A method and device for dynamic handling of alarms in a nuclear power plant

Abstract

This application belongs to the field of nuclear power plant security technology, specifically relating to a dynamic alarm handling method and device for nuclear power plants. The method includes: when an alarm is detected, acquiring real-time alarm information collected at the alarm time point; the real-time alarm information includes: an image of the alarm scene, the alarm time, and the alarm location; extracting core feature information from the alarm scene image; inputting the core feature information into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information; when the alarm trigger type matches a preset trigger type, determining the alarm priority of the real-time alarm information based on the alarm time and the alarm location; obtaining push alarm information corresponding to the real-time alarm information based on the alarm priority; and pushing the push alarm information to the user. This application can fundamentally reduce the false alarm rate, improve handling efficiency, and enhance system security.

Description

Technical Field

[0001] This application belongs to the field of nuclear power security technology, specifically relating to a method and device for dynamic handling of alarms in nuclear power plants. Background Technology

[0002] With the rapid development of security intrusion alarm detection technology, the number of related manufacturers and products continues to increase. Although the stability and reliability of the equipment have been improved to a certain extent, there are still inherent defects such as susceptibility to interference from environmental factors, and the tendency to generate false alarms and noise alarms. This not only reduces the efficiency of alarm handling personnel in handling alarm events, but also has an adverse effect on the safe operation of the system. Therefore, there is an urgent need for a method that can help users identify and eliminate noise alarm interference in intrusion alarm detection systems in order to improve the efficiency of alarm event handling. Summary of the Invention

[0003] The purpose of this application is to provide a method and apparatus for dynamic handling of alarms in nuclear power plants, which solves the problem of how to identify and eliminate noise alarm interference in intrusion alarm detection systems in the prior art.

[0004] The technical solution to achieve the purpose of this application is as follows: The first aspect of this application provides a method for dynamic handling of alarms in nuclear power plants, the method comprising: When an alarm is detected, real-time alarm information collected at the alarm time point is acquired; the real-time alarm information includes: alarm scene image, alarm time, and alarm location; Extract the core feature information from the alarm scene image; the core feature information includes the shape, size, color, and movement trajectory of the object; The core feature information is input into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information; the pattern matching model is trained based on a pre-set feature dataset of a nuclear power plant scenario and its corresponding alarm trigger types as training samples. When the alarm trigger type is consistent with the preset trigger type, the alarm priority of the real-time alarm information is determined according to the alarm time and the alarm location; Based on the alarm priority, the push alarm information corresponding to the real-time alarm information is obtained; The push alarm information will be sent to the user.

[0005] Optionally, determining the alarm priority of the real-time alarm information based on the alarm time and the alarm location specifically includes: When both the alarm time and the alarm location are within the corresponding preset filtering range, it is determined to be of normal priority. When the alarm time or the alarm location is not within the corresponding preset filtering range, it is determined to be of high priority.

[0006] Optionally, obtaining the push alarm information corresponding to the real-time alarm information based on the alarm priority specifically includes: When the alarm priority of the real-time alarm information is high priority, the pushed alarm information includes the alarm scene image and the alarm trigger type; When the alarm priority of the real-time alarm information is normal priority, the push alarm information includes the alarm trigger type.

[0007] Optionally, after pushing the alarm information to the user, the method further includes: Receive feedback data from the user regarding the pushed alarm information; the feedback data includes either confirmation or cancellation, as well as information on the processing result. The pattern matching model is updated based on the feedback data, the core feature information, and the alarm trigger type.

[0008] Optionally, updating the pattern matching model based on the feedback data, the core feature information, and the alarm trigger type specifically includes: Based on the feedback data, update the alarm trigger type corresponding to the real-time alarm information; The updated alarm trigger type and the core feature information are entered into the training sample; The pattern matching model is updated using the input training samples.

[0009] Optionally, the step of inputting the core feature information into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information further includes: When the alarm trigger type is inconsistent with the preset trigger type, it is determined to be an invalid alarm.

[0010] Optionally, the determination of an invalid alarm further includes: The alarm trigger type is pushed to the user.

[0011] A second aspect of this application provides a dynamic alarm handling device for nuclear power plants, the device comprising: The information acquisition module is used to acquire real-time alarm information collected at the alarm time point when an alarm is detected; the real-time alarm information includes: alarm scene image, alarm time and alarm location; The feature extraction module is used to extract the core feature information of the alarm scene image; the core feature information includes the shape, size, color, and movement trajectory of the object; The type acquisition module is used to input the core feature information into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information; the pattern matching model is trained based on a pre-set feature dataset of nuclear power plant scenarios and its corresponding alarm trigger types as training samples. The priority determination module is used to determine the alarm priority of the real-time alarm information based on the alarm time and the alarm location when the alarm trigger type is consistent with the preset trigger type. An alarm information acquisition module is used to obtain push alarm information corresponding to the real-time alarm information based on the alarm priority. The alarm push module is used to push the alarm information to the user.

[0012] The third aspect of this application provides a computer-readable storage medium storing computer program code thereon; when the computer program code is executed, it implements any one of the dynamic alarm handling methods for nuclear power plants provided in the first aspect of this application.

[0013] A fourth aspect of this application provides a controller, including a memory and a processor; the memory stores computer program code; when the processor reads and executes the computer program code, it implements any one of the dynamic alarm handling methods for nuclear power plants provided in the first aspect of this application.

[0014] The beneficial technical effects of this application are as follows: This application provides a method and apparatus for dynamic handling of alarms in nuclear power plants. The method includes: when an alarm is detected, acquiring real-time alarm information collected at the alarm time point; the real-time alarm information includes: an image of the alarm scene, the alarm time, and the alarm location; extracting core feature information from the alarm scene image; the core feature information includes the shape, size, color, and motion trajectory of an object; inputting the core feature information into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information; the pattern matching model is trained based on a preset feature dataset of nuclear power plant scenarios and its corresponding alarm trigger types as training samples; when the alarm trigger type is consistent with the preset trigger type, determining the alarm priority of the real-time alarm information according to the alarm time and the alarm location; obtaining push alarm information corresponding to the real-time alarm information according to the alarm priority; and pushing the push alarm information to the user. This application embodiment achieves efficient identification of the target object that triggers the alarm by performing feature extraction and pattern matching on the alarm scene images collected at the alarm time point: when an intrusion alarm signal is triggered, feature extraction and pattern matching are performed on the triggering target to clarify the type of the triggering target, and the effective alarm data and images with associated enhanced alarm information are filtered by time and location to make dual judgments on the alarm validity and push them to the user, reminding the user to take priority action. This achieves accurate identification, dynamic handling and hierarchical push of intrusion alarms, fundamentally reducing the false alarm rate and improving handling efficiency and system security. Attached Figure Description

[0015] Figure 1 A flowchart illustrating a dynamic alarm handling method for nuclear power plants provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a dynamic alarm handling device for a nuclear power plant provided in an embodiment of this application. Detailed Implementation

[0016] To enable those skilled in the art to better understand this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only a part of the embodiments of this application, and not all of them. Based on the embodiments described in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] The inventors of this application discovered in their research that traditional intrusion alarm detection systems can only directly push all generated alarm information to the user terminal, resulting in the following technical defects: 1) Alarm information lacks grading, filtering, and image verification, requiring on-duty personnel to handle invalid alarms one by one, leading to low efficiency; 2) Alarm data lacks image evidence at key time points, making it impossible to quickly confirm the accuracy and authenticity of alarms, easily overlooking real intrusion events; 3) The system lacks self-learning and model optimization capabilities, resulting in a persistently high false alarm rate; 4) Centralized data upload and processing leads to high transmission latency, failing to meet the low latency and high reliability requirements of nuclear power plant security. In summary, how to achieve accurate identification of alarm authenticity, intelligent filtering of invalid alarms, priority push of valid alarms, and dynamic iterative optimization of the system model, while simultaneously reducing processing latency and maintenance pressure, has become an urgent technical problem to be solved in the field of nuclear power plant intrusion alarms.

[0018] To address this, this application provides a method and apparatus for dynamic alarm handling in nuclear power plants. By integrating background modeling technology, multi-sensor fusion technology, and intelligent filtering algorithms from video analysis, a method for alarm identification and dynamic handling in nuclear power plant intrusion alarm detection systems is proposed. This method effectively solves the core pain point of existing systems where it is difficult to distinguish between false alarms, noise alarms, and real alarms, which seriously affects the efficiency of alarm handling. At the same time, it enables dynamic iterative optimization of the system model, reduces handling delay and maintenance pressure, and significantly improves system security.

[0019] Based on the above, in order to clearly and in detail illustrate the advantages of this application, the specific embodiments of this application will be described below in conjunction with the accompanying drawings.

[0020] See Figure 1 The figure is a flowchart illustrating a dynamic alarm handling method for nuclear power plants provided in an embodiment of this application.

[0021] This application provides a method for dynamic handling of nuclear power plant alarms, including: Step S101: When an alarm is detected, acquire the real-time alarm information collected at the alarm time point.

[0022] In this embodiment, the real-time alarm information includes: alarm scene image, alarm time, and alarm location. In practical applications, sensors can be used to detect intrusion and trigger an alarm. The alarm time and location are determined based on the alarm scene image and sensor data collected in real time by cameras and other sensing devices at the alarm trigger time.

[0023] In one example, for alarm scene images, preprocessing operations such as image denoising, contrast enhancement, normalization, and background suppression can be performed to highlight the feature information of the intrusion target and provide a basis for subsequent target comparison and classification.

[0024] Step S102: Extract the core feature information of the alarm scene image.

[0025] In this embodiment of the application, the core feature information includes the object's shape, size, color, and motion trajectory.

[0026] In practical implementation, deep learning object detection algorithms, such as YOLO (You Only Look Once) and Faster Region-based Convolutional Neural Network (Faster R-CNN), can be used to construct a feature extraction model. This model can be trained and optimized using a labeled nuclear power plant scene dataset to improve capture accuracy and efficiency. The feature extraction model extracts key features of targets from alarm scene images, accurately capturing core feature information such as the shape, size, color, and motion trajectory of target objects, providing initial target information for subsequent target comparison.

[0027] Step S103: Input the core feature information into the pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information.

[0028] In this embodiment of the application, the pattern matching model is trained based on a preset feature dataset of a nuclear power plant scenario and its corresponding alarm triggering types as training samples.

[0029] In practical implementation, a recurrent neural network (RNN) is used to construct a pattern matching model. A pre-defined feature dataset of labeled nuclear power plant scenarios is used as training samples to build a dynamic target feature library containing common targets in nuclear power plant scenarios. The pattern matching model is then trained and iteratively optimized. The extracted core feature information is input into the trained pattern matching model and accurately matched with the feature library to determine the type of target triggering the alarm and output the alarm trigger type.

[0030] In one example, historical image data of the protected area of ​​a nuclear power plant can be collected, labeled and classified to form training samples containing human bodies, animals, interfering objects, and foreign objects.

[0031] Step S104: When the alarm trigger type is consistent with the preset trigger type, determine the alarm priority of the real-time alarm information according to the alarm time and the alarm location.

[0032] It is understood that if there is a target category in the preset alarm target categories that matches the type of this alarm trigger, then the alarm priority of the real-time alarm information is determined based on the alarm time and the alarm location. In one example, the preset alarm target categories could be humans, animals, and foreign objects with a diameter greater than 30cm, etc., which will not be listed here.

[0033] In this embodiment, the alarm priority of the real-time alarm information is determined based on the alarm time and the alarm location. By supporting custom filtering of time and location, it can adapt to the differentiated security needs of different areas of a nuclear power plant.

[0034] Step S105: Obtain the push alarm information corresponding to the real-time alarm information according to the alarm priority.

[0035] Step S106: Push the alarm information to the user.

[0036] In this embodiment, image and sensor data fusion is employed, and feature extraction and pattern matching are used to significantly reduce false alarms and noise alarms caused by environmental interference. Based on the principle of hierarchical push, a differentiated push logic is used according to the difference in alarm priority to push different alarm information to users. This can identify and eliminate noise alarm interference from the intrusion alarm detection system, significantly improving the efficiency of handling. Target type matching and comparison with preset alarm categories complete the first level of target type determination; push filtering completes the second level of alarm validity screening through a dual-dimensional judgment of "time / location interval filtering". The two work together to form a hierarchical alarm push mechanism, solving the core problems of excessive false alarms, noise alarms, and low handling efficiency in traditional systems, and achieving accurate hierarchical push and efficient handling of alarm information.

[0037] In practical implementation, the nuclear power plant alarm dynamic handling method provided in this application embodiment can be deployed on an edge controller. The edge controller, as the hardware carrier for the entire method, provides the necessary computing power support for each functional module and is configured with standard physical interfaces such as network interfaces and video input / output interfaces to achieve physical link connections between data acquisition devices and user terminal devices. Localized processing at the edge ensures low latency and high reliability, meeting the stringent security requirements of nuclear power plants. When a sensor triggers an alarm, the edge controller simultaneously acquires on-site image data and sensor alarm data at the time of the alarm, constructs a multi-source fusion dataset, and performs localized processing.

[0038] In some possible implementations of the embodiments of this application, step S104 may specifically include: When both the alarm time and the alarm location are within the corresponding preset filtering range, it is determined to be of normal priority. When the alarm time or the alarm location is not within the corresponding preset filtering range, it is determined to be of high priority.

[0039] In this embodiment, based on alarm time and alarm location, a second determination of the effectiveness of real-time alarm information can be made. Based on the user-preset time interval and location interval, i.e. the corresponding preset filtering range, the pushed alarm information can be further precisely controlled.

[0040] The alarm time and location of real-time alarm information are compared with user-preset parameters, i.e., the corresponding preset filtering range. If they match, meaning the alarm falls within the preset filtering time and location range, it is determined to be of normal priority; if they do not match, it is determined to be of high priority. For example: the user presets the time filtering range to be 13:00–14:00 on a certain day; the location filtering range to be a non-critical fenced area outside a nuclear power plant. If the alarm occurs at 13:30 on the same day and the location is within that fenced area, it is determined to be of normal priority. Otherwise, it is determined to be of high priority.

[0041] In one example, step S105 may specifically include: When the alarm priority of the real-time alarm information is high priority, the pushed alarm information includes the alarm scene image and the alarm trigger type; When the alarm priority of the real-time alarm information is normal priority, the push alarm information includes the alarm trigger type.

[0042] This application embodiment can be based on the principle of hierarchical push, and adopt differentiated push logic according to the difference in alarm priority: for high priority alarms, the alarm scene image and alarm data (including alarm trigger type) are pushed to the user interface image area to remind the on-duty personnel to handle them first; for ordinary priority alarms, only the text information of alarm trigger type is pushed, and no related image is pushed to reduce interference from invalid information.

[0043] In some possible implementations of the embodiments of this application, step S106 may be followed by: Receive feedback data from the user regarding the pushed alarm information; the feedback data includes either confirmation or cancellation, as well as information on the processing result. The pattern matching model is updated based on the feedback data, the core feature information, and the alarm trigger type.

[0044] This application embodiment can be based on the principles of model iterative optimization and self-learning. By acquiring user confirmation, cancellation operations, and handling results feedback data of pushed alarm information, the real feedback data is used as the basis for model optimization to update training samples and adjust the core feature information, thereby realizing system self-learning and iterative optimization and continuously improving recognition accuracy and alarm identification precision.

[0045] In one example, updating the pattern matching model based on the feedback data, the core feature information, and the alarm trigger type may specifically include: Based on the feedback data, update the alarm trigger type corresponding to the real-time alarm information; The updated alarm trigger type and the core feature information are entered into the training sample; The pattern matching model is updated using the input training samples.

[0046] In practical implementation, the pattern matching model parameters and matching thresholds can be adjusted based on actual alarm feedback data to optimize the pattern matching model. This can also be used to build a dynamically updatable nuclear power plant target feature library and optimize and update the feature extraction model.

[0047] In some possible implementations of the embodiments of this application, step S103 may be followed by: When the alarm trigger type is inconsistent with the preset trigger type, it is determined to be an invalid alarm.

[0048] It is understandable that if there is no category in the preset trigger types that matches the current alarm trigger type, it can be determined that the alarm is an invalid alarm caused by a false alarm or noise alarm.

[0049] In one example, the determination that an alarm is invalid may be followed by: The alarm trigger type is pushed to the user.

[0050] In this embodiment of the application, for invalid alarms, the alarm trigger type can be pushed to the user normally without enhancing the alarm scene image at the time of the alarm, thus avoiding invalid interference.

[0051] This application embodiment achieves efficient identification of the target object that triggers the alarm by performing feature extraction and pattern matching on the alarm scene images collected at the alarm time point: when an intrusion alarm signal is triggered, feature extraction and pattern matching are performed on the triggering target to clarify the type of the triggering target, and the effective alarm data and images with associated enhanced alarm information are filtered by time and location to make dual judgments on the alarm validity and push them to the user, reminding the user to take priority action. This achieves accurate identification, dynamic handling and hierarchical push of intrusion alarms, fundamentally reducing the false alarm rate and improving handling efficiency and system security.

[0052] The following is a detailed explanation of a dynamic alarm handling method for nuclear power plants provided in this application, using a specific example.

[0053] First, an initial model needs to be built: a YOLO object detection network is used to construct a feature extraction model, and a recurrent neural network is used to construct a pattern matching model. Training samples are prepared: historical image data of the nuclear power plant's protected area is collected, labeled, and classified to form a training sample set containing humans, animals, interfering objects, and foreign objects. Model training: the training sample set is input into the feature extraction model, and iterative training is performed according to preset weight coefficients until the model converges. Model optimization: based on actual alarm feedback data, the model parameters and matching thresholds are adjusted to obtain the final usable feature extraction model and pattern matching model, and a dynamically updatable nuclear power plant target feature library is constructed.

[0054] This application provides a dynamic alarm handling method for nuclear power plants, which is deployed on an edge controller data acquisition device and connected to the edge controller via a physical interface; the method may specifically include the following steps: When the intrusion alarm detection system detects an alarm, it transmits the real-time alarm information collected at that alarm time point to the edge controller; The edge controller preprocesses information collected in real time by the data acquisition device, such as alarm scene images and sensor data. The target recognition technology is used to extract features and match patterns of targets in the alarm scene image. If there is no target category in the preset trigger type that matches the current alarm trigger type, it is judged as an invalid alarm caused by system false alarm or noise alarm. The alarm data is pushed to the user normally but the image at the time of the alarm is not enhanced and the process ends. If a target category matching the current alarm trigger type exists in the preset trigger types, it is determined to be a genuine alarm caused by a clear intrusion of a person or object. The alarm time and location of the real-time alarm information are used to determine the alarm priority of the real-time alarm information. If both the alarm time and alarm location are within the corresponding preset filtering range, only the alarm data is pushed to the user normally, but no image of the alarm time point is pushed, and the process ends. If the alarm time or alarm location is not within the corresponding preset filtering range, it is determined to be a high-priority valid alarm. The valid alarm data and the associated enhanced alarm information image are pushed to the user interface image area, and the process ends.

[0055] Based on the above embodiments of the method for dynamic handling of nuclear power plant alarms, this application also provides a device for dynamic handling of nuclear power plant alarms.

[0056] See Figure 2The figure is a schematic diagram of the structure of a dynamic alarm handling device for nuclear power plants provided in an embodiment of this application.

[0057] This application provides an embodiment of a dynamic alarm handling device for nuclear power plants, comprising: The information acquisition module 100 is used to acquire real-time alarm information collected at the alarm time point when an alarm is detected; the real-time alarm information includes: alarm scene image, alarm time and alarm location; The feature extraction module 200 is used to extract the core feature information of the alarm scene image; the core feature information includes the shape, size, color, and movement trajectory of the object; The type acquisition module 300 is used to input the core feature information into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information; the pattern matching model is trained based on a pre-set feature dataset of a nuclear power plant scenario and its corresponding alarm trigger types as training samples. The priority determination module 400 is used to determine the alarm priority of the real-time alarm information based on the alarm time and the alarm location when the alarm trigger type is consistent with the preset trigger type. The alarm information acquisition module 500 is used to obtain the push alarm information corresponding to the real-time alarm information according to the alarm priority. The alarm push module 600 is used to push the alarm information to the user.

[0058] In some possible implementations of this application, the priority determination module 400 can be specifically used for: When both the alarm time and the alarm location are within the corresponding preset filtering range, it is determined to be of normal priority. When the alarm time or the alarm location is not within the corresponding preset filtering range, it is determined to be of high priority.

[0059] In one example, the alarm information acquisition module 500 can be specifically used for: When the alarm priority of the real-time alarm information is high priority, the pushed alarm information includes the alarm scene image and the alarm trigger type; When the alarm priority of the real-time alarm information is normal priority, the push alarm information includes the alarm trigger type.

[0060] In some possible implementations of the embodiments of this application, the apparatus may further include: The feedback receiving module is used to receive feedback data from the user regarding the pushed alarm information; the feedback data includes either confirmation or cancellation, as well as information on the processing result. The model update module is used to update the pattern matching model based on the feedback data, the core feature information, and the alarm trigger type.

[0061] In one example, the model update module can be used specifically for: Based on the feedback data, update the alarm trigger type corresponding to the real-time alarm information; The updated alarm trigger type and the core feature information are entered into the training sample; The pattern matching model is updated using the input training samples.

[0062] In some possible implementations of the embodiments of this application, the apparatus may further include: The invalidity determination module is used to determine an invalid alarm when the alarm trigger type is inconsistent with the preset trigger type.

[0063] In one example, the invalid determination module can also be used for: Once an alarm is determined to be invalid, the alarm trigger type will be pushed to the user.

[0064] This application embodiment can achieve efficient identification of alarm-triggered target objects by performing feature extraction and pattern matching on alarm images collected at alarm time points on the edge controller: when an intrusion alarm signal is triggered, feature extraction and pattern matching are performed on the triggering target to clarify the type of the triggering target, and the effective alarm data and images with associated enhanced alarm information are double-discriminated and pushed to the user interface image area through time / location dual-dimensional filtering to remind the user to take priority action. At the same time, the target feature library and model parameters are dynamically optimized based on the processing results of the alarm event, so as to achieve accurate identification, dynamic handling and hierarchical push of intrusion alarms, fundamentally reducing the false alarm rate, improving handling efficiency and system security.

[0065] Based on the nuclear power plant alarm dynamic handling method and apparatus provided in the above embodiments, this application also provides a computer-readable storage medium storing computer program code thereon; when the computer program code is executed, it implements any one of the nuclear power plant alarm dynamic handling methods provided in the above embodiments.

[0066] Based on the nuclear power plant alarm dynamic handling method and apparatus provided in the above embodiments, this application also provides a controller, including a memory and a processor; the memory stores computer program code; when the processor reads and executes the computer program code, it implements any one of the nuclear power plant alarm dynamic handling methods provided in the above embodiments.

[0067] The present application has been described in detail above with reference to the accompanying drawings and embodiments. However, the present application is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present application. All content not described in detail in this application can be derived from existing technology.

Claims

1. A method for dynamic handling of alarms in nuclear power plants, characterized in that, The method includes: When an alarm is detected, real-time alarm information collected at the alarm time point is acquired; the real-time alarm information includes: alarm scene image, alarm time, and alarm location; Extract the core feature information from the alarm scene image; the core feature information includes the shape, size, color, and movement trajectory of the object; The core feature information is input into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information; the pattern matching model is trained based on a pre-set feature dataset of a nuclear power plant scenario and its corresponding alarm trigger types as training samples. When the alarm trigger type is consistent with the preset trigger type, the alarm priority of the real-time alarm information is determined according to the alarm time and the alarm location; Based on the alarm priority, the push alarm information corresponding to the real-time alarm information is obtained; The push alarm information will be sent to the user.

2. The method for dynamic handling of nuclear power plant alarms according to claim 1, characterized in that, The step of determining the alarm priority of the real-time alarm information based on the alarm time and the alarm location specifically includes: When both the alarm time and the alarm location are within the corresponding preset filtering range, it is determined to be of normal priority. When the alarm time or the alarm location is not within the corresponding preset filtering range, it is determined to be of high priority.

3. The method for dynamic handling of nuclear power plant alarms according to claim 2, characterized in that, The step of obtaining the push alarm information corresponding to the real-time alarm information based on the alarm priority specifically includes: When the alarm priority of the real-time alarm information is high priority, the pushed alarm information includes the alarm scene image and the alarm trigger type; When the alarm priority of the real-time alarm information is normal priority, the push alarm information includes the alarm trigger type.

4. The method for dynamic handling of nuclear power plant alarms according to claim 1, characterized in that, The process of pushing the alarm information to the user further includes: Receive feedback data from the user regarding the pushed alarm information; the feedback data includes either confirmation or cancellation, as well as information on the processing result. The pattern matching model is updated based on the feedback data, the core feature information, and the alarm trigger type.

5. The method for dynamic handling of nuclear power plant alarms according to claim 4, characterized in that, The step of updating the pattern matching model based on the feedback data, the core feature information, and the alarm trigger type specifically includes: Based on the feedback data, update the alarm trigger type corresponding to the real-time alarm information; The updated alarm trigger type and the core feature information are entered into the training sample; The pattern matching model is updated using the input training samples.

6. The method for dynamic handling of nuclear power plant alarms according to any one of claims 1-5, characterized in that, The process of inputting the core feature information into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information further includes: When the alarm trigger type is inconsistent with the preset trigger type, it is determined to be an invalid alarm.

7. The method for dynamic handling of nuclear power plant alarms according to claim 6, characterized in that, The process of determining an alarm as invalid then includes: The alarm trigger type is pushed to the user.

8. A dynamic alarm handling device for nuclear power plants, characterized in that, The device includes: The information acquisition module is used to acquire real-time alarm information collected at the alarm time point when an alarm is detected; the real-time alarm information includes: alarm scene image, alarm time and alarm location; The feature extraction module is used to extract the core feature information of the alarm scene image; the core feature information includes the shape, size, color, and movement trajectory of the object; The type acquisition module is used to input the core feature information into a pre-trained pattern matching model to obtain the alarm trigger type corresponding to the real-time alarm information; the pattern matching model is trained based on a pre-set feature dataset of nuclear power plant scenarios and its corresponding alarm trigger types as training samples. The priority determination module is used to determine the alarm priority of the real-time alarm information based on the alarm time and the alarm location when the alarm trigger type is consistent with the preset trigger type. An alarm information acquisition module is used to obtain push alarm information corresponding to the real-time alarm information based on the alarm priority. The alarm push module is used to push the alarm information to the user.

9. A computer-readable storage medium, characterized in that, It stores computer program code; when the computer program code is executed, it implements the dynamic alarm handling method for nuclear power plants as described in any one of claims 1-7.

10. A controller, characterized in that, It includes a memory and a processor; the memory stores computer program code; when the processor reads and executes the computer program code, it implements the dynamic alarm handling method for nuclear power plants as described in any one of claims 1-7.

Images