Artificial intelligence data acquisition method for smart power grid

Abstract

The invention relates to the technical field of smart power grid data, and discloses an artificial intelligence data acquisition method for a smart power grid. The method comprises the following steps: starting real-time scanning of a target acquisition area when monitoring that the running state of a power grid node fluctuates; constructing a dynamic environment topology model containing electric energy quality characteristic values based on the parameter set captured by scanning; deploying an intelligent terminal equipment cluster and activating a multi-dimensional performance test protocol to obtain data capture efficiency curves under different communication protocols; mapping the equipment cluster to an associated node of the dynamic environment topology model; calculating a global data feature demand total amount in a specified time window, and deducing a theoretical acquisition load value of the equipment cluster in combination with an efficiency curve; generating an acquisition parameter regulation and control instruction sequence according to the load value; executing the instruction to drive the equipment cluster acquisition unit to switch the working state; and after the acquisition is completed, performing quantitative evaluation analysis on the acquisition efficiency index of the equipment cluster. The method can optimize the equipment operation efficiency and adapt to the dynamic change of the power grid.

Description

Technical Field

[0001] This invention relates to the field of smart grid data technology, specifically to an artificial intelligence data acquisition method for smart grids. Background Technology

[0002] As power systems rapidly evolve towards intelligence and automation, the role of smart grids in ensuring energy security and optimizing energy allocation is becoming increasingly prominent. Data acquisition, as a fundamental component of smart grid operation monitoring and control, directly impacts the stable operation and efficient management of the grid due to its accuracy, real-time nature, and comprehensiveness. Traditional power grid data acquisition methods often employ fixed-period acquisition patterns, which are ill-suited to handle sudden fluctuations in the operational status of power grid nodes. When abnormal situations such as sudden load changes or equipment failures occur in the power grid, fixed acquisition frequencies often fail to capture critical data changes in a timely manner, leading to delays in power grid dispatching decisions and even triggering safety accidents. Existing data acquisition systems lack the ability to dynamically evaluate and adapt the performance of terminal devices. Smart grids cover a wide area, and the grid topology and communication environment vary significantly across different regions. The acquisition efficiency and data transmission stability of the same batch of terminal devices can differ drastically in different scenarios. If the device acquisition parameters cannot be dynamically adjusted according to the actual environment, data redundancy or loss can easily occur, wasting system resources and failing to meet the needs of refined grid management. Traditional data acquisition methods have significant shortcomings in predicting data demand and matching equipment load. The characteristic demand for power grid operation data fluctuates dynamically with time and load changes. If the total global data characteristic demand within a specific time window cannot be accurately calculated and the acquisition load of terminal equipment cannot be reasonably allocated accordingly, some equipment may fail due to overload or become idle due to underload, seriously affecting the overall efficiency of the data acquisition system. Summary of the Invention

[0003] The purpose of this invention is to provide an artificial intelligence data acquisition method for smart grids to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides an artificial intelligence data acquisition method for smart grids, the method comprising: When a fluctuation in the operating status of a power grid node is detected, a real-time scanning operation of the target data collection area is initiated. Based on the set of power grid node parameters captured by the real-time scanning operation, a dynamic environmental topology model containing power quality characteristic values ​​is constructed. Deploy a cluster of intelligent terminal devices and activate a multi-dimensional performance testing protocol to obtain the data capture efficiency curves of the intelligent terminal device cluster under different communication protocols; The tested cluster of smart terminal devices is mapped to the associated topology node positions in the dynamic environment topology model. Calculate the total global data feature demand of the target acquisition area within a specified time window, and derive the theoretical acquisition load value of the intelligent terminal device cluster within the specified time window based on the data capture efficiency curve; Based on the theoretically acquired load value, a sequence of acquisition parameter adjustment instructions is generated for the cluster of intelligent terminal devices; Execute the sequence of acquisition parameter adjustment instructions to drive the data acquisition unit of the smart terminal device cluster to switch its working state; After the data collection is completed, the data collection efficiency indicators of the smart terminal device cluster are quantitatively evaluated and analyzed.

[0005] Preferably, the construction of the dynamic environment topology model includes: Analyze the physical connection architecture of the target acquisition area and identify key topological association information in the physical connection architecture; Extract the grid node identifiers corresponding to each key topology association information to form a power quality feature set; Quantify the data fluctuation tolerance threshold of each power quality feature set, and simultaneously detect the types of external interference factors acting on the power quality feature set; Based on the coupling relationship between the data fluctuation tolerance threshold and the external interference factor category, a data flow balance equation for the target acquisition area is constructed.

[0006] Preferably, the deployment of the intelligent terminal device cluster and activation of the multi-dimensional performance testing protocol includes: The integrity of the original data capture of the intelligent terminal device cluster was verified in high-speed data transmission mode, low-power standby mode, and protocol switching transition state, respectively. Obtain the latency characteristic parameter package of the intelligent terminal device cluster in three operating modes; The latency feature parameter package is matched and verified with a preset benchmark performance index to generate a confidence rating label for the data capture efficiency curve.

[0007] Preferably, mapping the tested cluster of smart terminal devices to the associated topology node positions in the dynamic environment topology model includes: Analyze the data fluctuation amplitude of the power quality feature set in the dynamic environment topology model and locate the feature set identifier of the dominant data fluctuation source; Spatial calibration is performed between the physical deployment coordinates of the smart terminal device cluster and the topological coordinates corresponding to the feature set identifier of the dominant data fluctuation source, and the calibration result is written into the node configuration register of the edge computing layer.

[0008] Preferably, the total global data feature requirement of the target acquisition area within the specified time window includes: By aggregating the estimated data feature generation of each power quality feature set within the specified time window, the total global data feature demand of the target acquisition area is obtained. The inherent metadata generation amount of the intelligent terminal device cluster within the specified time window is obtained, and it is superimposed with the total global data feature demand to output the quantization matrix of the theoretical collection load value. Based on the mapping relationship between the quantization matrix and the data capture efficiency curve, a data acquisition traffic allocation scheme for the smart terminal device cluster within the specified time window is generated.

[0009] Preferably, the step of generating a sequence of acquisition parameter control instructions for the smart terminal device cluster based on the theoretical acquisition load value includes: Test the maximum effective data throughput of the intelligent terminal device cluster within a single instruction cycle, and associate it with the time dimension parameter in the data collection and traffic allocation scheme; Calculate the cumulative time slice within which the acquisition unit of the intelligent terminal device cluster needs to remain active within the specified time window; The accumulated time slices are converted into timing control codes in the acquisition strategy matrix.

[0010] Preferably, the step of executing the sequence of acquisition parameter adjustment instructions to drive the data acquisition unit of the smart terminal device cluster to switch its working state includes: The specified time window is divided into multiple consecutive execution time period units, and the timing control code is allocated to each execution time period unit according to a preset priority queue; After the current execution time period ends, it is checked whether the actual amount of data collected by the smart terminal device cluster meets the time period allocation threshold. Based on the detection results, the timing control codes in the next execution period unit are dynamically weighted and adjusted until all acquisition and scheduling of the specified time window is completed.

[0011] Preferably, the method further includes an exception handling step: When a communication link interruption alarm event is detected, the timing control coding segment that has not been executed in the acquisition strategy matrix is ​​extracted; Based on the historical execution records of the intelligent terminal device cluster, reconstruct the redundant transmission path of the unexecuted timing control coding segment; The reconstructed redundant transmission path parameters are injected into the fault-tolerant scheduling module of the intelligent terminal device cluster to generate an abnormal event set and synchronize it to the central control unit.

[0012] Preferably, the step of quantitatively evaluating and analyzing the data acquisition performance indicators of the smart terminal device cluster after completing the data acquisition action includes: Obtain the data quality difference value of the intelligent terminal device cluster before and after the execution of the data acquisition task; calculate the actual feature acquisition load of the target acquisition area based on the data quality difference value; Compare the deviation rate between the actual characteristic acquisition load and the theoretical acquisition load value; The acquisition control weight coefficient of the intelligent terminal device cluster is determined based on the deviation rate; An incremental learning algorithm is used to optimize and iterate the parameter layer of the dynamic environment topology model, and the optimization results are fed back to the generation logic of the acquisition parameter control command sequence.

[0013] Preferably, the method further includes a system update step: An acquisition strategy evaluation report is established based on the acquisition control weight coefficients; the evaluation report is input into the model version control engine to trigger the topology node weight update operation of the dynamic environment topology model; The updated topology node weights are rebound to the deployment coordinates of the smart terminal device cluster, generating a new node configuration register version number; When the version number change event is detected, the new version of the register parameters is automatically loaded into the data acquisition and scheduling system of the edge computing layer.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This smart grid employs AI-powered data acquisition methods to initiate real-time scanning by monitoring fluctuations in the operational status of grid nodes. This enables it to keenly capture abnormal changes in grid operation, making data acquisition more targeted and timely, and avoiding the problem of delayed response to anomalies in traditional fixed-period acquisition modes. Based on the parameter set captured by real-time scanning, a dynamic environmental topology model containing power quality characteristic values ​​is constructed. This model comprehensively reflects the current operating environment and topology of the grid, providing accurate environmental basis for the subsequent deployment and parameter adjustment of terminal equipment, allowing equipment to better function in complex and ever-changing grid environments.

[0015] Deploying a cluster of intelligent terminal devices and activating multi-dimensional performance testing protocols allows for the acquisition of data capture efficiency curves under different communication protocols. This provides a clear understanding of the terminal devices' performance under various communication conditions, offering detailed references for device adaptation to the power grid environment and facilitating the selection of the most suitable device operating mode for the current communication environment. Mapping the tested terminal device cluster to the associated topology node positions in the dynamic environment topology model enables precise matching between the devices and the power grid topology, making device deployment more aligned with the actual operational needs of the power grid and improving the spatial accuracy of data acquisition. The system calculates the total global data feature demand of the target acquisition area within a specified time window and derives the theoretical acquisition load value of the terminal device cluster based on the data capture efficiency curve. This enables a scientific match between data demand and device acquisition capabilities, avoiding unreasonable device load allocation and making device operation more efficient and stable. Based on the theoretical acquisition load value, a sequence of acquisition parameter control instructions is generated, allowing for dynamic adjustment of the terminal device acquisition parameters. This ensures that the device acquisition status always adapts to data demand, enhancing the flexibility and adaptability of the data acquisition system. Executing a sequence of acquisition parameter control commands drives the switching of the data acquisition unit's working state within the terminal device cluster, ensuring that the devices acquire data according to optimal parameters, thus improving the quality and efficiency of data acquisition. Quantitatively evaluating and analyzing the acquisition performance indicators of the terminal device cluster after completion provides practical evidence for optimizing subsequent data acquisition strategies. By continuously summarizing experience and improving acquisition methods, the overall efficiency of the data acquisition system can be enhanced. This method forms a complete closed loop from the data acquisition initiation mechanism, environment modeling, equipment adaptation, load allocation to parameter control and performance evaluation. Each link cooperates and works together to enable the smart grid data acquisition system to better adapt to the dynamic changes in grid operation, improve the accuracy, efficiency and reliability of data acquisition, and create favorable conditions for the stable operation and refined management of the smart grid. Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating the working principle of an artificial intelligence data acquisition method for smart grids according to the present invention. Figure 2 A flowchart for constructing a dynamic environment topology model; Figure 3 A flowchart for deploying a cluster of smart terminal devices and activating a multi-dimensional performance testing protocol; Figure 4 A flowchart illustrating the process of executing a sequence of commands to adjust acquisition parameters in order to drive the switching of the working state of the data acquisition unit. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 This invention provides an artificial intelligence data acquisition method for smart grids, the method comprising: When a fluctuation in the operating status of a power grid node is detected, a real-time scan of the target acquisition area is triggered. The parameter set of the power grid nodes is captured through real-time scanning, and a dynamic environmental topology model containing power quality characteristic values ​​is constructed. A cluster of intelligent terminal devices is deployed and a multi-dimensional performance testing protocol is activated to obtain the data acquisition efficiency curves of the cluster under different communication protocols. The tested intelligent terminal device cluster is mapped to the associated topology node positions in the dynamic environmental topology model. The total global data feature demand of the target acquisition area within a specified time window is calculated, and the theoretical acquisition load value of the intelligent terminal device cluster within this time window is derived based on the data acquisition efficiency curves. Based on the theoretical acquisition load value, a sequence of acquisition parameter adjustment instructions is generated to drive the data acquisition units of the intelligent terminal device cluster to switch working states. After acquisition is completed, the acquisition performance indicators of the intelligent terminal device cluster are quantitatively evaluated and analyzed.

[0019] Example 1: See Figure 2 The process of constructing a dynamic environmental topology model begins with analyzing the physical connectivity architecture of the target data acquisition area. This analysis is based on the actual deployment locations and electrical connections of power grid nodes. By traversing the node and branch information within the power grid topology, a complete network connectivity graph is established. During the analysis, key topological relationships in the network are identified. These relationships include, but are not limited to, electrical distances between nodes, power transmission paths, and impedance matching relationships. The identification of key topological relationships relies on real-time monitoring of power grid operation data and historical data analysis. By comparing parameter changes under different operating conditions, key nodes and branches that significantly impact power grid stability are identified.

[0020] Each key topology association corresponds to a grid node identifier. These identifiers are encoded according to the grid's hierarchical structure, forming a power quality feature set. The power quality feature set includes parameters such as voltage amplitude, frequency deviation, harmonic content, and three-phase imbalance. These parameters are collected by sensors deployed on grid nodes, and the sensor data is preprocessed and stored in a distributed database. During the construction of the power quality feature set, the timeliness and completeness of the data must be considered. A sliding time window mechanism is used to segment and aggregate the data, ensuring that the feature set reflects the dynamic characteristics of the grid operation.

[0021] When quantifying the data fluctuation tolerance threshold of power quality feature sets, it is necessary to combine the safety specifications of power grid operation with historical operating data. The data fluctuation tolerance threshold is not a fixed value, but is dynamically adjusted based on factors such as grid load levels, ambient temperature, and equipment aging. For example, under heavy load operation, the voltage amplitude fluctuation tolerance threshold may be tightened, while it may be relatively lenient under light load operation. The quantification process uses statistical analysis methods to calculate the distribution characteristics of power quality parameters under different operating scenarios, thereby determining a reasonable fluctuation range.

[0022] Detecting the types of external interference factors acting on the power quality feature set is a crucial step in constructing a dynamic environmental topology model. External interference factors include electromagnetic interference (EMI), load surges, communication delays, and ambient temperature variations. EMI primarily originates from the start-up and shutdown of nearby electrical equipment or lightning activity, manifesting as voltage waveform distortion or increased signal noise. Load surges are typically caused by the switching of high-power equipment, potentially leading to voltage sags or frequency fluctuations. Communication delays are related to the congestion level of the data transmission path and may affect the real-time performance of control commands. Ambient temperature variations affect the heat dissipation performance of electrical equipment, leading to parameter drift. Detecting these interference factors requires combining multi-source data, including power grid monitoring data, meteorological data, and communication network status data, using data fusion techniques to identify the type and intensity of the interference factors.

[0023] Based on the coupling relationship between the data fluctuation tolerance threshold and the categories of external interference factors, a data flow balance equation for the target acquisition area is constructed. This equation describes the stability boundary conditions of the power quality feature set in a dynamic environment. During construction, the influence mechanisms of external interference factors on power quality parameters are first analyzed; for example, electromagnetic interference may lead to an increase in voltage harmonic content, while sudden load changes may cause frequency deviations. Then, these influence mechanisms are correlated with the data fluctuation tolerance threshold to establish constraints between parameters. The specific form of the data flow balance equation depends on the actual needs of the power grid operation and may contain linear or nonlinear relationships. For example, in voltage stability analysis, the equation may reflect the dynamic balance between voltage amplitude and load power; in frequency regulation scenarios, the equation may describe the response relationship between generator output and load changes.

[0024] Updating and maintaining a dynamic environment topology model is an ongoing process. As the power grid's operating state changes, the data fluctuation tolerance threshold and external disturbance factor categories in the power quality feature set may also change. Therefore, it is necessary to periodically recalculate the parameters of the data flow balance equations to ensure that the model accurately reflects the actual operating conditions of the power grid. During the update process, an incremental learning algorithm is used to optimize the model parameters, recalculating only the changed parts to improve computational efficiency. Simultaneously, a version control mechanism ensures that the updated topology model can be correctly loaded and applied by the system after each update.

[0025] The application of dynamic environmental topology models is not limited to data acquisition task scheduling; it can also support other advanced functions of the power grid. For example, in fault location and isolation scenarios, the model can be used to quickly determine the area affected by a fault; in power quality management scenarios, the model can provide a basis for the switching strategies of reactive power compensation equipment. The model's versatility allows it to adapt to power grid application scenarios of different scales, from distribution networks to transmission networks, and can be adjusted according to actual needs.

[0026] In the actual deployment of the model, the allocation and optimization of computing resources need to be considered. The computation of dynamic environment topology models may involve the real-time processing of large amounts of data; therefore, it is necessary to make reasonable use of edge computing and cloud computing resources. Edge computing nodes are responsible for the rapid processing of local data and the preliminary calculation of model parameters, while the cloud computing platform undertakes the task of integrating and optimizing the global model. This layered computing architecture can effectively balance the contradiction between real-time performance and computational complexity, ensuring the efficient operation of the model in various application scenarios.

[0027] The construction and maintenance of dynamic environmental topology models rely heavily on the support of intelligent algorithms. Machine learning algorithms can be used to identify abnormal patterns in power quality feature sets, deep learning algorithms can be used to predict the development trends of external interference factors, and optimization algorithms can be used to solve parameter combinations in data flow balance equations. The selection and configuration of these algorithms need to be adjusted according to the specific needs of the power grid, and the algorithm parameters also need to be dynamically updated as the power grid's operating status changes.

[0028] Example 2: See Figure 3 The process of deploying a cluster of intelligent terminal devices and activating a multi-dimensional performance testing protocol begins with the initial configuration of the devices. The cluster consists of data acquisition units distributed across key nodes of the power grid, each unit containing a sensor module, a communication module, and a data processing module. During deployment, each device in the cluster is first physically installed and connected to the network to ensure a stable communication connection with the central control system. After installation, device parameters are uniformly configured using a remote configuration tool, including basic parameters such as sampling frequency, data format, and communication protocol. Parameter configuration uses a template-based approach, selecting the appropriate configuration template based on the device's location and functional requirements, ensuring both consistency in device functionality and consideration of the specific needs of different nodes.

[0029] The multi-dimensional performance testing protocol is activated after the device completes its basic configuration. The testing protocol includes verification of three main operating modes: high-speed data transmission mode, low-power standby mode, and protocol switching transition state. High-speed data transmission mode simulates the device's operation under abnormal power grid conditions, requiring the device to transmit data at the highest sampling frequency and maximum communication bandwidth. During the test, the device generates simulated data streams according to preset test vectors, while simultaneously recording key indicators during data transmission. Low-power standby mode verifies the device's energy-saving performance under stable power grid operation. The device enters a sleep state, maintaining only necessary monitoring functions; the test focuses on wake-up response time and data integrity under low-power conditions. Protocol switching transition state tests the device's performance when switching between different communication protocols, including parameters such as switching time and data loss rate.

[0030] The verification of the integrity of the raw data capture employs a multi-factor verification mechanism. In high-speed data transmission mode, the test data packets generated by the device are appended with a sequence number and a checksum. The receiving end assesses data integrity by comparing the continuity of the sequence number and the correctness of the checksum. Data verification in low-power standby mode focuses on the accuracy and timeliness of the first data packet after the device wakes up. Verification during protocol switching transitions focuses on the continuity of data packets before and after the switch, confirming data integrity through timestamp comparison and content continuity checks. All verification results are recorded in the device's local storage for subsequent analysis.

[0031] The acquisition and testing of latency characteristic parameters are performed simultaneously. In high-speed data transmission mode, the entire process time from data sampling to reception confirmation is recorded, broken down into sub-indicators such as sampling delay, processing delay, and transmission delay. In low-power standby mode, the main measurement is the time delay from event triggering to full device activation. Latency parameters for protocol switching transition states include protocol negotiation time, parameter reconfiguration time, and link reconstruction time. These latency parameters are acquired using a high-precision timer built into the device, and timestamp synchronization is calibrated using a network time protocol to ensure the comparability of data acquired by different devices.

[0032] The matching verification of preset benchmark performance indicators adopts a tiered evaluation strategy. The collected latency parameters are initially compared with the nominal values ​​in the equipment's technical specifications to confirm whether the equipment meets the basic performance requirements. Then, more stringent performance level standards are set according to the actual operation requirements of the power grid, and the equipment is graded and evaluated. The matching process considers the characteristics of different operating modes: high-speed data transmission mode focuses on throughput and real-time performance indicators, low-power standby mode focuses on energy consumption and response speed, and protocol switching transition states emphasize stability and reliability. The matching results generate a confidence rating label for the data capture efficiency curve, which comprehensively reflects the equipment's performance under different operating conditions.

[0033] The mapping deployment of the intelligent terminal device cluster is based on the analysis results of a dynamic environmental topology model. First, the data fluctuation characteristics of each power quality feature set in the model are analyzed, and time series analysis algorithms are used to identify node regions where fluctuation amplitudes exceed thresholds. Spectral analysis and correlation calculations are performed on the feature sets of these regions to determine the location and impact range of the dominant data fluctuation source. Then, the physical location information of the intelligent terminal devices is matched with the topological coordinates of the fluctuation source feature sets, taking into account factors such as device communication radius, installation conditions, and signal coverage. For critical fluctuation areas, a redundant device deployment strategy is adopted to ensure the reliability of data acquisition.

[0034] The spatial calibration process employs an iterative optimization method. Initial calibration establishes a mapping model based on the device deployment coordinates and the theoretical positional relationship of topology nodes. Through feedback from actual operational data, the correspondence between devices and topology nodes is continuously adjusted to resolve mapping deviations caused by installation errors, signal attenuation, and other factors. The calibration algorithm comprehensively considers the spatiotemporal characteristics of the data collected by the devices. For data streams with strong time correlation, it emphasizes time synchronization calibration; for data with obvious spatial distribution characteristics, it emphasizes position matching calibration. The calibration results are written to the node configuration register of the edge computing layer in the form of configuration parameters. The register content includes information such as device identifier, topology node mapping relationship, and data acquisition parameters.

[0035] The edge computing layer employs a distributed management architecture for node configuration. Each edge computing node manages the smart terminal devices within its jurisdiction and maintains its local configuration register. The central control system ensures the consistency of configuration information across all edge nodes through a periodic synchronization mechanism. When the topology or device deployment changes, configuration update commands are first sent to the relevant edge nodes, which then distribute the updated content to their respective terminal devices. This hierarchical management mechanism guarantees both the timeliness of configuration updates and avoids communication congestion at the central node.

[0036] Monitoring the operational status of the intelligent terminal device cluster is conducted throughout the entire implementation process. Devices periodically report operational status data, including basic parameters such as operating temperature, power supply voltage, and signal strength, as well as performance indicators such as data acquisition quality and communication success rate. This status monitoring data is used to assess device health, predict potential faults, and provide a basis for subsequent performance optimization. For devices in abnormal states, the system automatically triggers a diagnostic process, taking countermeasures such as parameter adjustments, switching to backup equipment, or on-site maintenance based on the diagnostic results.

[0037] Remote firmware updates ensure continuous system optimization. When device performance needs improvement or software bug fixes are required, update packages are distributed to target devices via a secure channel. The update process employs a phased, rolling upgrade strategy, first verifying the update's effectiveness on a small number of devices, and then gradually expanding the upgrade scope after confirmation. Each firmware update retains a rollback mechanism to ensure a quick revert to a stable version in case of compatibility issues.

[0038] The optimization and adjustment of the communication network are carried out simultaneously with equipment deployment. Based on the distribution density of smart terminal devices and data traffic characteristics, the topology and parameter configuration of the communication network are dynamically adjusted. For areas with dense data transmission, communication channel bandwidth is increased or relay equipment is deployed; for coverage edge areas, signal modulation methods are optimized or antenna orientation is adjusted. Network optimization aims to ensure real-time and reliable data transmission, balancing communication quality and energy consumption.

[0039] Test and operational data from the intelligent terminal device cluster are entered into the system's knowledge base. This data is used to improve device performance models, optimize test plans, and refine deployment strategies. The knowledge base uses a time-series database to store historical data, facilitating trend analysis and pattern mining. Based on the data analysis results from the knowledge base, performance testing protocols and device configuration templates are revised regularly, forming a closed-loop management mechanism for continuous improvement.

[0040] Equipment deployment and performance testing mutually validate each other; test results guide deployment optimization, and deployment effectiveness is fed back to adjust the test plan. The dynamic environment topology model provides the theoretical basis for equipment mapping, while actual operational data, in turn, corrects and refines the model parameters. This systematic implementation approach ensures that the intelligent terminal equipment cluster can adapt to the dynamic needs of power grid operation and maintain reliable data acquisition capabilities under various operating conditions.

[0041] Example 3: When calculating the total global data feature demand of the target acquisition area within a specified time window, it is first necessary to establish a mapping relationship between the time window and the data features. The division of the time window is based on the frequency of changes in the power grid operating state, usually using a sliding window mechanism, with the window length dynamically adjusted according to the time-varying characteristics of the data features. For rapidly changing power quality parameters, such as voltage sags or harmonic distortion, a shorter time window is used; for slowly changing parameters, such as equipment temperature or load trends, the window length is appropriately extended. The amount of data features generated within each time window is estimated through a feature extraction algorithm. The algorithm input is the original sampled data, and the output is a normalized feature vector. The feature vector contains parameters in multiple dimensions, each corresponding to a change characteristic of a power quality index.

[0042] The aggregation process for the total global data feature requirements adopts a hierarchical accumulation strategy. First, the feature generation volume of a single power grid node within each time segment is calculated. Then, node data within the same topological region are spatially aggregated. Finally, the results from all regions are integrated along the time dimension. During the aggregation process, the weight allocation of different data features needs to be considered; key features have higher weight coefficients, while secondary features have relatively lower weight coefficients. The weight coefficients are set with reference to power grid operation procedures and expert experience, and are dynamically adjusted based on actual operating conditions. To quantify this process, the following calculation formula is introduced: , in, This represents the total demand for global data features. The total number of power grid nodes. The number of segments into which the time window is divided. Representing the The node at the th Feature weight coefficients for each time segment This represents the estimated amount of data features generated for the corresponding node and time segment. The weighting coefficients in this formula... It is a dynamic variable, and its value depends on the importance of the node and the criticality of the time segment.

[0043] The inherent metadata generation of intelligent terminal device clusters includes two types of data: device status information and system operation logs. Device status information records the operating parameters of the acquisition units, such as sampling accuracy, communication quality, and power supply voltage; system operation logs contain management information such as task scheduling records and abnormal event reports. Although this metadata does not directly reflect the power grid's operating status, it is crucial for assessing data quality and system reliability. The estimation of metadata generation employs statistical analysis methods, establishing a predictive model based on parameters such as device model, deployment environment, and runtime. The model input consists of device configuration parameters and environmental monitoring data, and the output is a metadata generation rate curve.

[0044] The quantization matrix construction process for theoretically collected load values ​​employs a spatiotemporal two-dimensional modeling method. The rows of the matrix correspond to the physical distribution of intelligent terminal devices, while the columns represent the segmented sequence of time windows. Each matrix element's value consists of two parts: the predicted data feature demand for the corresponding node and time period, and the basic generation amount of device metadata. The matrix filling algorithm considers differences in device performance, allocating larger load values ​​to high-performance nodes and appropriately reducing load requirements for performance-constrained nodes. Matrix sparsity control is achieved by setting a minimum load threshold; elements below the threshold are set to zero to reduce unnecessary data collection tasks.

[0045] The traffic allocation scheme is generated based on the convolution operation of the quantization matrix and the data capture efficiency curve. The data capture efficiency curve reflects the performance of the device under different operating conditions, including key indicators such as sampling success rate and transmission rate. The convolution kernel design takes into account the time decay effect, assigning higher priority to recent data and gradually decreasing the weight of historical data. After normalization, the calculation results generate the acquisition task intensity coefficient for each device in each time period. The coefficient matrix is ​​decomposed into specific acquisition instructions through a distributed scheduling algorithm and sent to each intelligent terminal device for execution.

[0046] Single-instruction cycle testing of the intelligent terminal device cluster was conducted in a laboratory environment. The test platform simulated actual power grid operation scenarios, generating test cases covering various typical operating conditions. During the test, the devices recorded complete work logs, including detailed parameters such as instruction response time, data processing time, and cache usage. Test data analysis employed statistical methods such as box plots to identify the distribution characteristics and outliers of performance indicators. The determination of maximum effective data throughput considered worst-case performance, i.e., stable processing capability under multiple concurrent interferences. The test results formed a device performance benchmark library, serving as a reference for subsequent load calculations.

[0047] The cumulative time slice calculation for the active state of the acquisition unit employs a discrete-time integration method. The integration interval covers the entire specified time window, and the integrand is the task load coefficient of the device at each time point. The integration step size is determined based on the frequency of power grid state changes; a smaller step size is used during rapid fluctuations, while the step size is appropriately increased during stable operation. The allocation of cumulative time slices follows the load balancing principle to avoid situations where some devices are under high load for extended periods while others are idle. The time slice adjustment algorithm is adaptive, capable of dynamically reallocating the workload based on the actual performance of the devices.

[0048] The timing control encoding conversion process employs a state machine model. The model input is the cumulative time slice distribution sequence, and the output is binary control instructions that can be directly executed by the device. The encoding rules consider the characteristics of the device hardware, such as the processor's instruction set architecture and memory access timing. Instruction optimization is performed during the conversion process, merging similar operations in adjacent time slices to reduce unnecessary state transitions. The generated acquisition strategy matrix contains complete timing control information; the rows of the matrix represent control time slots, and the columns correspond to different device functional units. Sparsity analysis of the matrix is ​​used to identify optimizable instruction sequences, improving control efficiency.

[0049] The mapping relationship between the quantization matrix and the data capture efficiency curve is maintained using a version control mechanism. Each time the power grid topology or equipment configuration changes, the system automatically generates a new version of the mapping relationship. Version difference analysis helps identify affected acquisition tasks and triggers local recalculations. The mapping relationship is stored using graph database technology, which facilitates the representation of complex relationships. The version rollback function ensures rapid recovery to a previous stable state in the event of calculation errors.

[0050] The dynamic adjustment of the data acquisition and allocation scheme is achieved through a feedback control loop. The input to the control loop is the actual data acquisition performance of the equipment, and the output is the correction parameters for the allocation scheme. The control algorithm employs a gradual adjustment strategy to avoid system oscillations caused by a single large adjustment. Historical data during the adjustment process is stored in a knowledge base for optimizing the parameter settings of the control algorithm. The knowledge base uses a time-series data model, supporting the analysis and prediction of complex patterns.

[0051] The real-time update mechanism for theoretically acquired load values ​​is linked to the power grid condition monitoring system. When a significant condition change is detected, such as the switching of large-capacity loads or network reconfiguration, the system immediately initiates a recalculation of the load values. The update process uses an incremental calculation method, reassessing only the affected portion to improve response speed. Comparative analysis of the old and new load values ​​is used to assess the impact range of the condition change, providing a reference for subsequent dispatching decisions. The load forecasting model is periodically retrained using the latest operating data to maintain forecast accuracy.

[0052] Example 4: See Figure 4This will be illustrated through a typical data acquisition case in a power distribution network. A 110kV substation has 12 intelligent terminal devices deployed in its distribution network area, responsible for monitoring the operational status of 8 key distribution nodes. After detecting a voltage fluctuation event in the area, the system initiates a data acquisition task with a time window set at 30 minutes, divided into 6 consecutive execution time units, each unit lasting 5 minutes.

[0053] The time-slot allocation scheme adopts priority queue management, and the collection priority is determined according to the voltage fluctuation amplitude of the nodes. As shown in Table 1, the time-slot allocation of three typical power distribution nodes is illustrated.

[0054] Table 1: Time Period Allocation Table for 3 Typical Distribution Nodes Node number Voltage fluctuation Period 1 Period 2 Time period 3 Period 4 Period 5 Period 6 NODE_07 8.2% high high middle middle Low Low NODE_12 5.7% middle high high middle middle Low NODE_15 3.1% Low middle middle high middle middle Priority settings are based on a dynamic weighting algorithm, considering three key factors: the importance of a node's location in the power grid topology, the magnitude of real-time voltage fluctuations, and the frequency of historical data anomalies. High-priority nodes receive more data acquisition time slices in the time slot allocation, with their sampling frequency increased to three times the normal value; medium-priority nodes maintain the standard sampling frequency; and low-priority nodes adopt a reduced-frequency sampling mode. At the end of each time slot, the system automatically compares the actual amount of collected data with the expected value, triggering a weighting correction mechanism when the difference rate exceeds 15%.

[0055] The handling process for communication link interruptions is demonstrated through a real-world example using node NODE_07. During the third execution period, an interruption of the 4G communication link between this node and the edge computing node was detected. The system immediately activated the redundant transmission scheme: first, it retrieved the node's historical communication records and found that it had successfully transmitted data via the LoRa wireless link; then, it activated the backup communication module, re-encapsulating the unexecuted timing control codes into LoRa protocol data packets; finally, it established a multi-hop transmission path through the relay function of the neighboring node NODE_12. The fault-tolerant scheduling module recorded detailed information about the interruption event, including the occurrence time, duration, and backup scheme used, generating an abnormal event report that was synchronized to the central control system.

[0056] The dynamic weight correction process is reflected in the scheduling adjustment of node NODE_12. In time period 2, the actual data collected by this node was only 82% of the expected value. Analysis revealed that this was due to increased electromagnetic interference caused by the switching of high-power equipment from adjacent nodes. Based on the node's historical performance data, the system automatically reduced the expected sampling frequency for time period 4 and extended the sampling interval. The correction algorithm references the historical performance of the equipment under similar interference environments, adopting a conservative adjustment strategy to avoid over-adjustment due to a single anomaly.

[0057] The division of execution time periods takes into account the characteristics of power grid operation. The first two time periods focus on capturing the initial stage characteristics of voltage fluctuations, the middle time period monitors the fluctuation propagation process, and the last time period records the recovery status. Each time period's task allocation retains a 10% time margin to handle unexpected situations. When switching time periods, the system performs an integrity check to confirm that all critical data has been cached and preprocessed, and then smoothly transitions to the acquisition strategy of the next time period.

[0058] The construction of the abnormal event set includes multi-dimensional diagnostic information. Taking the communication delay event of node NODE_15 as an example, the set records the following elements: the power grid load level at the time of the event, the status of adjacent nodes, ambient temperature and humidity, signal strength indicators, and other related data. This information is stored through classification and coding to facilitate subsequent root cause analysis. After receiving the abnormal event set, the central control system initiates a correlation analysis process to identify common problems and potential risks across nodes.

[0059] The timing control coding of the acquisition strategy matrix adopts a hierarchical structure. The top-level coding defines the acquisition mode, including three basic types: high-speed sampling, standard sampling, and energy-saving sampling. The middle-level coding specifies specific parameters, such as sampling interval, data compression algorithm, and verification method. The bottom-level coding contains the device's operation instructions, such as maintenance commands like sensor calibration and cache clearing. This hierarchical structure ensures that control instructions remain standardized while adapting to the characteristics and requirements of different devices.

[0060] The selection criteria for redundant transmission paths are illustrated in the case of node NODE_07. Path evaluation considers four factors: historical transmission success rate, current signal quality, path delay characteristics, and energy consumption level. The system prioritizes paths with a good historical transmission record, followed by real-time signal strength. For time-sensitive data, direct paths are chosen at the expense of some reliability; for non-real-time data, multi-hop relays can be used to improve transmission success rate. Path switching decisions are completed within 200 milliseconds to ensure continuous data acquisition.

[0061] The node configuration register update process in the edge computing layer demonstrates dynamic adjustment capabilities. When node NODE_12 switches its communication mode from 4G to LoRa, the register automatically updates the following parameters: communication protocol version, radio frequency parameters, encryption key, etc. The update employs a transaction processing mechanism to ensure the atomicity of parameter changes. The register retains three versions of historical configurations, supporting fast rollback operations. After each update, the system automatically verifies the validity of the new configuration, and only submits the changes after confirming that the device can communicate normally.

[0062] The fault-tolerant scheduling module's workflow embodies the concept of fault isolation. When a smart terminal device malfunctions, the module first limits the scope of the fault's impact by transferring relevant data collection tasks to nearby devices. Then, it diagnoses the fault type, distinguishing between permanent damage and temporary anomalies. Finally, based on the diagnostic results, it takes appropriate measures: setting up a retry mechanism for temporary faults and marking the device as needing maintenance for permanent faults. The entire process does not affect the operation of other normally functioning devices.

[0063] The central control unit's synchronization mechanism ensures system state consistency. Abnormal event sets are synchronized incrementally, transmitting only changed data items. The synchronization protocol includes sequence number checksums and content digests to ensure data transmission integrity and correct sequence. For critical configuration changes, a two-phase commit protocol is used: pre-execution is performed on edge nodes, and the changes officially take effect only after central confirmation. This mechanism effectively solves the configuration inconsistency problem caused by network latency.

[0064] The state switching process of a cluster of intelligent terminal devices emphasizes a smooth transition. When a device switches from high-speed sampling mode to low-power mode, it performs the following steps: completes the current sampling cycle, clears the transmission buffer, saves the operating state, shuts down unnecessary circuits, and starts a sleep timer. During reverse switching, the device first restores basic functions, performs a rapid self-test, and then gradually increases the sampling frequency. The state switching algorithm considers the thermal characteristics of the devices to avoid overheating of components due to frequent switching in a short period.

[0065] The dynamic adjustment of time-period allocation thresholds reflects the system's adaptive capability. The initial threshold value is set based on equipment specifications and continuously optimized during operation based on actual performance. For equipment that consistently reaches the threshold, the system appropriately increases the expected value; for equipment that frequently fails to reach the threshold, the reasons are analyzed and the threshold is adjusted. The threshold management algorithm avoids a static, "one-size-fits-all" setting, instead differentiating between different equipment models, deployment locations, and years of use, implementing differentiated strategies.

[0066] Resource balancing in data acquisition task scheduling is illustrated using the example of node NODE_15. When the system detects a fault on a line adjacent to this node, it automatically reallocates acquisition resources: reducing the sampling density of nodes far from the faulty line and concentrating resources on monitoring key parameters near the fault point. This dynamic adjustment ensures that limited resources are prioritized for the most critical data acquisition while avoiding equipment overload. The resource balancing algorithm's cycle is synchronized with the power grid condition assessment, typically set to a cycle of 10-15 seconds.

[0067] Example 5: Evaluation of Data Acquisition Performance and System Updates for Smart Terminal Device Clusters. After completing the data acquisition task within a specified time window, the system initiates a multi-level quantitative evaluation process. The calculation of data quality difference values ​​begins with a comparison between the original sampled data and the reconstructed data, using waveform similarity algorithms to analyze the characteristic differences between the two in the time and frequency domains. The reconstructed data is obtained through cross-validation of redundant deployment nodes to ensure the reliability of the reference data. The difference value calculation covers three key dimensions: numerical accuracy, temporal consistency, and feature integrity, with independent evaluation criteria set for each dimension. Numerical accuracy focuses on the absolute error range of the sampled values; temporal consistency checks the continuity of the timestamps of data packets; and feature integrity assesses the degree of preservation of key waveform features.

[0068] The derivation of the actual characteristic sampling load combines equipment operation logs and network transmission records. The system analyzes the detailed logs generated by the equipment during the sampling process, extracting key indicators such as the number of valid sampling points, retransmission counts, and abnormal interruption events. Network transmission records provide the data packet delivery rate and transmission delay distribution. After normalization, these data are input into the load calculation model, outputting a comprehensive score of the actual sampling load. The scoring system uses a percentage system, with data integrity accounting for 40%, timeliness for 30%, and energy efficiency for 30%. The load calculation results are compared item by item with the theoretical expected values, identifying indicators where the deviation exceeds the allowable range.

[0069] The determination of the control weight coefficients for data acquisition employs fuzzy logic inference. The system establishes an inference rule base containing multiple input variables, including load deviation rate, equipment health status, and environmental interference level. Each variable is divided into several fuzzy sets; for example, the load deviation rate is divided into "slight deviation," "moderate deviation," and "severe deviation." The inference engine activates the corresponding fuzzy rules based on the current operating status and outputs specific weight coefficients through defuzzification. The weight coefficients are adjusted between 0 and 1 in increments of 0.1; a higher coefficient indicates that the equipment plays a more important data acquisition role in subsequent tasks.

[0070] The optimization iteration of the dynamic environment topology model adopts an incremental learning strategy. The model parameter layer is divided into two categories: core parameters and auxiliary parameters. Core parameters remain relatively stable, while auxiliary parameters are dynamically adjusted based on new data. The incremental learning process first selects the feature dimensions most correlated with the acquisition performance deviation, and then fine-tunes the parameters on these dimensions. An adaptive mechanism is used for the learning rate; when the improvement after multiple consecutive iterations is less than a threshold, the learning rate is automatically reduced. Parameter updates employ a sliding window mechanism, with new data gradually replacing old data to maintain the representativeness of the training samples. The optimized model parameters are updated using shadow copying technology. The effect is first verified on a newly created model copy, and the main model in the production environment is only replaced after the performance improvement is confirmed.

[0071] The data acquisition strategy evaluation report is generated using a structured document format. The main body of the report is divided into three parts: execution overview, performance analysis, and improvement recommendations. The execution overview describes the basic information and completion status of the data acquisition task; the performance analysis displays quantitative comparisons and trend charts of key indicators; and the improvement recommendations propose specific parameter adjustment schemes. The report appendix includes detailed data statistics tables and a summary of equipment operation logs. The evaluation report is versioned, with each newly generated report automatically linked to previous versions, forming a complete improvement trajectory. The report content is automatically generated using natural language generation technology, and key data visualization uses interactive charts.

[0072] The model version control engine's workflow comprises four phases. The first phase involves change impact analysis, identifying key improvements requiring a response in the assessment report. The second phase initiates the recalculation of topology node weights, with the weighting algorithm considering multiple factors such as device acquisition efficiency, network load, and power grid operational requirements. The third phase executes version switchover preparation, including verification testing of the new model and the development of rollback plans. The fourth phase implements the version release, employing a canary release mechanism to initially deploy the new version on select edge nodes, and then fully roll it out after stable operation. The version control engine maintains a complete change log, recording the specific content and implementation effects of each update.

[0073] The node configuration register update process emphasizes transaction integrity. New versions of register parameters use a differential update method, modifying only the changed configuration items. The update operation ensures consistency through a two-phase commit protocol: the preparation phase verifies the readiness of all relevant nodes, and the commit phase atomically applies all changes. The register version number uses a combination of timestamp and sequence number encoding, ensuring both uniqueness and chronological order. Version change event detection is achieved through a periodic checksum mechanism, where nodes periodically calculate the hash value of configuration parameters and reconcile it with the central system.

[0074] The edge computing layer's data acquisition and scheduling system employs hot-loading technology for seamless upgrades. New register parameters are first loaded into a spare memory area, and the processing logic is gradually migrated to the new version during system operation. This incremental update approach avoids service interruptions and ensures continuous data acquisition. Strict compatibility checks are performed during loading to ensure the new parameters match existing hardware devices and communication protocols. The system maintains a lightweight version adaptation layer to handle the conversion and mapping relationships between different version parameters.

[0075] The application of acquisition control weighting coefficients is reflected in the adjustment of task scheduling algorithms. High-weight devices are given priority for critical data acquisition tasks and receive more active time slices in time slot allocation. Weighting coefficients also affect the allocation of communication resources to devices; high-weight devices can use higher-priority transmission channels. The system periodically recalculates weighting coefficients to dynamically reflect changes in device performance. The weighting adjustment algorithm considers the aging characteristics of devices, appropriately relaxing evaluation criteria for devices with longer service lives.

[0076] The feedback mechanism of the dynamic environment topology model forms a closed-loop optimization system. The model continuously adjusts node weights and associated parameters based on the data acquisition performance evaluation results, and these adjustments in turn affect the generation of subsequent data acquisition strategies. The response time of the feedback loop is dynamically adjusted according to the power grid operating status; a longer response period is used during stable operation to reduce computational overhead, while a shorter response time is used under abnormal conditions to improve adaptability. The model optimization process retains complete decision-making data; every parameter change can be traced back to specific evaluation data and operational logs.

[0077] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0078] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence data collection method for a smart grid, characterized by, The method comprises the following steps: When a running state fluctuation event of a power grid node is monitored, a real-time scanning operation on a target collection area is started; Based on a set of power grid node parameters captured by the real-time scanning operation, a dynamic environment topology model containing power quality characteristic values is constructed; A cluster of intelligent terminal devices is deployed and a multi-dimensional performance test protocol is activated to obtain a data capture efficiency curve of the cluster of intelligent terminal devices under different communication protocols; The tested cluster of intelligent terminal devices is mapped to associated topology node positions of the dynamic environment topology model; A total amount of global data characteristic demand of the target collection area within a specified time window is calculated, and a theoretical collection load value of the cluster of intelligent terminal devices within the specified time window is derived in combination with the data capture efficiency curve; A collection parameter regulation instruction sequence for the cluster of intelligent terminal devices is generated according to the theoretical collection load value; The collection parameter regulation instruction sequence is executed to drive a state switching of a data collection unit of the cluster of intelligent terminal devices; After the collection action is completed, a collection efficiency index of the cluster of intelligent terminal devices is quantitatively evaluated and analyzed.

2. The artificial intelligence data collection method for a smart grid according to claim 1, wherein, The construction of the dynamic environment topology model comprises: The physical connection architecture of the target collection area is analyzed to identify key topology association information in the physical connection architecture; Power grid node identifiers corresponding to each key topology association information are extracted to form a power quality characteristic set; Data fluctuation tolerance thresholds of each power quality characteristic set are quantified, and external interference factor categories acting on the power quality characteristic set are detected; Based on a coupling relationship between the data fluctuation tolerance thresholds and the external interference factor categories, a data flow balance equation of the target collection area is constructed.

3. The artificial intelligence data collection method for a smart grid according to claim 2, wherein, The deployment of the cluster of intelligent terminal devices and the activation of the multi-dimensional performance test protocol comprise: The original data capture integrity of the cluster of intelligent terminal devices in a high-speed data transmission mode, a low-power standby mode and a protocol switching transition state is verified respectively; Time delay characteristic parameter packages of the cluster of intelligent terminal devices in the three operating modes are obtained; The time delay characteristic parameter packages are matched and verified with preset benchmark performance indexes to generate a confidence rating label of the data capture efficiency curve. 4.The artificial intelligence data collection method for smart grid according to claim 1, wherein, The mapping of the tested cluster of intelligent terminal devices to the associated topology node positions of the dynamic environment topology model comprises: The data fluctuation amplitudes of the power quality characteristic sets in the dynamic environment topology model are analyzed to locate a characteristic set identifier of a dominant data fluctuation source; Physical deployment coordinates of the cluster of intelligent terminal devices are spatially calibrated with topology coordinates corresponding to the characteristic set identifier of the dominant data fluctuation source, and the calibration result is written into a node configuration register of an edge computing layer. 5.The artificial intelligence data collection method for smart grid according to claim 1, wherein, The calculation of the total amount of global data characteristic demand of the target collection area within the specified time window comprises: The data characteristic generation estimates of each power quality characteristic set within the specified time window are aggregated to obtain the total amount of global data characteristic demand of the target collection area; Obtaining the inherent metadata generation amount of the intelligent terminal device cluster within the specified time window, superimposing it with the total amount of global data feature requirements, and outputting a quantization matrix of the theoretical collection load value; Based on the mapping relationship between the quantization matrix and the data capture efficiency curve, a collection flow allocation scheme of the intelligent terminal device cluster within the specified time window is generated.

6. The artificial intelligence data collection method for a smart grid according to claim 5, wherein, The collection parameter control instruction sequence for the intelligent terminal device cluster is generated according to the theoretical collection load value, which includes: Testing the maximum effective data throughput of the intelligent terminal device cluster within a single instruction cycle, and associating the time dimension parameter in the collection flow allocation scheme; Calculating the cumulative time slice that the collection unit of the intelligent terminal device cluster needs to remain in an active state within the specified time window; Converting the cumulative time slice into a time sequence control code in the collection strategy matrix.

7. The artificial intelligence data collection method for a smart grid according to claim 6, wherein, The execution of the collection parameter control instruction sequence to drive the working state switching of the data collection unit of the intelligent terminal device cluster includes: Dividing the specified time window into multiple continuous execution period units, and distributing the time sequence control code to each execution period unit according to a preset priority queue; After the current execution period unit ends, it is detected whether the actual collection data amount of the intelligent terminal device cluster meets the period allocation threshold; Based on the detection result, the time sequence control code in the next execution period unit is dynamically weighted and corrected until the entire collection scheduling of the specified time window is completed. 8.The smart grid artificial intelligence data collection method of claim 7, wherein, The method further includes an exception handling step: When a communication link interruption alarm event is monitored, the unexecuted time sequence control code segment in the collection strategy matrix is intercepted; According to the historical execution record of the intelligent terminal device cluster, the redundant transmission path of the unexecuted time sequence control code segment is reconstructed; The reconstructed redundant transmission path parameters are injected into the fault tolerance scheduling module of the intelligent terminal device cluster, an exception event set is generated and synchronized to the central control unit. 9.The artificial intelligence data collection method for smart grid according to claim 1, wherein, After the collection action is completed, the collection efficiency indicators of the intelligent terminal device cluster are quantitatively evaluated and analyzed, which includes: Obtaining the data quality difference value of the intelligent terminal device cluster before and after the collection task execution; calculating the actual feature collection load of the target collection area based on the data quality difference value; Comparing the deviation rate of the actual feature collection load and the theoretical collection load value; According to the deviation rate, the collection control weight coefficient of the intelligent terminal device cluster is determined; An incremental learning algorithm is used to optimize and iterate the parameter layer of the dynamic environment topology model, and the optimization result is fed back to the generation logic of the collection parameter control instruction sequence. 10.The artificial intelligence data collection method for smart grid according to claim 9, wherein, The method further includes a system updating step: Based on the collection control weight coefficient, a collection strategy evaluation report is established; the evaluation report is input into the model version control engine to trigger the topology node weight updating operation of the dynamic environment topology model; The updated topology node weight is re-bound with the deployment coordinates of the intelligent terminal device cluster to generate a new node configuration register version number; When the version number change event is detected, automatically load the new version of the register parameters to the data collection scheduling system of the edge computing layer.

Images