification model, identify whether the alarm signal is a fixed value modification signal sent by a relay protection device, and obtain a signal identification result, wherein the trained signal identification model is trained based on historical alarm signals carrying fixed value modification signal category labels; The fixed value modification behavior category identification module is used to identify the fixed value modification behavior category corresponding to each fixed value modification signal when the signal identification result indicates that the alarm signal includes multiple fixed value modification signals.

Images