Power anti-violation recognition computing power dynamic allocation method and system based on concurrent perception

Abstract

The invention provides a power anti-violation recognition computing power dynamic allocation method and system based on concurrent perception, and relates to the technical field of computing power allocation, and the method comprises the steps: carrying out the interaction of concurrent perception task information, carrying out the execution parameter, processing amount and computing power analysis of the perception task information, and obtaining the perception task computing power description information; obtaining the concurrent processing amount of the system, and carrying out multi-dimensional analysis on the concurrent processing amount of the system; according to the concurrency constraint parameters, computing power surface layer magnitude distribution is carried out on the sensing task computing power description information; and carrying out concurrent processing optimization based on a surface level matching relationship, obtaining a final computing power distribution strategy by taking maximization of a perception task completion effect and minimization of a concurrent processing amount as targets, and carrying out computing power configuration of power violation detection. According to the method and the device, the technical problem of unreasonable power anti-violation recognition calculation power dynamic distribution in the prior art can be solved, and the technical effect of improving the power anti-violation recognition calculation power dynamic distribution reasonability is achieved.

Description

A Method and System for Dynamic Allocation of Computational Power for Power Violation Identification Based on Concurrency Awareness Technical Field

[0001] This application relates to the field of computing power allocation technology, and in particular to a method and system for dynamic allocation of computing power for power violation identification based on concurrency awareness. Background Technology

[0002] With the continuous expansion of the power system and the ongoing improvement of its intelligence and informatization levels, the requirements for operational standardization and safety in power production, transmission, distribution, and operation and maintenance are increasing. Power violation detection technology based on video perception and intelligent recognition is gradually becoming an important technical means to ensure the safe operation of the power grid. However, power violation detection often faces complex scenarios involving multiple monitoring points, multiple video streams, and multiple recognition algorithms operating concurrently. Significant differences in video quality, monitoring time span, and recognition targets across different monitoring areas make the scheduling and allocation of computing resources a key factor affecting the effectiveness of violation detection and system stability.

[0003] Currently, most existing power violation detection systems employ fixed configurations or static rules for computing power allocation. They allocate computing resources to video analysis tasks based on equipment specifications or empirical thresholds, lacking refined modeling of the complexity of the sensing tasks themselves, differences in computing power requirements, and the relationships between tasks. When processing multiple video streams concurrently, existing technologies often use only the number of parallel streams or a single load indicator as the scheduling basis, failing to comprehensively consider the impact of multiple dimensions such as video stream span, monitoring scene complexity, algorithm execution path, and deployment location on computing power consumption. This leads to uneven computing power allocation, delayed response of critical tasks, or low system resource utilization in high-concurrency scenarios. Furthermore, existing solutions typically treat each sensing task as an independent execution unit, lacking analysis of common computing power nodes and cross-computing power requirements among multiple tasks. This fails to effectively mine shareable or reusable computing power resources, resulting in redundant calculations and wasted computing power.

[0004] In summary, existing technologies suffer from several technical problems. These include a lack of a refined description mechanism for the full-cycle computing power requirements of concurrent sensing tasks, a lack of multi-dimensional concurrency constraint modeling methods that integrate video stream span, parallel number of objects, and algorithm deployment location, and a lack of dynamic computing power optimization strategies based on task relationships and historical violation patterns. These issues lead to unreasonable computing power configuration, low resource utilization efficiency, and untimely response to key violations in power grid anti-violation detection systems under high-concurrency scenarios. This further affects the accuracy, real-time performance, and overall stability and security of power grid anti-violation detection systems. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for dynamic allocation of computing power for power violation detection based on concurrent awareness. This is to solve the technical problems in the existing technology, which are caused by the lack of a refined description mechanism for the computing power requirements of the entire life cycle of concurrent awareness tasks, the lack of a multi-dimensional concurrent constraint modeling method for fusion video stream span, parallel number of objects and algorithm deployment location, and the lack of a dynamic computing power optimization strategy based on task correlation and historical violation patterns. These problems lead to unreasonable computing power configuration, low resource utilization efficiency, and untimely response to key violations in power violation detection systems under high-concurrency scenarios, which further affect the accuracy, real-time performance, and overall stability and security of power violation detection.

[0006] In view of the above problems, this application provides a method and system for dynamic allocation of computing power for power violation identification based on concurrency awareness.

[0007] Firstly, this application provides a method for dynamic allocation of computing power for power violation detection based on concurrent sensing. This method is implemented through a system for dynamic allocation of computing power for power violation detection based on concurrent sensing. The method includes: interacting with concurrent sensing task information; parsing the sensing task information for execution parameters, processing volume, and computing power to obtain a description of the sensing task's computing power; obtaining the system's concurrent processing volume; performing multi-dimensional analysis of the system's concurrent processing volume to construct concurrent constraint parameters; allocating computing power at the surface level according to the concurrent constraint parameters to the computing power description information of the sensing task, establishing a surface level matching relationship; and optimizing concurrent processing based on the surface level matching relationship, aiming to maximize the sensing task completion effect and minimize the concurrent processing volume, to obtain a final computing power allocation strategy for configuring computing power for power violation detection.

[0008] Preferably, the method for dynamic allocation of computing power for power violation identification based on concurrent perception further includes: parsing the data perception processing chain based on the perception task information to establish a full-cycle data link for the task; analyzing the node execution parameters, processing volume, and computing power requirements of the full-cycle data link for the task to establish task execution factors for each node; and parsing the task factor computing power based on the task execution factors of each node to obtain the computing power description information of the perception task.

[0009] Preferably, the method for dynamic allocation of computing power for power violation identification based on concurrent perception further includes: taking each node as the main node in the graph, taking the task execution factor as the child node, taking the computing power of the task factor as the attribute of the child node, and establishing node connection edges according to the data flow relationship of each node; and constructing a perception task computing power description graph based on the main node, child node and node connection edges to represent the description information of perception task computing power.

[0010] Preferably, the method for dynamic allocation of computing power for power violation identification based on concurrent sensing further includes: obtaining the computing power description map of each parallel sensing task, tracing the cross-relationship of the map nodes, and identifying the cross-sensing nodes; and connecting the computing power description maps of each parallel sensing task through the cross-sensing nodes to obtain the global computing power description map of parallel sensing.

[0011] Preferably, the method for dynamic allocation of computing power for power violation identification based on concurrent perception further includes: establishing a response quantification relationship between concurrent processing volume and video stream span, number of parallel objects, and algorithm deployment equipment; using the concurrent threshold of the system's concurrent processing volume, performing multi-dimensional response analysis on the response quantification relationship, determining the response parameter constraint quantity or parameter constraint combination of each dimension, and obtaining the concurrent constraint parameters.

[0012] Preferably, the power grid anti-violation identification dynamic allocation method based on concurrent perception further includes: calculating the video stream complexity and required processing resources based on the monitoring area, monitoring time period, and video quality of the video stream span; establishing a quantitative mapping relationship between the video stream span characteristics and processing resources; obtaining the response quantitative relationship corresponding to the video stream span; establishing a response quantitative relationship for the number of parallel video streams based on the changing relationship between the number of concurrent video streams and the system processing volume; and establishing a response quantitative relationship corresponding to the algorithm deployment equipment according to the changing relationship between the location of the algorithm deployment equipment and the system processing volume.

[0013] Preferably, the method for dynamically allocating computing power for power violation identification based on concurrent awareness further includes: the location of the algorithm deployment device includes: edge deployment and cloud deployment.

[0014] Preferably, the power dynamic allocation method for power violation identification based on concurrency awareness further includes: using the maximum concurrent processing capacity of the system as a concurrency threshold, and setting the concurrency threshold as a constant constraint condition; dynamically mapping the concurrency threshold to a response quantization space composed of three dimensions: video stream span, number of parallel devices, and algorithm deployment devices, based on the response quantization relationship, wherein, by threshold truncation analysis of the response curves of each dimension, the corresponding single-dimensional parameter constraint quantities are derived respectively; based on the mutual influence between the multi-dimensional parameters of video stream span, number of parallel devices, and algorithm deployment devices, constraint conditions under different combinations of dimension parameters are synthesized; and the single-dimensional parameter constraint quantities and constraint conditions under different combinations of dimension parameters are integrated to form the concurrency constraint parameters.

[0015] Preferably, the method for dynamic allocation of computing power for power violation identification based on concurrent sensing further includes: extracting key attributes to characterize computing power requirements based on the computing power description information of the sensing task, and constructing a feature vector of sensing task requirements; calculating computing power resources based on the feature vector of sensing task requirements to obtain the computing power required for the task; and, under the constraints of the concurrent constraint parameters, matching the computing power required for the task using the standardized resource units currently available in the system to obtain the surface-level matching relationship.

[0016] Preferably, the method for dynamic allocation of computing power for power violation identification based on concurrent sensing further includes: analyzing the priority and high-risk degree of concurrent sensing tasks, and configuring the allocation weights for sensing tasks; performing cross-computing power analysis on sensing tasks based on a global computing power description graph to obtain common computing power nodes and cross-computing power requirements, and extracting redundant computing power based on the common computing power nodes and cross-computing power requirements; constructing an effect evaluation function for parallel sensing tasks according to the allocation weights for sensing tasks; optimizing the allocation of redundant computing power based on the effect evaluation function under the resource boundary constraints formed by the concurrency constraint parameters and the surface-level matching relationship, with the goal of maximizing the completion effect of sensing tasks and minimizing the concurrent processing volume, optimizing the computing power allocation strategy, and searching for the computing power allocation strategy with the best target result as the final computing power allocation strategy.

[0017] Preferably, the power grid anti-violation identification computing power dynamic allocation method based on concurrent sensing further includes: extracting co-occurrence features and temporal patterns of violation tasks from a historical power grid violation event database, wherein the co-occurrence features include the probability and conditional probability of different violation types occurring simultaneously at the same monitoring point or adjacent points, and the temporal patterns are the time interval distribution, sequential order, and periodicity of different violation events; establishing a linkage relationship between corresponding sensing tasks based on the co-occurrence features and temporal patterns; determining the spatiotemporal correlation of computing power based on the linkage relationship, wherein when any sensing task in a high-frequency co-occurrence feature task combination is triggered, computing power is pre-allocated to the associated sensing task; for task combinations with strong temporal relationships, a computing power off-peak scheduling window is set according to the temporal window; and updating the global computing power description graph using the spatiotemporal correlation of computing power for real-time dynamic allocation of computing power.

[0018] Secondly, this application also provides a dynamic computing power allocation system for power violation detection based on concurrent sensing, used to execute the dynamic computing power allocation method for power violation detection based on concurrent sensing as described in the first aspect, including: an information acquisition module, used to interact with concurrent sensing task information, analyze the execution parameters, processing volume, and computing power of the sensing task information to obtain computing power description information of the sensing task; a parameter construction module, used to obtain the system concurrent processing volume, analyze the system concurrent processing volume in multiple dimensions, and construct concurrent constraint parameters; a relationship establishment module, used to allocate computing power at the surface level according to the concurrent constraint parameters to the computing power description information of the sensing task, and establish a surface level matching relationship; and a computing power configuration module, used to optimize concurrent processing based on the surface level matching relationship, with the goal of maximizing the completion effect of the sensing task and minimizing the concurrent processing volume, to obtain the final computing power allocation strategy and configure the computing power for power violation detection.

[0019] The technical solution provided in this application has at least the following technical effects or advantages: by realizing the full-cycle, structured and quantifiable description of the computing power requirements of concurrent sensing tasks, and dynamically, collaboratively and adaptively allocating computing power resources under multi-dimensional concurrency constraints, the technical effects of improving the utilization efficiency of computing power resources, enhancing the priority identification capability of high-risk violation tasks, reducing the overall concurrent processing pressure, and improving the accuracy and real-time response of power anti-violation detection are achieved while ensuring the system's concurrent carrying capacity and operational stability.

[0020] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0022] Figure 1 is a flowchart illustrating the dynamic allocation method of computing power for power violation identification based on concurrent sensing in this application.

[0023] Figure 2 is a schematic diagram of the structure of the power dynamic allocation system for identifying power violations based on concurrent perception in this application.

[0024] Figure labeling: Information acquisition module 1, parameter construction module 2, relationship establishment module 3, computing power configuration module 4. Detailed Implementation

[0025] This application provides a method and system for dynamic allocation of computing power for power grid violation detection based on concurrency awareness. It addresses existing technologies that suffer from inefficient computing power allocation, low resource utilization, and delayed response to critical violations in high-concurrency scenarios. These issues stem from a lack of refined description mechanisms for the full-cycle computing power requirements of concurrent awareness tasks, a lack of multi-dimensional concurrency constraint modeling methods that integrate video stream span, parallel processing volume, and algorithm deployment location, and a lack of dynamic computing power optimization strategies based on task relationships and historical violation patterns. These shortcomings further impact the accuracy, real-time performance, and overall stability and security of power grid violation detection systems. The application achieves a full-cycle, structured, and quantifiable description of the computing power requirements of concurrent awareness tasks, and dynamically, collaboratively, and adaptively allocates computing resources under multi-dimensional concurrency constraints. This results in improved computing resource utilization efficiency, enhanced priority identification capabilities for high-risk violations, reduced overall concurrent processing pressure, and improved accuracy and real-time response of power grid violation detection, all while ensuring system concurrency capacity and operational stability.

[0026] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0027] Example 1, please refer to Figure 1. This application provides a method for dynamic allocation of computing power for power violation identification based on concurrent sensing, which is applied to a power violation identification dynamic allocation system based on concurrent sensing. Specifically, it includes the following steps: interactive concurrent sensing task information, parsing the execution parameters, processing volume, and computing power of the sensing task information to obtain the computing power description information of the sensing task.

[0028] Furthermore, this application also includes: parsing the data perception processing chain based on the perception task information to establish a full-cycle data link for the task; analyzing the node execution parameters, processing volume, and computing power requirements of the full-cycle data link for the task to establish task execution factors for each node; and parsing the task factor computing power based on the task execution factors of each node to obtain the computing power description information of the perception task.

[0029] Furthermore, this application also includes: taking each node as the main node in the graph, taking the task execution factor as the child node, taking the task factor computing power as the attribute of the child node, and establishing node connection edges according to the data flow relationship of each node; and constructing a perception task computing power description graph based on the main node, child node and node connection edges, which is used to characterize the description information of perception task computing power.

[0030] Furthermore, this application also includes: obtaining the computing power description graph of each parallel sensing task, tracing the cross-relationships of the graph nodes, and identifying the cross-sensing nodes; and connecting the computing power description graphs of each parallel sensing task through the cross-sensing nodes to obtain the global computing power description graph of parallel sensing.

[0031] Specifically, the interactive concurrent sensing task information involves unified interaction and access to sensing task information triggered concurrently from multiple monitoring points and business scenarios. It provides a structured description of the input data types, data source devices, triggering conditions, and identification targets involved in each sensing task. Based on the sensing task information, it analyzes the data sensing processing chain to establish a complete data link throughout the task's lifecycle. In the power grid anti-violation identification process, this involves breaking down and analyzing the data flow of the sensing task throughout the entire identification process. From data acquisition, data preprocessing, feature extraction, model inference to result output, and other stages, a complete data sensing processing chain is formed, spanning the entire lifecycle of the sensing task. This constructs a data link that reflects the entire lifecycle of the sensing task, which is then used for subsequent computing power demand analysis and scheduling decisions.

[0032] Furthermore, analyzing the node execution parameters, processing volume, and computing power requirements of the entire task lifecycle data link, and establishing task execution factors for each node, involves abstracting each processing stage of the link into an independent execution node based on the already constructed task lifecycle data link. This analysis covers the execution parameters of each node, such as algorithm type, execution device, input data scale, processing complexity, number of concurrent calls, and real-time requirements. Simultaneously, considering the amount of data and computational complexity required for each execution node, the corresponding computing power consumption level is assessed. This establishes a task execution factor for each execution node that comprehensively represents its execution characteristics and resource consumption features, enabling a refined depiction of the differences in computing power requirements across different processing stages within the sensing task.

[0033] Furthermore, the execution nodes corresponding to different processing stages such as data acquisition, data preprocessing, feature analysis, model inference, and result output are abstracted as master nodes in a graph structure. The task execution factors used to characterize the computational characteristics of each execution node are mapped as child nodes attached to the corresponding master nodes. The computing power of the task factors, which reflects the computational load of each execution factor, is labeled as the attribute parameters of the child nodes. Then, based on the flow order, dependency relationship, and calling relationship of data between processing nodes during the actual execution of the perception task, node connection edges are established between master nodes and between master nodes and child nodes to describe the correlation relationship of computing power consumption changing with data flow.

[0034] Furthermore, by unifying the processing stages corresponding to the master node, the computing power impact factors corresponding to the child nodes, and the data dependencies reflected by the node connection edges, a structured graph model is formed. This enables the computing power description graph of the perception task to reflect the distribution of computing power demand, computing power transmission relationships, and computing power coupling characteristics of a single perception task under different processing stages and execution links. Thus, the computing power description information of the perception task is systematically expressed in the form of a graph structure, providing a traceable and computable basic representation for subsequent multi-task concurrent computing power analysis and optimization allocation.

[0035] Furthermore, in scenarios where multiple power violation detection tasks are executed concurrently, the computing power description graphs corresponding to each parallel detection task are retrieved separately. The main nodes, child nodes, and their attribute information in the graphs are used as the analysis objects. The node execution stages, data sources, algorithm models, device calls, and computing power consumption characteristics between different graphs are compared and traced. This identifies processing nodes that are repeatedly called or have highly similar execution characteristics in multiple detection tasks and determines them as cross-sensing nodes, which are used to characterize the shared or overlapping parts of multiple parallel detection tasks at the computing power usage level.

[0036] Furthermore, by using the identified cross-sensing nodes as connection anchors, the computing power description maps of the originally independent sensing tasks are structurally connected, so that the correlation between different sensing tasks in the data processing stage, computing power demand distribution and execution path can be uniformly expressed. This will construct a global computing power description map that can reflect the computing power resource sharing relationship, computing power coupling characteristics and potential redundant computing power distribution under the concurrent execution of multiple sensing tasks, and provide a global perspective modeling foundation for system-level computing power optimization and scheduling decisions.

[0037] Obtain the system's concurrent processing capacity, perform multi-dimensional analysis on the system's concurrent processing capacity, and construct concurrent constraint parameters.

[0038] Furthermore, this application also includes: establishing a response quantization relationship between concurrent processing volume and video stream span, number of parallel objects, and algorithm deployment device; using the concurrent threshold of the system's concurrent processing volume, performing multi-dimensional response analysis on the response quantization relationship, determining the response parameter constraint quantity or parameter constraint combination of each dimension, and obtaining the concurrent constraint parameters.

[0039] Furthermore, this application also includes: calculating the video stream complexity and required processing resources based on the monitoring area, monitoring time period, and video quality of the video stream span; establishing a quantitative mapping relationship between the video stream span characteristics and processing resources; and obtaining the response quantitative relationship corresponding to the video stream span; establishing a response quantitative relationship for the number of parallel video streams based on the changing relationship between the number of concurrent video streams and the system processing volume; and establishing a response quantitative relationship corresponding to the algorithm deployment device according to the changing relationship between the location of the algorithm deployment device and the system processing volume.

[0040] Furthermore, this application also includes: the location of the algorithm deployment device includes: edge deployment and cloud deployment.

[0041] Furthermore, this application also includes: using the system's maximum concurrent processing capacity as a concurrency threshold, and setting the concurrency threshold as a constant constraint condition; dynamically mapping the concurrency threshold to a response quantization space composed of three dimensions: video stream span, number of parallel devices, and algorithm deployment devices, based on the response quantization relationship, wherein, by threshold truncation analysis of the response curves of each dimension, corresponding single-dimensional parameter constraints are derived respectively; based on the mutual influence between the multi-dimensional parameters of video stream span, number of parallel devices, and algorithm deployment devices, constraints under different combinations of dimensional parameters are synthesized; and the single-dimensional parameter constraints and constraints under different combinations of dimensional parameters are integrated to form the concurrency constraint parameters.

[0042] Specifically, the monitoring area covered by the video stream, the duration of continuous monitoring, and video quality parameters such as video resolution, frame rate, and encoding method are used as components of the video stream span. Factors such as target density, scene change frequency, and lighting change complexity in the video content are comprehensively analyzed to assess the overall processing complexity of the video stream. Based on this, the scale of computing resources required to complete target recognition and analysis is calculated, thereby establishing a quantitative correspondence between video stream span characteristics and processing resource consumption. This shows that the larger the video stream span and the more complex the scene, the higher the corresponding processing difficulty and computing resource requirements, and forms a quantitative relationship of video stream span response that can reflect this trend.

[0043] Furthermore, under the condition of concurrent operation of the system processing multiple video streams, by analyzing the changes in the overall system processing volume, computing power utilization, processing latency, and resource scheduling pressure as the number of concurrent video streams increases, the load status and response capability of the system under different parallel stream counts are determined. Based on this, a threshold for the number of parallel processing streams that the system can stably bear is set to ensure that the system does not exceed the existing computing power and resource carrying capacity when processing multiple video streams simultaneously. Among them, the increase in the number of parallel streams will directly increase the system load and affect the response speed. The impact boundary of the number of parallel streams on the system processing performance can be clarified through the response quantification relationship.

[0044] The execution mode of the perception algorithm under different deployment locations is included in the scope of processing volume analysis. Different scenarios such as algorithm deployment on edge devices, local computing nodes, or cloud computing platforms are distinguished. By analyzing the differences in data transmission path, data preprocessing ratio, and local computing load under different deployment methods, the impact of algorithm deployment location on system processing volume and computing power consumption is quantified. Specifically, for data that has been partially or fully processed on the edge side and the processing results are directly uploaded, the local computing power requirement is relatively low. However, for tasks that need to be fully processed directly on the local or central node, the local computing power consumption will be significantly increased. Therefore, under the same concurrency conditions, the number of parallel processing targets that can be supported needs to be reduced accordingly to form a response quantification relationship that matches the location of the algorithm deployment device.

[0045] The algorithms used for video analysis and violation identification are deployed on edge computing devices or centralized cloud computing platforms close to the data collection source, depending on the distribution of computing resources and system architecture. Edge deployment refers to running the perception algorithm directly on the front-end monitoring equipment, the nearest edge server, or the local computing node, so that the video data can be fully or partially analyzed and processed locally after collection, thereby reducing the amount of data transmission, reducing transmission latency, and alleviating the pressure on the central computing power. Cloud deployment refers to running the perception algorithm on the cloud computing platform or the central server, and transmitting the collected video data or intermediate results to the cloud for centralized processing, so as to utilize the stronger computing power resources and higher elastic scalability of the cloud to complete complex identification tasks.

[0046] Furthermore, based on the maximum concurrent processing capacity that the system can stably support under given hardware configuration, network conditions, and algorithm load, a concurrent processing threshold is determined to characterize the system's limit processing capacity. This threshold is then introduced as a global constant constraint during the computing power allocation and scheduling process to ensure that the system's processing load does not exceed its maximum capacity under any concurrent operating state, thereby guaranteeing the stability and reliability of the system operation.

[0047] Furthermore, based on the established quantitative relationship between video stream span, number of parallel processing devices, and location of algorithm deployment equipment and system processing capacity, the system-level concurrent processing capacity threshold is projected into a multi-dimensional parameter space. By analyzing the response curves of system processing capacity changes with each single-dimensional parameter, and using the concurrent threshold as the cutoff point on the response curves, the maximum or optimal range of each dimension parameter under the condition of satisfying the system's concurrent capability limit is determined, thereby obtaining the corresponding single-dimensional parameter constraint results such as video stream span constraint, number of parallel processing devices constraint, and algorithm deployment equipment constraint.

[0048] Based on the mutual influence of various parameters such as video stream span, number of parallel processing devices, and algorithm deployment devices, the constraints under different combinations of parameters are synthesized. This means further considering that the parameters in each dimension are not independent of each other in actual operation, but rather have a coupling relationship. By analyzing the superimposed impact of different video stream span levels, different numbers of parallel processing devices, and different algorithm deployment locations on the system's processing volume and computing load, a comprehensive constraint model under the joint change of multi-dimensional parameters is constructed. This yields constraints applicable to different parameter combination scenarios, reflecting the true carrying capacity boundary of the system under complex concurrent operation.

[0049] Furthermore, by integrating single-dimensional parameter constraints and constraints under different combinations of parameters, concurrent constraint parameters are formed. This means unifying and integrating the parameter constraint results derived from a single dimension with the combined constraint conditions formed from the coupling analysis of multi-dimensional parameters, forming a set of concurrent constraint parameters that can simultaneously constrain the video stream span, parallel processing scale, and algorithm deployment method. This set of concurrent constraint parameters serves as a unified constraint basis in subsequent computing power allocation and scheduling, thereby achieving a comprehensive characterization and effective control of the system's concurrent processing capabilities.

[0050] Based on the concurrency constraint parameters, the computing power description information of the perception task is allocated at the surface level, and a surface level matching relationship is established.

[0051] Furthermore, this application also includes: extracting key attributes to characterize computing power requirements based on the perceived task computing power description information, and constructing a perceived task requirement feature vector; performing computing power resource calculation based on the perceived task requirement feature vector to obtain the task requirement computing power; and, under the constraints of the concurrency constraint parameters, using the standardized resource units currently available in the system to perform computing power matching on the task requirement computing power to obtain the surface-level matching relationship.

[0052] Specifically, based on the obtained computing power description map or structured computing power description information of the perception task, core attribute parameters that can directly reflect the computing load characteristics of the perception task are selected and used as the representation dimension of computing power demand. Among them, key attributes include, but are not limited to, the computational complexity of each processing node, data processing volume, real-time requirements, concurrent execution frequency, algorithm type, and execution device characteristics. By unifying the scale and structuring the key attributes, a demand feature vector is constructed to describe the overall computing power demand characteristics of a single perception task, thereby providing standardized input for computing power calculation.

[0053] Furthermore, based on the feature vector of the perception task requirements, computing power resource calculation is performed to obtain the computing power required for the task. This means that by performing weighted analysis, model calculation, or rule mapping on the features of each dimension in the requirement feature vector, the computing load required by the perception task in different processing stages is comprehensively quantified, thereby obtaining the computing power scale required to complete the perception task per unit time. The computing power required for the task can reflect the overall level of computing resource requirements of the perception task under the current operating conditions, providing a quantitative basis for subsequent computing power allocation.

[0054] Under the constraints of concurrency constraints, the computing power required for the task is matched with the computing power of the currently available standardized resource units in the system to obtain a surface-level matching relationship. This means that, under the constraints of the system's concurrent processing capacity and multi-dimensional parameter constraints, the calculated computing power required for the task is aligned with the schedulable computing resources in the system. The system's computing power resources are abstracted into several standardized resource units with fixed computing power specifications, and the combination of resource units that meet the concurrency constraints is determined by the matching algorithm so that their computing power scale corresponds to the computing power required for the task. This forms a surface-level matching relationship between the perceived task and the available computing power resources, which is used to guide the subsequent more refined computing power optimization and scheduling process.

[0055] Based on the aforementioned surface-level matching relationship, concurrent processing optimization is performed to maximize the completion effect of the perception task and minimize the amount of concurrent processing, thereby obtaining the final computing power allocation strategy for configuring computing power for power violation detection.

[0056] Furthermore, this application also includes: analyzing the task priority and high-risk level of concurrent sensing tasks, and configuring the sensing task allocation weights; performing cross-computing power analysis on sensing tasks based on the global computing power description graph, obtaining common computing power nodes and cross-computing power requirements for tasks, and extracting redundant computing power based on the common computing power nodes and cross-computing power requirements for tasks; constructing an effect evaluation function for parallel sensing tasks according to the sensing task allocation weights; optimizing the allocation of redundant computing power based on the effect evaluation function under the resource boundary constraints formed by the concurrency constraint parameters and the surface-level matching relationship, optimizing the computing power allocation strategy with the goal of maximizing the completion effect of sensing tasks and minimizing the concurrent processing volume, and searching for the computing power allocation strategy with the best target result as the final computing power allocation strategy.

[0057] Furthermore, this application also includes: extracting co-occurrence characteristics and temporal patterns of violation tasks from a historical power violation event database, wherein the co-occurrence characteristics include the probability and conditional probability of different violation types occurring simultaneously at the same monitoring point or adjacent points, and the temporal patterns are the time interval distribution, sequential order, and periodic characteristics of different violation events; establishing linkage relationships for corresponding sensing tasks based on the co-occurrence characteristics and temporal patterns; determining the spatiotemporal correlation of computing power based on the linkage relationships, wherein when any sensing task in a high-frequency co-occurrence characteristic task combination is triggered, computing power is pre-allocated to the associated sensing task; for task combinations with strong temporal relationships, a computing power off-peak scheduling window is set according to the temporal window; and updating the global computing power description graph using the spatiotemporal correlation of computing power for real-time dynamic allocation of computing power.

[0058] Specifically, in scenarios where multiple power violation detection tasks are triggered and executed concurrently, a comprehensive analysis is conducted on the violation types, risk levels, impact ranges, and real-time requirements of different detection tasks. By quantitatively assessing the importance of tasks and the degree of potential security risks, the priority order of each detection task in the computing power allocation process is determined, and an allocation weight reflecting its scheduling weight is configured for each detection task accordingly. This is used to guide resources towards high-priority, high-risk tasks under the condition of limited computing power resources.

[0059] Furthermore, based on the global computing power description map, cross-computing power analysis is performed on the sensing tasks to obtain common computing power nodes and cross-computing power requirements. Redundant computing power is extracted based on these common computing power nodes and cross-computing power requirements. This means that, based on the constructed global computing power description map of parallel sensing tasks, the overlapping parts of different sensing tasks in terms of execution nodes, algorithm models, data processing links, and computing power consumption paths are analyzed. Processing nodes that multiple sensing tasks commonly depend on or repeatedly call are identified as common computing power nodes. The resulting cross-computing power requirements are further analyzed. By merging repeated calculations, sharing intermediate results, or coordinating execution timing, redundant computing power resources that can be uniformly scheduled and reallocated are extracted to reduce overall computing power waste.

[0060] The evaluation function for parallel sensing tasks is constructed by assigning weights to sensing tasks. This means introducing the weights assigned to each sensing task into the evaluation model. By constructing an evaluation function that can comprehensively reflect the quality of task completion, response timeliness, risk identification coverage, and resource utilization efficiency, different sensing tasks can make differentiated contributions in the evaluation process according to their importance and risk weights. This provides a unified objective function and evaluation standard for optimizing the allocation of computing power.

[0061] Furthermore, under the resource boundary constraints formed by the concurrency constraint parameters and the surface-level matching relationship, redundant computing power is optimized and allocated based on the effect evaluation function. The optimization of the computing power allocation strategy aims to maximize the completion effect of the sensing tasks and minimize the concurrent processing volume. The optimal computing power allocation strategy is searched and selected as the final computing power allocation strategy. This strategy involves reallocating and dynamically adjusting the extractable redundant computing power within the resource boundary defined by the system's concurrent processing capacity, multi-dimensional parameter constraints, and the initial computing power matching relationship. While meeting the basic computing power requirements of each sensing task, the optimization objective is to improve the overall execution effect of the sensing tasks and reduce the system's concurrent processing pressure. A search or optimization algorithm is used to solve for the optimal computing power allocation scheme, and the obtained optimal computing power configuration scheme is taken as the final computing power allocation strategy executed by the system. The computing power configuration for power violation detection is then performed according to the final computing power allocation strategy, and this configuration is used in the actual computing power configuration process for power violation detection.

[0062] Furthermore, co-occurrence characteristics and temporal patterns of violation tasks are extracted from the historical power violation event database. Co-occurrence characteristics include the probability and conditional probability of different violation types occurring simultaneously at the same or adjacent monitoring points. Temporal patterns refer to the distribution of time intervals, chronological order, and periodicity of different violation events. This involves mining and analyzing historically accumulated power violation event data, statistically analyzing the probability relationships of different violation types occurring simultaneously at the same or adjacent monitoring points, and the triggering probability under specific conditions to form co-occurrence characteristics among violation tasks. Furthermore, the occurrence intervals, chronological order, and periodic change patterns of various violation events in the time dimension are analyzed to construct a temporal pattern that reflects the spatiotemporal distribution characteristics of violation behavior.

[0063] Furthermore, establishing the linkage relationship between corresponding sensing tasks based on co-occurrence characteristics and temporal patterns means mapping co-occurrence characteristics and temporal patterns to the power grid anti-violation sensing task level, determining the degree of correlation between different sensing tasks in terms of spatial location and temporal evolution, thereby establishing a linkage relationship model between sensing tasks, enabling the identification of sensing tasks that are triggered simultaneously, mutually affect each other, or are triggered continuously in actual operation, and providing a basis for subsequent collaborative scheduling of computing power.

[0064] Based on the linkage relationship, the spatiotemporal correlation of computing power is determined. Specifically, when any sensing task in a high-frequency co-occurrence feature task combination is triggered, computing power is pre-allocated to the associated sensing task. For task combinations with strong temporal relationships, computing power staggered scheduling windows are set according to the temporal window. This means further transforming the linkage relationship between sensing tasks into spatiotemporal correlation rules at the computing power scheduling level. By sensing the triggering of any task in a high-frequency co-occurrence task combination, computing power resources are reserved or pre-allocated to its associated tasks in advance to reduce subsequent response latency. At the same time, for task combinations with clear sequence and time interval patterns, computing power scheduling windows are divided according to their temporal patterns, so that related tasks are executed in staggered time periods, thereby balancing system load and improving computing power utilization efficiency.

[0065] Furthermore, the global computing power description graph is updated using the spatiotemporal correlation of computing power for real-time dynamic allocation of computing power. This means introducing the spatiotemporal correlation rules of computing power as new knowledge elements into the global computing power description graph, and dynamically updating the correlation edges between task nodes in the graph, the prediction of computing power demand, and the resource scheduling strategy. This allows the global computing power description graph to continuously evolve with the accumulation of historical experience, thereby supporting the system to make more forward-looking and adaptive dynamic allocation of computing power resources during real-time operation.

[0066] In summary, the power grid anti-violation identification dynamic allocation method based on concurrency awareness provided in this application has the following technical effects: by realizing the full-cycle, structured, and quantifiable description of the computing power requirements of concurrency awareness tasks, and dynamically, collaboratively, and adaptively allocating computing resources under multi-dimensional concurrency constraints, it achieves the technical effects of improving the utilization efficiency of computing resources, enhancing the priority identification capability of high-risk violation tasks, reducing the overall concurrency processing pressure, and improving the accuracy and real-time response of power grid anti-violation detection, while ensuring the system's concurrency carrying capacity and operational stability.

[0067] Example 2: Based on the same inventive concept as the power grid violation detection dynamic allocation method based on concurrent sensing in the previous examples, this application also provides a power grid violation detection dynamic allocation system based on concurrent sensing, as shown in Figure 2. The system includes: an information acquisition module 1, used for interacting with concurrent sensing task information, parsing execution parameters, processing volume, and computing power of the sensing task information to obtain a sensing task computing power description; a parameter construction module 2, used for obtaining the system's concurrent processing volume, performing multi-dimensional parsing of the system's concurrent processing volume, and constructing concurrent constraint parameters; a relationship establishment module 3, used for allocating computing power at the surface level according to the concurrent constraint parameters to the sensing task computing power description information, and establishing a surface level matching relationship; and a computing power configuration module 4, used for optimizing concurrent processing based on the surface level matching relationship, aiming to maximize the sensing task completion effect and minimize the concurrent processing volume, to obtain a final computing power allocation strategy and configure the computing power for power grid violation detection.

[0068] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrent perception is also used for: parsing the data perception processing chain based on the perception task information to establish a full-cycle data link for the task; analyzing the node execution parameters, processing volume, and computing power requirements of the full-cycle data link for the task to establish the task execution factor for each node; and parsing the task factor computing power based on the task execution factor of each node to obtain the computing power description information of the perception task.

[0069] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrent perception is also used to: take each node as the main node in the graph, take the task execution factor as the child node, take the task factor computing power as the attribute of the child node, and establish node connection edges according to the data flow relationship of each node; and construct a perception task computing power description graph based on the main node, child node and node connection edges to represent the description information of perception task computing power.

[0070] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrent sensing is also used to: obtain the computing power description map of each parallel sensing task, trace the cross-relationship of the map nodes, and identify the cross-sensing nodes; and connect the computing power description maps of each parallel sensing task through the cross-sensing nodes to obtain the global computing power description map of parallel sensing.

[0071] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrency awareness is also used to: establish a response quantification relationship between concurrent processing volume and video stream span, number of parallel objects, and algorithm deployment equipment; and perform multi-dimensional response analysis on the response quantification relationship using the concurrency threshold of the system's concurrent processing volume to determine the response parameter constraint quantity or parameter constraint combination of each dimension, thereby obtaining the concurrency constraint parameters.

[0072] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrent perception is also used for: calculating the complexity of the video stream and the required processing resources based on the monitoring area, monitoring time period, and video quality of the video stream span; establishing a quantitative mapping relationship between the video stream span characteristics and processing resources; obtaining the response quantitative relationship corresponding to the video stream span; establishing a response quantitative relationship for the number of parallel video streams based on the changing relationship between the number of concurrent video streams and the system processing volume; and establishing a response quantitative relationship corresponding to the algorithm deployment equipment according to the changing relationship between the location of the algorithm deployment equipment and the system processing volume.

[0073] Furthermore, the concurrent awareness-based power violation identification computing power dynamic allocation system is also used for: the location of the algorithm deployment device includes: edge deployment and cloud deployment.

[0074] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrency awareness is also used for: using the system's maximum concurrent processing capacity as a concurrency threshold, and setting the concurrency threshold as a constant constraint condition; dynamically mapping the concurrency threshold to a response quantization space composed of three dimensions: video stream span, number of parallel devices, and algorithm deployment devices, based on the response quantization relationship, wherein, by threshold truncation analysis of the response curves of each dimension, the corresponding single-dimensional parameter constraint quantities are derived respectively; based on the mutual influence between the multi-dimensional parameters of video stream span, number of parallel devices, and algorithm deployment devices, the constraint conditions under different combinations of dimension parameters are synthesized; and the single-dimensional parameter constraint quantities and the constraint conditions under different combinations of dimension parameters are integrated to form the concurrency constraint parameters.

[0075] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrent perception is also used for: extracting key attributes to characterize computing power requirements based on the computing power description information of the perception task, and constructing a perception task requirement feature vector; calculating computing power resources based on the perception task requirement feature vector to obtain the computing power required for the task; and, under the constraints of the concurrent constraint parameters, matching the computing power required for the task using the standardized resource units currently available in the system to obtain the surface-level matching relationship.

[0076] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrent sensing is also used for: analyzing the task priority and high-risk degree of concurrent sensing tasks, and configuring the allocation weight of sensing tasks; performing cross-computing power analysis on sensing tasks based on the global computing power description graph, obtaining the common computing power nodes and cross-computing power requirements of the tasks, and extracting redundant computing power based on the common computing power nodes and cross-computing power requirements of the tasks; constructing an effect evaluation function for parallel sensing tasks according to the allocation weight of the sensing tasks; optimizing the allocation of redundant computing power based on the effect evaluation function under the resource boundary constraints formed by the concurrency constraint parameters and the surface-level matching relationship, optimizing the computing power allocation strategy with the goal of maximizing the completion effect of sensing tasks and minimizing the concurrent processing volume, and searching for the computing power allocation strategy with the best target result as the final computing power allocation strategy.

[0077] Furthermore, the power grid anti-violation identification computing power dynamic allocation system based on concurrent sensing is also used for: extracting co-occurrence features and temporal patterns of violation tasks from a historical power grid violation event database, wherein the co-occurrence features include the probability and conditional probability of different violation types occurring simultaneously at the same monitoring point or adjacent points, and the temporal patterns are the time interval distribution, sequential order, and periodicity of different violation events; establishing linkage relationships between corresponding sensing tasks based on the co-occurrence features and temporal patterns; determining the spatiotemporal correlation of computing power based on the linkage relationships, wherein when any sensing task in a high-frequency co-occurrence feature task combination is triggered, computing power is pre-allocated to the associated sensing task; for task combinations with strong temporal relationships, a computing power off-peak scheduling window is set according to the temporal window; and updating the global computing power description graph using the spatiotemporal correlation of computing power for real-time dynamic allocation of computing power.

[0078] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The method and specific examples of dynamic allocation of computing power for power violation identification based on concurrency awareness in the aforementioned embodiment 1 are also applicable to the dynamic allocation system of computing power for power violation identification based on concurrency awareness in this embodiment. Through the foregoing detailed description of the method of dynamic allocation of computing power for power violation identification based on concurrency awareness, those skilled in the art can clearly understand the dynamic allocation system of computing power for power violation identification based on concurrency awareness in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0079] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0080] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for dynamic allocation of computing power for power violation identification based on concurrent sensing, characterized in that, include: Interactive concurrent sensing task information is used to analyze the execution parameters, processing volume, and computing power of the sensing task information to obtain the computing power description information of the sensing task. Obtain the system's concurrent processing capacity, perform multi-dimensional analysis on the system's concurrent processing capacity, and construct concurrency constraint parameters; allocate computing power at the surface level to the computing power description information of the perception task based on the concurrency constraint parameters, and establish a surface level matching relationship; Based on the aforementioned surface-level matching relationship, concurrent processing optimization is performed to maximize the completion effect of the perception task and minimize the amount of concurrent processing, thereby obtaining the final computing power allocation strategy for configuring computing power for power violation detection.

2. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 1, characterized in that, The process involves analyzing the execution parameters, processing volume, and computing power of the sensing task information to obtain a description of the sensing task's computing power. This includes: analyzing the data sensing processing chain based on the sensing task information to establish a full-cycle data link for the task; analyzing the node execution parameters, processing volume, and computing power requirements of the full-cycle data link for the task to establish task execution factors for each node; and analyzing the task factor computing power based on the task execution factors for each node to obtain a description of the sensing task's computing power.

3. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 2, characterized in that, Based on the task execution factors of each node, the task factor computing power is analyzed to obtain the computing power description information of the perception task. This includes: taking each node as the main node in the graph, taking the task execution factor as the child node, taking the task factor computing power as the attribute of the child node, and establishing node connection edges according to the data flow relationship of each node; and constructing a perception task computing power description graph based on the main node, child node, and node connection edges to represent the description information of the perception task computing power.

4. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 3, characterized in that, Obtaining the computing power description information of the sensing task further includes: acquiring the computing power description map of each parallel sensing task, tracing the cross-relationships of the map nodes, and identifying the cross-sensing nodes; and connecting the computing power description maps of each parallel sensing task through the cross-sensing nodes to obtain the global computing power description map of parallel sensing.

5. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 1, characterized in that, The system concurrent processing volume is obtained, and the system concurrent processing volume is analyzed in multiple dimensions to construct concurrent constraint parameters, including: establishing a quantitative relationship between concurrent processing volume and video stream span, number of parallel objects, and response of algorithm deployment devices; using the concurrent threshold of the system concurrent processing volume, the quantitative relationship of response is analyzed in multiple dimensions to determine the response parameter constraint quantity or parameter constraint combination of each dimension, and the concurrent constraint parameters are obtained.

6. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 5, characterized in that, Establish a quantitative relationship between concurrent processing volume and video stream span, number of parallel streams, and response of algorithm deployment devices. This includes: calculating video stream complexity and required processing resources based on the monitoring area, monitoring time period, and video quality of the video stream span; establishing a quantitative mapping relationship between video stream span characteristics and processing resources; obtaining the quantitative relationship of response corresponding to the video stream span; establishing a quantitative relationship of response for the number of parallel streams based on the changing relationship between the number of concurrent video streams and the system processing volume; and establishing a quantitative relationship of response corresponding to the algorithm deployment devices based on the changing relationship between the location of the algorithm deployment devices and the system processing volume.

7. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 6, characterized in that, The algorithm is deployed on devices including: edge deployment and cloud deployment.

8. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 5, characterized in that, Using the concurrency threshold of the system's concurrent processing capacity, a multi-dimensional response analysis is performed on the response quantization relationship to determine the response parameter constraints or combinations of constraints for each dimension, thereby obtaining the concurrency constraint parameters. This includes: using the system's maximum concurrent processing capacity as the concurrency threshold and setting the concurrency threshold as a constant constraint condition; dynamically mapping the concurrency threshold to a response quantization space composed of three dimensions: video stream span, number of parallel devices, and algorithm deployment devices, based on the response quantization relationship; deriving the corresponding single-dimensional parameter constraints through threshold truncation analysis of the response curves for each dimension; synthesizing constraints under different combinations of dimension parameters based on the mutual influence between the multi-dimensional parameters of video stream span, number of parallel devices, and algorithm deployment devices; and integrating the single-dimensional parameter constraints and the constraints under different combinations of dimension parameters to form the concurrency constraint parameters.

9. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 2, characterized in that, The process of allocating computing power at the surface level based on the computing power description information of the sensing task according to the concurrency constraint parameters and establishing a surface level matching relationship includes: extracting key attributes to characterize computing power requirements based on the computing power description information of the sensing task and constructing a sensing task requirement feature vector; calculating computing power resources based on the sensing task requirement feature vector to obtain the task requirement computing power; and matching the task requirement computing power using the standardized resource units currently available in the system under the constraints of the concurrency constraint parameters to obtain the surface level matching relationship.

10. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 4, characterized in that, Concurrency optimization based on the surface-level matching relationship aims to maximize the completion effect of the perception task and minimize the concurrent processing volume, resulting in a final computing power allocation strategy. This includes: analyzing the task priority and high-risk level of concurrent perception tasks and configuring task allocation weights; performing cross-computing power analysis on perception tasks based on a global computing power description graph to obtain common computing power nodes and cross-computing power requirements, and extracting redundant computing power based on these requirements; constructing an effect evaluation function for parallel perception tasks according to the task allocation weights; and optimizing the allocation of redundant computing power based on the effect evaluation function under resource boundary constraints formed by concurrency constraint parameters and the surface-level matching relationship, aiming to maximize the completion effect of the perception task and minimize the concurrent processing volume. The optimal computing power allocation strategy is then searched for and selected as the final computing power allocation strategy.

11. The method for dynamic allocation of computing power for power violation identification based on concurrent sensing according to claim 10, characterized in that, Also includes: From the historical power violation event database, co-occurrence characteristics and temporal patterns of violation tasks are extracted. Co-occurrence characteristics include the probability and conditional probability of different violation types occurring simultaneously at the same or adjacent monitoring points. The temporal patterns are the distribution of time intervals, sequence patterns, and periodic characteristics of different violation events. Based on the co-occurrence characteristics and temporal patterns, a linkage relationship is established between corresponding sensing tasks. Based on this linkage relationship, a spatiotemporal correlation of computing power is determined. When any sensing task in a high-frequency co-occurrence characteristic task combination is triggered, computing power is pre-allocated to the associated sensing task. For task combinations with strong temporal relationships, a computing power off-peak scheduling window is set according to the temporal window. The spatiotemporal correlation of computing power is used to update the global computing power description graph for real-time dynamic allocation of computing power.

12. A dynamic computing power allocation system for power violation identification based on concurrent sensing, characterized in that, The steps for implementing the power grid violation detection dynamic allocation method based on concurrent sensing as described in any one of claims 1 to 11 include: an information acquisition module, used to interact with concurrent sensing task information, analyze the execution parameters, processing volume, and computing power of the sensing task information, and obtain computing power description information of the sensing task; a parameter construction module, used to obtain the system concurrent processing volume, analyze the system concurrent processing volume in multiple dimensions, and construct concurrent constraint parameters; a relationship establishment module, used to allocate computing power at the surface level according to the concurrent constraint parameters of the sensing task computing power description information, and establish a surface level matching relationship; and a computing power configuration module, used to optimize concurrent processing based on the surface level matching relationship, with the goal of maximizing the sensing task completion effect and minimizing the concurrent processing volume, to obtain the final computing power allocation strategy and configure the computing power for power grid violation detection.

Images