Method and system for intelligent research and judgment of monitoring alarm signals, and storage medium

Abstract

The application discloses a kind of monitoring alarm signal intelligent research and judge method, device, electronic equipment and storage medium, belong to data processing technical field, including steps S1: obtaining abnormal data, abnormal data include the running state of each fault type associated power equipment, and corresponding operation data and result data;Step S2: establish prediction model and auxiliary model, prediction model is divided into multiple prediction groups;Step S3: based on alarm signal type selection corresponding prediction group, prediction group exports fault type to each auxiliary model, and auxiliary model exports operation data and result data;Step S4: the output result is scored, and based on the scoring result, fault type, operation data and result data are filled into the first table to convert the first table into the auxiliary decision table for alarm signal.It can not only identify corresponding fault according to alarm signal according to the application, and can give corresponding solution according to fault.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically relating to a method, system, and storage medium for intelligent analysis of monitoring alarm signals. Background Technology

[0002] Currently, when abnormal situations occur in the power grid, monitoring alarm signals are generated. Power grid dispatchers can then perform corresponding operations based on these alarm signals to maintain the safe and stable operation of the power grid. However, after an alarm signal is generated, monitoring personnel still need to manually identify, analyze, and respond to each piece of information in the alarm signal. As the power grid continues to expand and its intelligence level improves, the number of power devices and faults that can be monitored is increasing, resulting in a greater number of monitoring alarm signals. This leads to the problem that manually verifying each monitoring alarm signal is both time-consuming and prone to errors and omissions when monitoring personnel lack experience. To address this issue and assist monitoring personnel in making judgments, the following technical solutions have been proposed in the existing technology.

[0003] For example, Chinese patent application "CN201910563074.9" discloses a method, system, and readable storage medium for classifying power grid monitoring alarm information. This method acquires raw power grid monitoring alarm information, preprocesses the acquired raw power grid monitoring alarm information to obtain monitoring alarm information, then extracts monitoring alarm event samples containing monitoring alarm information from the monitoring alarm information, constructs a training sample library, and quantifies the text information in each monitoring alarm event sample to obtain the feature vector of the monitoring alarm event sample. Finally, based on the training sample library and the feature vector of the monitoring alarm event samples, a machine learning model is established to obtain the monitoring alarm event classification result of the monitoring alarm information. This method, by classifying monitoring alarm events, facilitates centralized signal processing by monitoring personnel. For example, Chinese patent application "CN202010058261.4" discloses a method for identifying power grid monitoring alarm events based on convolutional and long short-term memory networks. This method generates information vectors by using historical monitoring alarm information and time stamps in the power grid monitoring system, and extracts event samples from the collected historical monitoring alarm information to construct an alarm event sample library. Secondly, it establishes a deep learning recognition model based on a combination of long short-term memory networks and convolutional neural networks, and trains the model using alarm event samples. Finally, it uses the trained deep learning model to identify monitoring alarm information, and takes the event category with the highest probability as the recognition result output, thereby achieving rapid identification of power grid alarm events.

[0004] However, the above methods only classify and identify alarms. In fact, after identifying the cause of the alarm signal, it is necessary to take corresponding measures to eliminate the alarm signal. The current technology does not disclose the corresponding technical solution. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a method, system, and storage medium for intelligent analysis of monitoring alarm signals, thereby resolving the issues in the prior art.

[0006] To achieve the aforementioned objectives, this invention proposes an intelligent analysis method for monitoring alarm signals, comprising: Step S1: Obtain historical data, which includes normal data and abnormal data. The abnormal data includes the operating status of power equipment associated with each fault type, fault data, retrospective data, operation data and result data corresponding to each fault type, as well as the first moment, second moment, third moment and fourth moment generated respectively, wherein the second moment is before the first moment, the third moment is after the first moment, and the fourth moment is after the third moment. Step S2: Establish multiple prediction models and auxiliary models. The prediction model is used to output the fault type, and the auxiliary model is used to output the operation data and the result data. Based on the type of alarm signal, the prediction model is divided into multiple prediction groups, and each prediction group includes multiple prediction models. Step S3: After the alarm signal is generated, the corresponding prediction group is selected based on the type of the alarm signal. The prediction model in the prediction group outputs the fault type that generated the alarm signal based on the fault data and the backtracking data corresponding to the alarm signal. The fault types output by multiple prediction models and the operating status of associated power equipment are merged and input into each auxiliary model to obtain multiple types of operation data and the result data corresponding to each type of operation data. Step S4: Establish a first table, score the output results of the prediction model and the auxiliary model, and fill the first table with the fault type, the operation data and the result data in descending order of the score results, so as to transform the first table into an auxiliary decision table for the alarm signal.

[0007] Furthermore, in step S2, the auxiliary model is established based on the following steps: A second table is created, and each fault type, the occurrence time of the fault type, the operating status of the power equipment associated with the fault type, and the corresponding operation data and result data are filled into the second table from the historical data. The operation data includes operation procedures arranged in chronological order for different power equipment. Set a first threshold, define the operation data in the second table that are of the same fault type and whose occurrence time interval is less than the first threshold as first data. If the first data and the operating status of each power device corresponding to the first data are not completely the same, define the fault type corresponding to the first data as the first type. Obtain the occurrence time corresponding to each first type and define the earliest occurrence time as the first time. New data is generated in the second table to transform the second table into a third table. The auxiliary model is established based on the data in the third table. The fault type in the new data is the first type, the occurrence time is the first time, the operating status of the associated power equipment and the operation data are generated by fusing the first data and the corresponding operating status of the power equipment. The result data is the result data corresponding to the first time.

[0008] Furthermore, in step S4, the output results of the prediction model and the auxiliary model are scored based on the following steps: Set basic values ​​for each power device. Based on the power devices included in the operational data and their corresponding basic values, calculate a first score for each operational data point. Based on the importance of each fault type, set a second score for each fault type. If the prediction model and the auxiliary model produce n output results, calculate the score for each output result based on the first formula. The first formula is: ,in, and These are the first score of the operation data i and the second score of the fault type j, respectively. Output the total number of operation data i for all the auxiliary models. Output the total number of fault types J for all the predicted models.

[0009] Furthermore, the auxiliary model outputs the operation data based on the following steps: The fault type that appears in the third table in a fixed order and whose first occurrence count is greater than the second threshold is defined as the second data. The second data in the first and last order within the same round and the area between them are defined as the attention area. A third threshold is set. If the second occurrence count of the fault type is greater than the third threshold in all the attention areas, then the fault type is set as a derived fault. Calculate the state transition probability based on the second formula. The second formula is: Where M is the first number of times, m is the second number of times, and a fourth threshold is set. If the state transition probability is greater than the fourth threshold, the auxiliary model outputs the second data and the operation data of the derived fault; otherwise, it only outputs the operation data corresponding to the second data.

[0010] Furthermore, in step S2, the prediction groups are divided based on the following steps: Based on the historical data, each prediction model is tested to obtain the prediction accuracy of each prediction model for the alarm signal. A fourth threshold is set, and prediction models with prediction accuracy greater than the fourth threshold are obtained and assigned to the prediction group corresponding to the alarm signal.

[0011] This invention also provides an intelligent analysis device for monitoring alarm signals. This system is used to implement the aforementioned intelligent analysis method for monitoring alarm signals. The system mainly includes: The data acquisition unit is used to acquire historical data, which includes normal data and abnormal data. The abnormal data includes the operating status of power equipment associated with each fault type, fault data, retrospective data, operation data and result data corresponding to each fault type, as well as the first moment, second moment, third moment and fourth moment generated respectively, wherein the second moment is located before the first moment, the third moment is located after the first moment, and the fourth moment is located after the third moment. A storage unit is used to store a prediction model and an auxiliary model. The prediction model is used to output the fault type, and the auxiliary model is used to output the operation data and the result data. Based on the type of alarm signal, the prediction model is divided into multiple prediction groups, and each prediction group includes multiple prediction models. The calculation unit, after generating the alarm signal, selects the corresponding prediction group based on the type of the alarm signal. The prediction model in the prediction group outputs the fault type that generated the alarm signal based on the fault data and the retrospective data corresponding to the alarm signal. The calculation unit merges the fault types output by multiple prediction models and the operating status of associated power equipment into each auxiliary model to obtain multiple types of operation data and the result data corresponding to each type of operation data. The display unit scores the output results of the prediction model and the auxiliary model, and fills the fault type, the operation data and the result data into a pre-generated first table based on the score results from largest to smallest, so as to transform the first table into an auxiliary decision table for the alarm signal.

[0012] The present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program stored in the memory to perform the above-described intelligent analysis method for monitoring alarm signals.

[0013] The present invention also provides a computer storage medium storing program instructions, wherein the program instructions, when executed, control the device where the computer storage medium is located to execute the above-described intelligent judgment method for monitoring and alarm signals.

[0014] Compared with the prior art, the beneficial effects of the present invention are at least as follows: This invention first establishes multiple prediction models and auxiliary models. When an alarm signal is generated, the prediction model can output the fault type of the alarm signal based on the normal data and fault data of the power grid. The auxiliary model can generate the operation method for eliminating the alarm signal based on the fault type output by the prediction model, as well as the result of using the operation method. This helps relevant monitoring personnel make decisions and reduces the operational burden on personnel.

[0015] This invention also establishes multiple prediction models and auxiliary models. By using multiple prediction models to predict and judge the same alarm signal, the accuracy of the model judgments can be mutually verified. In addition, before prediction, different prediction models are classified and grouped according to the alarm signal. After the alarm signal is generated, the multiple prediction models with the highest accuracy are selected to judge the fault type of the alarm signal, thereby further improving the accuracy of the judgment. Finally, this invention also scores the operational data output by the auxiliary model and sorts the operational data according to the scoring results, thereby prioritizing the display of reliable data to monitoring personnel and further helping them to quickly process alarm signals. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of intelligent analysis of monitoring alarm signals according to the present invention. Figure 2 This is a schematic diagram illustrating the generation of the third table in this invention; Figure 3 This is a schematic diagram illustrating the output results generated by the prediction model and auxiliary model of this invention; Figure 4 This is a schematic diagram of the structure of an intelligent analysis device for monitoring alarm signals according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] It is understood that the terms "first," "second," etc., used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are used only to distinguish one element from another. For example, without departing from the scope of this application, a first script may be referred to as a second script, and similarly, a second script may be referred to as a first script.

[0019] like Figure 1 As shown, an intelligent analysis method for monitoring alarm signals includes: Step S1: Obtain historical data, which includes normal data and abnormal data. Abnormal data includes the operating status of power equipment associated with each fault type, fault data, backtracking data, operation data and result data corresponding to each fault type, as well as the first moment, second moment, third moment and fourth moment generated respectively. The second moment is before the first moment, the third moment is after the first moment, and the fourth moment is after the third moment.

[0020] Historical data refers to data that has already occurred and been processed. Specifically, after the power monitoring system identifies abnormal data in the power grid, it issues an alarm signal. The time when the abnormal data occurs is defined as the fault moment. The first moment of the retrospective data is before the fault moment, for example, data from one hour before the fault moment. By capturing retrospective data for a certain period of time, the system can help analyze the cause of the fault. Retrospective data, along with fault data, includes data such as current, voltage, and phase sequence. Operational data refers to data generated after the fault moment when monitoring personnel operate the equipment. Specifically, it includes the order, name, time, and specific operation content of each power device operated by the monitoring personnel. Result data is the result data fed back by the power grid after the monitoring personnel's operation, including how much time has passed since the operation and which power devices' data have changed. Based on the result data, it can be determined whether the operation was effective and quickly restored the power grid to normal.

[0021] Step S2: Establish multiple prediction models and auxiliary models. The prediction model is used to output the fault type, and the auxiliary model is used to output the operation data and result data. Based on the type of alarm signal, the prediction model is divided into multiple prediction groups, and each prediction group includes multiple prediction models.

[0022] The prediction model can be built based on different machine learning models, such as BP neural networks, RBF neural networks, or k-means clustering algorithms. Historical data is then divided into training and testing sets to train the prediction model and auxiliary models. During testing, the prediction accuracy of the model for various alarm signals is obtained, specifically by the ratio of the number of data points correctly judged by the model to the total number of test data points. Furthermore, this embodiment divides the prediction model into multiple prediction groups for different alarm signals. For example, for current fluctuation alarm signals, models based on BP neural networks and RBF neural networks are more accurate in predicting based on the fluctuation of input current data, so these two models are placed in the same prediction group. Similarly, models based on CNN networks and SVM networks are more accurate in predicting based on the waveform curve of input current data, so these two models are also placed in the same prediction group. Therefore, by assigning different prediction models to different alarm signals through this step, the prediction accuracy can be improved. In addition, this embodiment also sets up multiple auxiliary models, which identify and judge the same set of data, thereby further ensuring the reliability of the operational data output results.

[0023] Step S3: After an alarm signal is generated, the corresponding prediction group is selected based on the type of the alarm signal. The prediction model in the prediction group outputs the fault type that generated the alarm signal based on the fault data and backtracking data corresponding to the alarm signal. The fault types output by multiple prediction models and the operating status of related power equipment are merged and input into each auxiliary model to obtain various operation data and the result data corresponding to each operation data.

[0024] Specifically, after an alarm signal is generated, such as a protection action alarm signal, the fault data that generated the alarm signal and the retrospective data from a period of time before the alarm signal was generated are collected and input into the prediction model in the pre-set prediction group. The prediction model outputs the reason for the alarm signal, that is, the corresponding fault type. In particular, the historical data in this embodiment includes not only the fault type, but also the time of the fault occurrence and the operating status of the power equipment associated with the fault type. Therefore, after the prediction model generates the fault type, the fault type and the current operating status of the power equipment associated with the fault type are input into each auxiliary model. The auxiliary model outputs the operation data for the fault type and the corresponding result data after using the operation data.

[0025] Step S4: Create a first table, score the output results of the prediction model and the auxiliary model, and fill the first table with the fault type, operation data and result data from largest to smallest based on the score results, so as to transform the first table into an auxiliary decision table for alarm signals.

[0026] By creating a first table and filling in the corresponding data, the monitoring personnel are shown the types of faults that may cause the alarm signal, the corresponding troubleshooting methods, and how long it will take to restore normal operation after using the method, for their reference. In addition, by scoring and sorting the operational data, the monitoring personnel can prioritize seeing the more reliable data. The specific scoring method will be described later.

[0027] This invention first establishes multiple prediction models and auxiliary models. When an alarm signal is generated, the prediction model can output the fault type of the alarm signal based on the normal data and fault data of the power grid. The auxiliary model can generate the operation method for eliminating the alarm signal based on the fault type output by the prediction model, as well as the result of using the operation method. This helps relevant monitoring personnel make decisions and reduces the operational burden on personnel.

[0028] This invention also establishes multiple prediction models and auxiliary models. By using multiple prediction models to predict and judge the same alarm signal, the accuracy of the model judgments can be mutually verified. In addition, before prediction, different prediction models are classified and grouped according to the alarm signal. After the alarm signal is generated, the multiple prediction models with the highest accuracy are selected to judge the fault type of the alarm signal, thereby further improving the accuracy of the judgment. Finally, this invention also scores the operational data output by the auxiliary model and sorts the operational data according to the scoring results, thereby prioritizing the display of reliable data to monitoring personnel and further helping them to quickly process alarm signals.

[0029] It is particularly noteworthy that this invention can not only identify the corresponding fault based on the alarm signal, but also provide a corresponding solution based on the fault.

[0030] In step S2, an auxiliary model is established based on the following steps.

[0031] Create a second table and fill it with each type of fault from the historical data, the time of occurrence of the fault type, the operating status of the power equipment associated with the fault type, and the corresponding operation data and result data. The operation data includes operation procedures arranged in chronological order for different power equipment.

[0032] like Figure 2 The diagram shown is a simplified illustration of the second table. This invention first categorizes historical data and fills it into the second table as training and testing data. The second table includes fault type, occurrence time, operating status of power equipment related to the fault type, operation data, and result data. Operating status refers to the state of the power equipment itself; for example, when the power equipment is a switch, its operating status is open or closed. Operation data is the sequence of operations for each power equipment; for example, when a fault occurs, the state of the corresponding equipment is changed sequentially. It is worth noting that the fault type in this invention is a fault type specific to a certain area. For example, if a tripping fault A1 occurs in area A, since area A includes multiple power equipment, different faults in different power equipment may lead to tripping. Therefore, the operation data needs to be determined according to the actual situation. For example, if the fault type is A1 and the operating status is y2, y2, and y2, operation sequence 1 is to operate power equipment a1 and a3 sequentially. If the fault type is also A1, but the operating status is y1, y2, and y2, operation sequence 2 is to operate power equipment a2 and a3 sequentially.

[0033] Set a first threshold, define the operation data of the same fault type in the second table and whose occurrence time interval is less than the first threshold as the first data. If the first data and the operating status of each power equipment corresponding to the first data are not completely the same, then define the fault type corresponding to the first data as the first type. Obtain the occurrence time corresponding to each first type, and define the earliest occurrence time as the first time.

[0034] New data is generated in the second table to transform the second table into the third table. An auxiliary model is built based on the data in the third table. The fault type in the new data is the first type, the occurrence time is the first time, the operating status of the associated power equipment and the operation data are the first data and the corresponding operating status of the power equipment, and the result data is the result data corresponding to the first time.

[0035] In practical applications, historical data cannot possibly include data for all situations. Therefore, this invention selects multiple data from historical data and merges them to generate new data through the above steps. The newly generated data has never appeared in the historical data. So when the corresponding fault type occurs, even if there is no corresponding operation in the collected historical data, the auxiliary model can still provide corresponding solutions.

[0036] The following is combined Figure 2 The above process will be explained in this embodiment. The first threshold is set to 1 hour. Setting this first threshold prevents the auxiliary model from fusing too much data into new data, thus avoiding an excessively large and unrealistic newly generated dataset. Figure 2 In the table, the occurrence times of the two fault types A1 are 12:30 and 13:20, respectively, with an interval of less than one hour. Therefore, the operation data corresponding to the above two fault types is defined as the first data. Then, if the two first data and the corresponding operating states are not completely the same, the fault type A1 corresponding to the first data is defined as the first type. Here, fault types with different operating states but the same operation data are excluded because the same operation data is taken under different operating states, indicating that the operation data is unrelated to the operating state. However, if the operating state is the same but the operation data is different, it indicates that this kind of fault requires handling measures to change according to the on-site situation and has an accidental nature. Then, the occurrence time of the first fault type A1 in the table, 12:30, is defined as the first time.

[0037] Then, new data is generated using the data obtained above, such as... Figure 2 As shown in the dashed box, in the newly generated data, the fault type is A1, the occurrence time is 12:30, and the equipment operating status and operation data are formed by merging two sets of first data and their corresponding operating statuses. Specifically, the merging process involves replacing two different sets of operating statuses to generate the data. For example... Figure 2The two operating states can be merged into y2, y2, y2, and y1, y1, y2, etc. The operation sequence is generated based on this method, according to the fusion method of the operating states, which will not be elaborated here.

[0038] In step S4 of this embodiment, the output results of the prediction model and the auxiliary model are scored based on the following steps.

[0039] Set the base values ​​for each power device. Based on the power devices and their corresponding base values ​​included in the operational data, calculate the first score for each operational data point. Based on the importance of each fault type, set a second score for each fault type. If the prediction model and auxiliary model produce n output results, calculate the score for each output result based on the first formula. The first formula is: ,in, and These represent the first score for operation data i and the second score for fault type j, respectively. Output the total number of operation data i for all auxiliary models. Output the total number of fault types J for all prediction models.

[0040] The steps described above are explained below. First, based on the importance of each power device, a base value is set for each device. For example, power device a1 is of high importance and should be operated with caution or with relevant approvals, so its base value is set lower. Power device a2 is of lower importance, so its base value is set higher. Similarly, this invention sets a second score for each fault type based on the degree of impact of the fault type. For example, if the fault has a large impact and is urgent, its second score is set lower, because the handling of faults with a large impact should also be judged based on the actual situation. If the fault has a small impact and is not urgent, then the second score for the fault type is set higher. In addition, when calculating the first score, the base values ​​of each power device in the operation data are accumulated. For example, if the operation data requires operating devices a1 and a2 in sequence, then the base values ​​corresponding to devices a1 and a2 are added together to obtain the first score of the operation data.

[0041] The following is combined Figure 3 To explain the first formula, for example, a prediction group has three prediction models. When data is input into prediction model 1, prediction model 2, and prediction model 3 respectively, two of the prediction models output fault type A1, and one prediction model outputs fault type A2. The total number of fault types A1 is 2, and the total number of fault types A2 is 1. Then, if fault type A1 and the associated power equipment are input into auxiliary models 1, 2, and 3, two operation sequences 1 and one operation sequence 3 are obtained, as follows... Figure 3As shown by the solid line, similarly, after inputting fault type A2 and related power equipment into auxiliary model 1, auxiliary model 2 and auxiliary model 3, two operation sequences 2 and one operation sequence 4 are obtained, as shown by the dashed line in the figure. Thus, four output results will be generated according to the alarm signal, namely (fault type A1, operation sequence 1), (fault type A1, operation sequence 2), (fault type A2, operation sequence 2), and (fault type A2, operation sequence 4). Then, the above four results need to be scored.

[0042] Taking the first output as an example, when calculating the score, the number of fault type A1 and the number of operation sequence 1 are both obtained, which are 2. Therefore, in the first formula... and Both scores are 2. Then, by obtaining the first and second scores for each, the first result can be calculated. Similarly, based on this process, the scores for fault type A1 using operation sequence 2, fault type A2 using operation sequence 3, and fault type A2 using operation sequence 4 are calculated respectively, and sorted from high to low according to the score, so as to be displayed to the monitoring personnel.

[0043] Therefore, the higher the score calculated through the above steps, the more models output the result, which proves that the result is more reliable.

[0044] In this embodiment, the auxiliary model outputs operation data based on the following steps: In the third table, fault types that appear in a fixed order and whose first occurrence count is greater than the second threshold are defined as second data. Within the same round, the second data in the first and last order, as well as the area between them, are defined as areas of interest. A third threshold is set. If, within all areas of interest, there is a fault type whose second occurrence count is greater than the third threshold, then that fault type is set as a derived fault.

[0045] Continue to refer to Figure 2 If the second threshold is 5, for example, if fault type A1 and fault type A3 always appear in the third table in a certain order, with fault type A1 appearing first and then fault type A3, and the number of occurrences is 10, then fault type A1 and fault type A3 are set as the second data. Here, the sequential appearance of fault type A1 and fault type A3 is defined as one round. It is worth noting that a round begins with the first appearance of A1 and ends with the first appearance of A3. Figure 2The first round consists of the first four rows; the second data in the first sequence is the second data A1 that appears at 12:30, and the second data in the last sequence is the second data A3 that appears at 14:20. The area between them is set as the area of ​​interest; check the number of all areas of interest in the third table. If there are 10 areas of interest, and 4 of them contain fault type A2, and the third threshold is set to 3, then fault type A2 is set as a derivative fault of fault types A1 and A3.

[0046] Calculate the state transition probability based on the second formula. The second formula is: Where M is the first number, m is the second number, and a fourth threshold is set. If the state transition probability is greater than the fourth threshold, the auxiliary model outputs the second data and the operation data of the derived fault; otherwise, it only outputs the operation data corresponding to the second data.

[0047] Next, the transition probability of fault type A2 after fault type A1 is generated is calculated using the second formula. Since the number of areas of interest is 10 and the second number is 4, the calculation result is 40%, which means that after fault type A1 is generated, there is a 40% chance of generating fault type A2. If the fourth threshold is set to 20%, then the auxiliary model will not only output the state transition probability, but also output the operation data for fault type A2, thereby informing the monitoring personnel in advance and improving the efficiency of fault handling.

[0048] In step S2 of this embodiment, prediction groups are divided based on the following steps.

[0049] Based on historical data, each prediction model is tested to obtain the prediction accuracy of each prediction model for the alarm signal. A fourth threshold is set, and prediction models with prediction accuracy greater than the fourth threshold are obtained and assigned to the prediction group corresponding to the alarm signal.

[0050] Specifically, firstly, various prediction models are trained using fault data and normal data for a specific alarm signal. Then, data from the test set is input into the model for testing. For example, if 100 data points are input and the model successfully identifies the fault type corresponding to 95 of them, then the prediction accuracy of the prediction model is 95%. After that, the fourth threshold is set to 90%, and prediction models with a prediction accuracy greater than 90% are grouped into prediction groups, thus making the prediction groups include multiple prediction models.

[0051] This invention also provides an intelligent analysis device for monitoring alarm signals. This system is used to implement the aforementioned intelligent analysis method for monitoring alarm signals. The system mainly includes: The acquisition unit is used to acquire historical data, which includes normal data and abnormal data. The abnormal data includes the operating status of the power equipment associated with each fault type, the fault data, retrospective data, operation data and result data corresponding to each fault type, as well as the first moment, second moment, third moment and fourth moment generated respectively. The second moment is located before the first moment, the third moment is located after the first moment, and the fourth moment is located after the third moment. The storage unit is used to store the prediction model and the auxiliary model. The prediction model is used to output the fault type, and the auxiliary model is used to output the operation data and result data. Based on the type of alarm signal, the prediction model is divided into multiple prediction groups, and each prediction group includes multiple prediction models. The calculation unit selects the corresponding prediction group based on the type of alarm signal after generating an alarm signal. The prediction model in the prediction group outputs the fault type that generated the alarm signal based on the fault data and backtracking data corresponding to the alarm signal. The fault types output by multiple prediction models and the operating status of related power equipment are merged and input into each auxiliary model to obtain various operation data and the result data corresponding to each operation data. The display unit scores the output results of the prediction model and the auxiliary model, and fills the fault type, operation data and result data into the pre-generated first table based on the score results from largest to smallest, so as to transform the first table into an auxiliary decision table for alarm signals.

[0052] The present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program stored in the memory to perform the above-described intelligent analysis method for monitoring alarm signals.

[0053] The present invention also provides a computer storage medium storing program instructions, wherein the program instructions, when executed, control the device where the computer storage medium is located to execute the above-described intelligent judgment method for monitoring and alarm signals.

[0054] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0055] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0056] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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 specification.

[0057] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

[0058] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent analysis of monitoring alarm signals, characterized in that, include: Step S1: Obtain historical data, which includes normal data and abnormal data. The abnormal data includes the operating status of power equipment associated with each fault type, fault data, retrospective data, operation data and result data corresponding to each fault type, as well as the first moment, second moment, third moment and fourth moment generated respectively, wherein the second moment is before the first moment, the third moment is after the first moment, and the fourth moment is after the third moment. Step S2: Establish multiple prediction models and auxiliary models. The prediction model is used to output the fault type, and the auxiliary model is used to output the operation data and the result data. Based on the type of alarm signal, the prediction model is divided into multiple prediction groups, and each prediction group includes multiple prediction models. Step S3: After the alarm signal is generated, the corresponding prediction group is selected based on the type of the alarm signal. The prediction model in the prediction group outputs the fault type that generated the alarm signal based on the fault data and the backtracking data corresponding to the alarm signal. The fault types output by multiple prediction models and the operating status of associated power equipment are merged and input into each auxiliary model to obtain multiple types of operation data and the result data corresponding to each type of operation data. Step S4: Establish a first table, score the output results of the prediction model and the auxiliary model, and fill the fault type, the operation data and the result data into the first table in descending order of the score results, so as to transform the first table into an auxiliary decision table for the alarm signal; In step S2, the auxiliary model is established based on the following steps: A second table is created, and each fault type, the occurrence time of the fault type, the operating status of the power equipment associated with the fault type, and the corresponding operation data and result data are filled into the second table from the historical data. The operation data includes operation procedures arranged in chronological order for different power equipment. Set a first threshold, define the operation data in the second table that are of the same fault type and whose occurrence time interval is less than the first threshold as first data. If the first data and the operating status of each power device corresponding to the first data are not completely the same, define the fault type corresponding to the first data as the first type. Obtain the occurrence time corresponding to each first type and define the earliest occurrence time as the first time. New data is generated in the second table to transform the second table into a third table. The auxiliary model is established based on the data in the third table. The fault type in the new data is the first type, the occurrence time is the first time, the operating status of the associated power equipment and the operation data are generated by fusing the first data and the operating status of the power equipment, and are fused into two different sets of operating statuses that are mutually replaced. The result data is the result data corresponding to the first time. In step S4, the output results of the prediction model and the auxiliary model are scored based on the following steps: Set basic values ​​for each power device. Based on the power devices included in the operational data and their corresponding basic values, calculate a first score for each operational data point. Based on the importance of each fault type, set a second score for each fault type. If the prediction model and the auxiliary model generate n output results, calculate the score for each output result based on the first formula. The first formula is: ,in, and These are the first score of the operation data i and the second score of the fault type j, respectively. Output the total number of operation data i for all the auxiliary models. Output the total number of fault types J for all the predicted models.

2. The intelligent analysis method for monitoring alarm signals according to claim 1, characterized in that, The auxiliary model outputs the operation data based on the following steps: The fault type that appears in the third table in a fixed order and whose first occurrence count is greater than the second threshold is defined as the second data. The second data in the first and last order within the same round and the area between them are defined as the attention area. A third threshold is set. If the second occurrence count of the fault type is greater than the third threshold in all the attention areas, then the fault type is set as a derived fault. Calculate the state transition probability based on the second formula. The second formula is: Where M is the first number of times, m is the second number of times, and a fourth threshold is set. If the state transition probability is greater than the fourth threshold, the auxiliary model outputs the second data and the operation data of the derived fault; otherwise, it only outputs the operation data corresponding to the second data.

3. The intelligent analysis method for monitoring alarm signals according to claim 2, characterized in that, In step S2, the prediction groups are divided based on the following steps: Based on the historical data, each prediction model is tested to obtain the prediction accuracy of each prediction model for the alarm signal. A fourth threshold is set, and prediction models with prediction accuracy greater than the fourth threshold are obtained and assigned to the prediction group corresponding to the alarm signal.

4. A monitoring alarm signal intelligent analysis device, used to implement the monitoring alarm signal intelligent analysis method as described in any one of claims 1-3, characterized in that, include: The data acquisition unit is used to acquire historical data, which includes normal data and abnormal data. The abnormal data includes the operating status of power equipment associated with each fault type, fault data, retrospective data, operation data and result data corresponding to each fault type, as well as the first moment, second moment, third moment and fourth moment generated respectively, wherein the second moment is located before the first moment, the third moment is located after the first moment, and the fourth moment is located after the third moment. A storage unit is used to store a prediction model and an auxiliary model. The prediction model is used to output the fault type, and the auxiliary model is used to output the operation data and the result data. Based on the type of alarm signal, the prediction model is divided into multiple prediction groups, and each prediction group includes multiple prediction models. The calculation unit, after generating the alarm signal, selects the corresponding prediction group based on the type of the alarm signal. The prediction model in the prediction group outputs the fault type that generated the alarm signal based on the fault data and the retrospective data corresponding to the alarm signal. The calculation unit merges the fault types output by multiple prediction models and the operating status of associated power equipment into each auxiliary model to obtain multiple types of operation data and the result data corresponding to each type of operation data. The display unit scores the output results of the prediction model and the auxiliary model, and fills the fault type, the operation data and the result data into a pre-generated first table based on the score results from largest to smallest, so as to transform the first table into an auxiliary decision table for the alarm signal.

5. An electronic device comprising a processor and a memory, the memory storing a computer program, characterized in that, The processor executes the computer program stored in the memory to implement the intelligent analysis method for monitoring alarm signals as described in any one of claims 1-3.

6. A computer storage medium, characterized in that, The computer storage medium stores program instructions, wherein when the program instructions are executed, the device where the computer storage medium is located is controlled to execute the intelligent judgment method for monitoring alarm signals as described in any one of claims 1-3.

Images