Electric power big data edge collaborative efficient processing method, system and medium

Abstract

The application provides a power big data edge collaborative efficient processing method, system and medium, and relates to the technical field of power system data processing. The method comprises the following steps: monitoring a target node whose running data exceeds a parameter out-of-limit threshold in the process of executing an initial task scheduling scheme; if the target node meets a collaborative regulation trigger level, screening out associated nodes corresponding to the target node based on a multi-parameter sensitivity matrix and a distribution network topology structure; constructing a global optimization model with the target of restoring the normal operation of the target node, and decomposing the global optimization model into local sub-problems of the associated nodes; adjusting an iteration step length coefficient according to the communication quality scores between the associated nodes; solving the local sub-problems according to the corrected iteration coefficient, and performing iteration through data interaction between the nodes to obtain collaborative optimal control solutions of the associated nodes, and then generating regulation and control instructions for execution. The method can improve the efficient collaborative processing efficiency of edge-side power big data.

Description

Technical Field

[0001] This application relates to the field of power system data processing technology, and in particular to a method, system, and medium for efficient edge collaborative processing of power big data. Background Technology

[0002] With the deepening of smart grid construction and the large-scale integration of new energy equipment, the distribution network is gradually evolving into a bidirectional energy flow system with massive monitoring nodes. In order to cope with grid parameter exceeding the limit events caused by high-frequency fluctuations of distributed power sources, the power system has introduced an edge computing architecture. By pushing computing power down to edge nodes close to the data source, the collaborative processing of big data in the distribution network and the generation of control strategies are completed by data interaction between multiple nodes and distributed computing.

[0003] In related technologies, when a power system detects abnormal conditions such as parameter exceeding limits in a local power grid, a topology partitioning mechanism based on a fixed geographical grid or physical distribution area is typically used. This mechanism directly activates all or some edge computing nodes within the jurisdiction of the abnormal area to participate in collaborative decision-making. During task execution, these selected nodes must participate in a distributed optimization algorithm and uniformly follow a pre-set fixed iteration cycle. After each round of calculation, each node frequently exchanges intermediate state variables with neighboring nodes according to a unified time rhythm. Through multiple rounds of repeated iterations, the computational data of all nodes tends to be consistent, ultimately achieving a global collaborative decision-making scheme.

[0004] However, with the increasing scale of distribution networks and the limitations of data carrying capacity in edge communication environments (such as wireless carriers and narrowband IoT), in mechanisms involving all or many nodes and fixed-frequency interaction, each node's local computation and data broadcasting consume fixed communication time and computing resources. This linear superposition effect is rapidly amplified in complex topology networks, resulting in significant time costs for the system in each round of synchronization waiting and data transmission. This cumulative resource consumption directly prolongs the overall time from anomaly detection to the generation of the final control strategy, leading to low collaborative processing efficiency of the entire system when dealing with high-frequency fluctuating power data. Summary of the Invention

[0005] This application provides a method, system, and medium for efficient edge collaborative processing of power big data, which addresses the problem of low coordination efficiency caused by the static and fixed characteristics of the collaborative response range and computational iteration mechanism in the edge collaborative processing schemes of distribution networks when facing complex network environments.

[0006] Firstly, this application provides a method for efficient edge collaborative processing of power big data, the method comprising: During the execution of the initial task scheduling scheme, the operating data of the distribution network edge nodes are monitored in real time to determine the target nodes whose operating data exceeds the preset parameter limit threshold, and the corresponding limit deviation rate of the target nodes. If the detected out-of-limit deviation rate meets the preset collaborative control trigger level, then based on the preset multi-parameter sensitivity matrix and distribution network topology, the associated nodes that have electrical coupling influence on the target node are selected from all distribution network edge nodes to construct an edge collaborative node group. A global optimization model is constructed with the goal of restoring the normal operation of the target node, and the global optimization model is decomposed into local sub-problems corresponding to each associated node in the edge collaborative node group based on a distributed collaborative optimization algorithm; The iteration step size coefficient is adjusted based on the communication quality score between each associated node in the edge collaborative node group to obtain the corrected iteration coefficient, which is positively correlated with the communication score. The local subproblems of each associated node in the edge collaborative node group are solved based on the corrected iteration coefficients, and the optimal collaborative control solution of each associated node is obtained through data interaction between nodes. Based on the cooperative optimal control solution, control commands are generated for each controllable device in the edge cooperative node group and then issued for execution.

[0007] By adopting the above technical solution, the system selects electrically coupled nodes based on the sensitivity matrix to build edge collaboration groups, and dynamically adjusts the iteration step size of solving local subproblems according to the communication quality. This reduces the computational participation of irrelevant nodes, reduces the time loss of data interaction and synchronization, accelerates the algorithm convergence speed, and thus achieves efficient collaborative processing of power big data on the edge side.

[0008] In some embodiments, before the step of real-time monitoring of the operating data of the distribution network edge nodes during the execution of the initial task scheduling scheme, determining the target node whose operating data exceeds a preset parameter limit threshold, and the limit deviation rate corresponding to the target node, the method further includes: Measure the performance parameters corresponding to each distribution network edge node. The performance parameters include computing power indicators, storage performance indicators, communication capability indicators, and reliability indicators calculated based on historical failure rate and historical task success rate. The performance parameters are weighted and summed according to the preset first dimension weighting coefficient to obtain the comprehensive capability index corresponding to each distribution network edge node; Obtain the task feature vector corresponding to each task in the task request queue to be processed. The task feature vector represents the latency sensitivity, computational complexity and spatial correlation. Based on the task feature vector and the comprehensive capability index, hierarchical matching between tasks and distribution network edge nodes is performed to obtain an initial task scheduling scheme.

[0009] By adopting the above technical solution, the system extracts multi-dimensional performance indicators of edge nodes for comprehensive quantification, and performs hierarchical matching with the latency sensitivity, complexity and other characteristics of tasks, so that power tasks are distributed more reasonably at the beginning, the node load of the underlying computing devices is balanced, and the resource allocation efficiency of the initial scheduling scheme of the entire computing network is improved.

[0010] In some embodiments, after the step of real-time monitoring of the operating data of the distribution network edge nodes during the execution of the initial task scheduling scheme, determining the target node whose operating data exceeds a preset parameter limit threshold, and the limit deviation rate corresponding to the target node, the method further includes: Based on the deviation rate of the target node's operating data relative to the parameter limit threshold, the current alarm level corresponding to the target node is determined. The current alarm level includes a warning level, a coordinated control trigger level, and a severe limit violation level. In response to the warning level, local controllable devices within the jurisdiction of the target node are invoked to perform local control. In response to the aforementioned severe limit violation level, the joint cloud platform makes global collaborative decisions on the edge collaborative node group.

[0011] By adopting the above technical solutions, the system classifies alarm levels based on the over-limit deviation rate and dynamically executes a hierarchical response mechanism that integrates local local control or cloud platform collaboration. This allows minor state fluctuations to be resolved quickly on-site and introduces high-dimensional computing power when there are serious over-limit situations, ensuring the safe operation of the power distribution network while taking into account response timeliness.

[0012] In some embodiments, before the step of adjusting the iteration step size coefficient based on the communication quality score between each associated node in the edge collaborative node group to obtain the corrected iteration coefficient, the method further includes: Each associated node in the edge collaborative node group sends probe data packets to each other to establish a communication link probe connection between node pairs; Within a sliding time window of a preset window length, the communication quality parameters between each pair of associated nodes are obtained. The communication quality parameters include average communication latency, latency jitter standard deviation, packet loss rate, and available bandwidth. The communication quality parameters are weighted and summed according to the preset second-dimensional weighting coefficients to obtain the communication quality score between each pair of associated nodes.

[0013] By adopting the above technical solution, the system extracts multi-dimensional feature parameters such as latency and packet loss of communication links using probe data packets and performs weighted comprehensive quantization, thereby realizing real-time all-round perception of the dynamic network status between edge nodes.

[0014] In some embodiments, the step of adjusting the iteration step size coefficient based on the communication quality score between each associated node in the edge collaborative node group to obtain the corrected iteration coefficient specifically includes: Obtain the standard deviation of delay jitter between each pair of associated nodes from the communication quality parameters; The preset baseline step size is attenuated and adjusted based on the standard deviation of the time delay jitter between each pair of associated nodes to obtain the initial iteration coefficient, which is negatively correlated with the standard deviation of the time delay jitter. The initial iteration coefficients are adjusted by gain based on the communication quality scores between each pair of associated nodes to obtain the corrected iteration coefficients. During the iterative process of solving the local subproblems of each associated node based on the corrected iterative coefficients, the changing trend of the original residuals between adjacent iteration steps is monitored. If the original residual continues to decrease in a predetermined number of iterations, then the corrected iteration coefficient is increased. If the original residual oscillates, then the corrected iteration coefficient is reduced.

[0015] By adopting the above technical solution, the system initially adjusts the iteration coefficients based on the objective score of communication, and adaptively corrects the step size according to the convergence and oscillation trend of the residuals during the solution process. This deeply couples the physical network state with the mathematical optimization characteristics, which not only ensures the robustness of the algorithm in the face of drastic environmental fluctuations, but also accelerates the process of the cooperative instructions approaching the global optimal solution to the greatest extent.

[0016] In some embodiments, the step of jointly making global collaborative decisions on the edge collaborative node group with the cloud platform in response to the severe limit violation level specifically includes: In response to the severe limit violation level, based on the multi-parameter sensitivity matrix locally cached by each associated node in the edge collaborative node group and the adjustment capability of the local controllable device corresponding to each associated node, the preliminary adjustment amount of each associated node to the target node is calculated, and the preliminary collaborative control scheme is obtained by summarizing. After sending the target node's operating data, the preliminary coordinated control scheme, and the global topology parameters of the distribution network to the cloud platform, the cloud platform performs a full-network power flow simulation on the preliminary coordinated control scheme based on the current operating data of the entire network, and obtains the predicted distribution of the network parameters after the execution of the preliminary coordinated control scheme. Based on the predicted distribution of parameters across the entire network, potential cascading risk nodes that generate operating data exceeding the corresponding parameter threshold due to the execution of the preliminary collaborative control scheme are identified, and these potential cascading risk nodes are included in the edge collaborative node group. Based on the results of the full network power flow simulation, preventive constraint margins are generated for each associated node in the extended edge collaborative node group and distributed to each associated node. The preventive constraint margins are used as additional safety constraints in the local subproblems corresponding to each associated node when constructing the global optimization model.

[0017] By adopting the above technical solutions, the system relies on the cloud platform to conduct full-network simulation verification of the initial collaborative scheme when the limit is significantly exceeded, and promptly issues safety constraint margins. The panoramic computing view compensates for the local decision-making blind spots caused by the limited field of view on the edge side, and blocks the risk of large-scale cascading failures in the distribution network that may be induced by one-sided excitation and control.

[0018] In some embodiments, the step of solving the local subproblems of each associated node in the edge collaborative node group based on the corrected iteration coefficients, and iterating through data interaction between nodes to obtain the collaborative optimal control solution of each associated node, specifically includes: After each iteration step is completed, the change in local solution of each associated node between the current iteration step and the previous iteration step is calculated; If there is a target associated node whose local solution change is lower than a preset marginal contribution threshold in a consecutive preset number of iteration steps, then the iteration participation state of the target associated node is switched from active state to locked state. In the locked state, the node no longer waits for the target associated node to send back local solution updates. Receive broadcast updates of global coordination variables, and calculate the coupling constraint residuals corresponding to the target associated node based on the received global coordination variables; If the coupling constraint residual exceeds the preset recovery threshold, the iteration participation state of the target associated node is restored from the locked state to the active state, and it re-participates in the iterative solution with the currently locked local solution as the initial value.

[0019] By adopting the above technical solution, the system temporarily locks nodes where local solution changes stagnate during iteration and wakes them up when the feedback of global constraint residuals comes into play. This directly eliminates the time spent waiting for invalid network synchronization in the later stages of distributed algorithm iteration, releasing computing resources and improving overall solution efficiency without sacrificing the accuracy of collaborative control decisions.

[0020] Secondly, this application provides a high-efficiency edge collaborative processing system for power big data, the system comprising: one or more processors and a memory; The memory is coupled to the one or more processors. The memory is used to store computer program code, which includes computer instructions. The one or more processors call the computer instructions so that the system can implement the power big data edge collaborative high-efficiency processing method provided in the above embodiments, which will not be described in detail here.

[0021] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on a power big data edge collaborative high-efficiency processing system, enable the system to implement a power big data edge collaborative high-efficiency processing method provided in the above embodiments, which will not be elaborated here.

[0022] Fourthly, this application provides a computer program product, including a computer program / instruction. When the computer program / instruction runs on the power big data edge collaborative high-efficiency processing system, the system can implement the power big data edge collaborative high-efficiency processing method provided in the above embodiments, which will not be elaborated here.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. In distributed computing, the system breaks through the limitations of static step size and fixed communication frequency in traditional algorithms, deeply integrating real-time states such as latency jitter and comprehensive scoring of the underlying communication link with the residual convergence trend in mathematical iteration. By constructing a dual feedback closed loop combining physical network sensing parameters and mathematical optimization characteristics, the system can adaptively and finely adjust the iteration step size for solving local subproblems. This avoids, to some extent, the algorithm oscillation and convergence difficulties caused by fixed step size in weak communication environments, while minimizing the time loss caused by frequent communication synchronization between nodes, thus improving the algorithm's robustness and collaborative processing efficiency in complex power grid environments.

[0024] 2. During iteration, the system proactively switches nodes with stagnant local solution changes to a locked state, pausing their network interactions to save bandwidth and synchronization waiting time. Simultaneously, utilizing the residual monitoring mechanism of global coordination variables, it accurately restores nodes to an active state to participate in the solution when strongly coupled changes are identified. This method directly eliminates invalid computation and communication overhead in the later stages of iteration, achieving full utilization of limited computing resources and a significant improvement in overall solution speed without reducing the accuracy of collaborative control.

[0025] 3. When encountering severe over-limit situations, the system does not rely on limited local data on the edge side to directly generate the final strategy. Instead, it incorporates the cloud into the edge collaboration group and uses the global data model of the cloud platform to perform a full-network power flow simulation of the preliminary control scheme. This allows for the accurate identification of the risk of cascading over-limit situations in related nodes that may be caused by the scheme, and the risk is transformed into preventive security margin constraints. This balances the core technical requirements of quickly resolving abnormal states at the lower level and ensuring the security defense of the entire network system. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating an efficient edge-collaborative processing method for power big data in an embodiment of this application. Figure 2 This is a schematic diagram of a system performing layered collaborative processing of power big data in an embodiment of this application; Figure 3 This is another flowchart illustrating a method for efficient edge collaborative processing of power big data in this application embodiment; Figure 4 This is a schematic diagram of the physical device structure of the power big data edge collaborative high-efficiency processing system in this application embodiment. Detailed Implementation

[0027] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0029] The large-scale integration of new energy sources into the power grid results in data that is massive, heterogeneous, and requires high real-time performance. Traditional centralized processing models suffer from high latency and bandwidth pressure. While edge computing can alleviate these problems, existing edge collaborative processing suffers from low node collaboration efficiency, rigid iteration mechanisms, and resource waste due to the participation of irrelevant nodes. The power big data edge collaborative high-efficiency processing system (hereinafter referred to as the system) provided in this application can accurately identify abnormal nodes based on real-time operational data from edge nodes. It filters collaborative nodes through electrical coupling characteristics, dynamically adjusts iterative parameters to complete distributed optimization solutions, and ultimately generates precise control commands. This enables efficient collaborative processing of power big data and rapid response to distribution network anomalies, thereby improving the stability and intelligence level of distribution network operation. For details, please refer to [link to details]. Figure 1 This is a flowchart illustrating a method for efficient edge collaborative processing of power big data in an embodiment of this application.

[0030] S101. During the execution of the initial task scheduling scheme, the operating data of the distribution network edge nodes are monitored in real time to determine the target nodes whose operating data exceeds the preset parameter limit threshold, and the corresponding limit deviation rate of the target nodes.

[0031] Among them, the distribution network edge node refers to the hardware node deployed near the data source in the distribution network, which has the ability to collect, calculate, and communicate data, and can complete local data processing and edge collaborative computing; the operation data refers to the various electrical parameter data collected by the distribution network edge node during the operation of the distribution network; the initial task scheduling scheme refers to the task allocation and execution scheme obtained by the system after hierarchical matching of distribution network big data processing tasks based on the comprehensive capabilities and task characteristics of the edge node; the parameter limit threshold refers to the critical value of the normal operating range preset by the system for various operating electrical parameters of the distribution network. If the value is exceeded, it is judged as a parameter abnormality; the limit deviation rate refers to the percentage of the abnormal operation data of the target node that exceeds the corresponding parameter limit threshold relative to the parameter limit threshold, which is used to quantitatively characterize the severity of the target node parameter deviating from the normal operating range.

[0032] This step is executed when the system has completed the initial task scheduling plan and entered the plan execution phase. The scenario is that the system performs routine real-time monitoring of the operating status of edge nodes across the entire distribution network and promptly identifies nodes with abnormal parameters in the operation of the distribution network.

[0033] Specifically, during the process of allocating and executing various power big data processing tasks according to the initial task scheduling scheme, the system continuously receives and monitors the distribution network operation data collected and uploaded by each distribution network edge node in real time. The monitored operation data covers core electrical parameters such as voltage deviation rate, current over-limit rate, power fluctuation rate, frequency deviation, and harmonic distortion rate. The system compares the real-time operation data of each edge node with the parameter over-limit thresholds set in advance for each parameter. When it is detected that any one or more operation data values ​​of an edge node exceed the corresponding parameter over-limit threshold, the system marks the edge node as a target node, thus completing the identification and determination of abnormal nodes.

[0034] Optionally, the system can first deploy a data monitoring agent program locally on each edge node. The agent program collects the operating data of the local jurisdiction in real time and compares it with the parameter limit thresholds cached locally. If the data exceeds the limit, the node is marked as a suspected target node. Then, the operating data and marking information of the suspected target node are uploaded to the fog node in the upper layer of the system. The fog node performs a second verification on the data. After the verification is passed, it is officially determined as a target node. Optionally, the system can first have fog nodes uniformly receive real-time operating data uploaded by each edge node, establish a centralized monitoring database at the fog node, divide the monitoring area according to the distribution network area, perform batch comparative analysis of the operating data of each area, directly identify edge nodes whose data exceeds the parameter limit threshold as target nodes, and send the target node information to the corresponding edge node.

[0035] It should be noted that the larger the deviation rate, the more severe the deviation of the target node's operating parameters from the normal range, and the greater the required control effort and coordination range. When multiple operating data of the target node simultaneously exceed the limits, the system can calculate the deviation rate corresponding to each parameter separately and take the maximum value as the comprehensive deviation rate of the target node for subsequent alarm level determination. In a specific embodiment, the formula for calculating the deviation rate can be: Exceedance deviation rate = (actual value of operating data - parameter exceedance threshold) / parameter exceedance threshold × 100%, where the actual value of operating data is the electrical parameter value that exceeds the threshold currently monitored by the target node, and the parameter exceedance threshold is the critical value of the normal operating range corresponding to this type of parameter preset by the system.

[0036] For example, if the actual voltage deviation rate of a feeder terminal node is +9.2%, while the system's preset over-limit threshold for voltage deviation rate is +7%, then the over-limit deviation rate of this node is (9.2% − 7%) / 7% × 100% ≈ 31.4%, indicating that the voltage deviation of this node has exceeded the over-limit threshold by approximately 31.4%. The system will match the corresponding alarm level in subsequent steps based on the magnitude of the over-limit deviation rate to determine the appropriate control response strategy.

[0037] S102. If the detected deviation rate meets the preset collaborative control trigger level, then based on the preset multi-parameter sensitivity matrix and distribution network topology, select the associated nodes that have electrical coupling effects on the target node from all distribution network edge nodes, and construct an edge collaborative node group.

[0038] Among them, the coordinated control trigger level refers to the parameter abnormality level preset by the system that triggers the coordinated control operation of multiple nodes, which is an abnormality level between the warning level and the severe over-limit level; the multi-parameter sensitivity matrix refers to the matrix preset by the system that describes the degree of mutual influence of electrical parameters between nodes in the distribution network, including voltage sensitivity matrix, frequency sensitivity matrix, harmonic propagation matrix, etc.; the distribution network topology refers to the connection method and layout relationship between electrical nodes, lines and equipment in the distribution network.

[0039] After identifying the target node, the system extracts the out-of-limit operation data of the target node and calculates its deviation rate relative to the parameter out-of-limit threshold. Based on the deviation rate, the system determines the abnormality level corresponding to the target node. If the determination result meets the preset collaborative control trigger level, the multi-node collaborative control process is initiated.

[0040] Specifically, the system pre-sets three alarm levels and corresponding over-limit deviation rate judgment ranges for various operating parameters of the distribution network. The three alarm levels, from low to high, are: early warning level, coordinated control trigger level, and severe over-limit level, corresponding to three abnormal degrees of slight deviation, moderate over-limit, and significant over-limit, respectively. The system configures a corresponding over-limit deviation rate range boundary for each alarm level. When the over-limit deviation rate of a target node falls into the range corresponding to a certain level, the target node is determined to meet that alarm level.

[0041] The determination process for the coordinated control trigger level is as follows: The system compares the out-of-limit deviation rate calculated in step S101 with the preset lower threshold values ​​for both the coordinated control trigger level and the severe out-of-limit level for the target node's abnormal operating data. If the target node's out-of-limit deviation rate is greater than or equal to the lower threshold value for the coordinated control trigger level, and less than the lower threshold value for the severe out-of-limit level, then the target node's out-of-limit deviation rate meets the coordinated control trigger level, and the system initiates the multi-node coordinated control process. If the out-of-limit deviation rate is lower than the lower threshold value for the coordinated control trigger level, it is determined to be a warning level, and the target node calls upon locally controllable devices within its local jurisdiction to perform local control. If the out-of-limit deviation rate is greater than or equal to the lower threshold value for the severe out-of-limit level, it is determined to be a severe out-of-limit level, and while initiating multi-node coordinated control, global coordinated decision-making is conducted in conjunction with the cloud platform. When multiple operating parameters of a target node exceed their limits simultaneously, the system responds by using the alarm level corresponding to the maximum deviation rate of each parameter as the final alarm level for that target node.

[0042] In one specific embodiment, the system can implicitly define the over-limit deviation rate range corresponding to each alarm level based on the absolute thresholds of various operating parameters. For example, for the voltage deviation rate parameter, the system presets the parameter over-limit threshold (i.e., the warning level threshold) to be ±5%, the absolute threshold corresponding to the coordinated control trigger level to be ±7%, and the absolute threshold corresponding to the severe over-limit level to be ±10%. When the actual value of the voltage deviation rate of a target node is +9.2%, the over-limit deviation rate relative to the parameter over-limit threshold of +5% is (9.2%−5%) / 5%×100%=84%, and the over-limit deviation rate relative to the absolute threshold of +7% corresponding to the coordinated control trigger level is (9.2%−7%) / 7%×100%≈31.4%. This actual value of +9.2% exceeds the absolute threshold of +7% corresponding to the coordinated control trigger level but does not reach the absolute threshold of +10% corresponding to the severe over-limit level. Therefore, the system determines that the over-limit deviation rate of the target node meets the coordinated control trigger level and initiates the multi-node coordinated control process.

[0043] Similarly, for the frequency deviation parameter, the warning level threshold is ±0.2Hz, the absolute threshold corresponding to the coordinated control trigger level is ±0.5Hz, and the absolute threshold corresponding to the severe over-limit level is ±1Hz; for the harmonic distortion rate parameter, the warning level threshold is 5%, the absolute threshold corresponding to the coordinated control trigger level is 8%, and the absolute threshold corresponding to the severe over-limit level is 12%.

[0044] Furthermore, the system retrieves a preset multi-parameter sensitivity matrix and real-time distribution network topology data from local storage. Based on the type of out-of-limit parameter of the target node, it selects the corresponding sensitivity matrix, searches for the column elements of the row corresponding to the target node in the matrix, and filters out edge nodes whose sensitivity coefficients exceed the preset threshold. These nodes have a strong electrical coupling effect with the target node. At the same time, combined with the distribution network topology and considering the parameter propagation characteristics of the radial topology, the upstream and downstream adjacent edge nodes of the target node are also included in the filtering range and identified as associated nodes. Finally, the target node and all associated nodes are integrated to form an edge collaborative node group.

[0045] In some embodiments, the acquisition and updating of the multi-parameter sensitivity matrix can be achieved in the following ways: A mathematical model of the distribution network is established, and a node admittance matrix is ​​constructed. This matrix reflects the topology and line parameters of the distribution network. Each element in the matrix represents the electrical connection relationship and impedance characteristics between nodes. A set of power flow equations is established, including active power balance equations and reactive power balance equations. The equations contain variables such as voltage amplitude, phase angle, and injected power of each node. A frequency dynamic equation is established to describe the dynamic relationship between system frequency and active power imbalance, taking into account the speed governor characteristics of generators and the frequency characteristics of loads.

[0046] Numerical calculations of the sensitivity matrix were performed. The Newton-Raphson power flow calculation method was used to solve for the voltage distribution at the current operating point. After power flow convergence, the Jacobian matrix of the power flow equations was calculated. The Jacobian matrix contains four submatrices: partial derivatives of voltage magnitude with respect to active power, partial derivatives of voltage magnitude with respect to reactive power, partial derivatives of voltage phase angle with respect to active power, and partial derivatives of voltage phase angle with respect to reactive power. By performing mathematical transformations and inversion operations on the Jacobian matrix, the voltage sensitivity matrix to power injection was obtained. This matrix describes the degree of influence of power changes at each node on the voltage at each node. Small-signal analysis was used to calculate frequency sensitivity. By applying small-amplitude active power disturbances to each node, the frequency response changes of the system were observed, and the sensitivity coefficient of frequency to active power injection at each node was calculated.

[0047] Harmonic propagation characteristics are analyzed, and a harmonic equivalent circuit model of the distribution network is established. Considering the frequency correlation characteristics of the lines and the harmonic impedance characteristics of the equipment, a harmonic power flow calculation method is adopted. Harmonic currents of a specific order are injected into each node, and the harmonic voltage distribution of each node is calculated. By changing the harmonic injection point and injection amplitude, a harmonic propagation coefficient matrix is ​​established. This matrix describes the propagation influence of harmonic sources at each node on the harmonic levels of other nodes.

[0048] For online updates of the sensitivity matrix, the system continuously monitors changes in the operating status of the distribution network. When changes in load level exceeding 10%, distributed power output exceeding 15%, or changes in network topology (such as switching operations or transformer tap adjustments) are detected, the sensitivity matrix is ​​automatically recalculated. Incremental calculation methods can be used to improve computational efficiency, recalculating only the changed parts while using historical calculation results for the unchanged parts.

[0049] S103. Construct a global optimization model with the goal of restoring the normal operation of the target node, and decompose the global optimization model into local sub-problems of each associated node in the corresponding edge collaborative node group based on the distributed collaborative optimization algorithm.

[0050] Among them, the global optimization model refers to the mathematical model constructed by the system to restore the operating parameters of the target node to normal, which includes the optimization objective and constraints; the distributed collaborative optimization algorithm refers to the optimization algorithm that can decompose the global optimization problem into multiple local sub-problems, which are solved independently by each node and then achieve the global optimal solution through data interaction; the local sub-problem refers to the sub-optimization problem of a single associated node in the edge collaborative node group after the global optimization model is decomposed, which only includes the local constraints and optimization objective of that node.

[0051] Specifically, the system aims to restore the out-of-limit operating parameters of the target node to the normal range, while also considering the overall operational stability and control costs of the distribution network. A global optimization model is constructed, and the optimization objective of the global optimization model is to minimize the weighted sum of squares of voltage deviation, frequency deviation, and harmonic distortion rate of each node, plus the weighted sum of control costs. The system sets weight coefficients for the sum of squares of voltage deviation, frequency deviation, harmonic distortion rate, and control costs. The weight coefficients are configured differently according to the type and severity of the out-of-limit parameters of the current target node. For example, when the target node is mainly experiencing voltage out-of-limit issues, the weight coefficient corresponding to the sum of squares of voltage deviation is increased, so that the optimization solution focuses more on the voltage recovery target.

[0052] It should be noted that the sum of squares of voltage deviation reflects the voltage quality level after coordinated regulation, the sum of squares of frequency deviation reflects the stability of the system frequency, and the sum of squares of harmonic distortion rate reflects the power quality level. These three measures the effectiveness of coordinated regulation in voltage management, frequency maintenance, and harmonic suppression, respectively. Control cost reflects the comprehensive cost incurred by controllable equipment in performing regulation actions, including battery depreciation costs caused by charging and discharging of energy storage systems, operating costs caused by continuous operation of active filters, and economic losses caused by interruption or reduction of flexible loads.

[0053] Furthermore, the constraints of the global optimization model include, but are not limited to, the following aspects: First, the upper and lower limits of the voltage of each node are constrained, requiring that the voltage deviation rate of each node be controlled within the preset normal range after coordinated regulation, for example, the voltage deviation rate is controlled within ±3%, to ensure that the regulation scheme will not cause the voltage of other nodes to exceed the normal range while restoring the voltage of the target node. Second, system frequency constraints require that the system frequency deviation be controlled within a preset range after coordinated regulation, for example, within ±0.2Hz, to ensure system active power balance and frequency stability. Third, harmonic constraints require that the harmonic distortion rate of each node be controlled within a preset limit after coordinated regulation, for example, the harmonic distortion rate be controlled within 5%, to prevent the introduction of new harmonic pollution during the regulation process. Fourth, the capacity constraint of line transmission power ensures that the transmission power of each line does not exceed its thermal stability limit after coordinated regulation, so as to avoid safety hazards caused by line overload. Fifth, constraints on the adjustment capabilities of controllable devices. For different types of controllable devices managed by each associated node in the edge collaborative node group, corresponding adjustment range limits are set, including active and reactive power adjustment range constraints for photovoltaic inverters, charging and discharging power range constraints and upper and lower limits of state of charge constraints for energy storage systems, harmonic compensation capacity range constraints for active filters, adjustable capacity limits and minimum power supply guarantee constraints for flexible loads, etc., to ensure that the control scheme obtained by optimization does not exceed the actual adjustment capabilities of each device.

[0054] Furthermore, the system employs the alternating direction multiplier method as the framework for the distributed collaborative optimization algorithm, decomposing the aforementioned global optimization model into local sub-problems for each associated node in the corresponding edge collaborative node group. Specifically, the system identifies coupling variables between associated nodes in the global optimization model. In the radial topology of the distribution network, adjacent associated nodes are coupled through line transmission power; that is, the transmission power of the same line, calculated from the perspectives of the nodes at both ends of the line, should be consistent. This consistency condition constitutes the coupling constraint between associated nodes. For each line connecting adjacent associated nodes, the system uses the line transmission power calculated by the nodes at both ends of the line as the coupling variable.

[0055] Then, the system introduces Lagrange multipliers to relax the aforementioned coupling constraints. Specifically, for each coupling constraint, the system assigns a corresponding Lagrange multiplier, transforming the coupling constraint from a hard constraint into a penalty term in the objective function. The objective function of the local subproblem of each associated node consists of three parts: the first part is the weighted sum of the squared multi-parameter deviations of the associated node and the weighted sum of the control costs of the local controllable equipment, reflecting the node's own control effect and control cost; the second part is the Lagrange multiplier penalty term related to the associated node, i.e., the product of the Lagrange multiplier and the difference of the node's coupling variables, used to transfer global coupling information to the local subproblem; the third part is a quadratic penalty term, i.e., the square of the difference of coupling variables multiplied by half the penalty coefficient, used to enhance the convergence performance of the algorithm. The constraints of the local subproblem of each associated node only include the local electrical parameter range constraints of the node and the adjustment capability constraints of the controllable equipment under the node's jurisdiction, and do not involve the variables and constraints of other associated nodes.

[0056] Through the above decomposition, the global optimization model is transformed into a set of local subproblems that can be solved independently and in parallel by each associated node in the edge collaborative node group. The local subproblems are coupled and coordinated through Lagrange multipliers and global coordination variables. In the subsequent iterative solution process, each associated node only needs to exchange local solutions and multiplier information related to the coupling variables, without sharing complete local running data and device parameters. This not only ensures the privacy of distributed solution, but also reduces the amount of communication data between nodes.

[0057] S104. Adjust the iteration step size coefficient based on the communication quality score between each associated node in the edge collaborative node group to obtain the corrected iteration coefficient.

[0058] Among them, the communication quality score refers to the score obtained by the system after quantitatively evaluating the quality of the communication link between each associated node in the edge collaborative node group, which comprehensively reflects indicators such as communication latency, packet loss rate, and bandwidth; the iteration step size coefficient refers to the coefficient used to control the update magnitude of variables during the iterative process of solving the optimization problem.

[0059] Specifically, the system obtains communication quality scores between associated nodes within the edge collaborative node group. These scores comprehensively consider communication quality parameters such as average communication latency, latency jitter standard deviation, packet loss rate, and available bandwidth. A higher score indicates better communication link quality. The system presets a baseline iteration step size coefficient, which serves as the basis for adjustment. Based on the positive correlation between communication quality scores and the iteration step size coefficient, the baseline iteration step size coefficient is adjusted. For node pairs with high communication quality scores, the system appropriately increases the iteration step size coefficient to accelerate the iterative solution speed; for node pairs with low communication quality scores, the system appropriately decreases the iteration step size coefficient to ensure the stability of the iterative solution. During the adjustment process, the system combines the differences in communication quality scores of each node pair to perform differentiated coefficient adjustments, avoiding the problem of insufficient algorithm adaptability caused by uniform adjustment, and finally generating corresponding corrected iteration coefficients for the iterative solution between nodes within the edge collaborative node group.

[0060] In some embodiments, the system can obtain the baseline iteration step size coefficient and the grading standard of the communication quality score, divide the score into different levels, set a corresponding coefficient adjustment ratio for each level, match the communication quality score of each node pair to the corresponding level, adjust the baseline iteration step size coefficient by multiplication according to the adjustment ratio to obtain the initial correction coefficient, and finally fine-tune the initial correction coefficient according to the standard deviation of the delay jitter between nodes. If the delay jitter is large, the coefficient is appropriately reduced to obtain the final corrected iteration coefficient.

[0061] S105. Solve the local subproblems of each associated node in the edge collaborative node group based on the corrected iteration coefficients, and iterate through data interaction between nodes to obtain the collaborative optimal control solution of each associated node.

[0062] Among them, the cooperative optimal control solution refers to the optimal value of the controllable equipment corresponding to each associated node, which can restore the target node parameters to normal and meet the global optimization objective and constraint conditions, obtained by the system through distributed iterative solution.

[0063] Specifically, the system distributes the corrected iteration coefficients to each associated node in the edge collaborative node group, and simultaneously distributes the corresponding local subproblem and initial Lagrange multipliers to each node. Each associated node uses the corrected iteration coefficients as parameters and independently solves its local subproblem using a sequential quadratic programming method. The nonlinear local subproblem is approximated twice at the current point, and the quadratic programming subproblem is solved to obtain the search direction. A line search is performed along the search direction to determine the step size, and the local control variable values ​​are updated to obtain the local preliminary solution.

[0064] Each associated node uploads its local preliminary solution to the main coordinating node within the group. After collecting the preliminary solutions from all nodes, the main coordinating node calculates the original residuals and dual residuals between the nodes. The original residuals measure the degree of inconsistency between the local solutions of each node, while the dual residuals measure the degree of change in the Lagrange multipliers. If the residuals do not reach the preset convergence threshold, the main coordinating node updates the Lagrange multipliers using the gradient ascent method based on the original residuals and the corrected iteration coefficients, and distributes the updated multipliers to each associated node. Each node then solves the local subproblem again based on the new multipliers, repeating the above iterative process of "local solution - data interaction - multiplier update".

[0065] When the residual reaches the convergence threshold, the iteration process terminates. The system determines the local solution of each associated node at this time as the cooperative optimal control solution. This solution satisfies the local constraints of each node and achieves the global optimization objective, so that the out-of-limit parameters of the target node return to normal.

[0066] In some embodiments, the system can first have the main coordinating node uniformly distribute the corrected iteration coefficients and initial parameters. Each associated node independently completes the solution of its local subproblem and uploads the results. The main coordinating node centrally summarizes the results and calculates the residuals. If convergence is not achieved, the Lagrange multipliers are uniformly updated and redistributed until convergence is achieved. The main coordinating node then organizes the final local solutions of each node into a cooperative optimal control solution and distributes it to each node.

[0067] It should be noted that, to ensure the real-time performance of distributed optimization, the system sets a maximum iteration limit, for example, 50 iterations. Even if the algorithm does not fully converge to the optimal solution, it will terminate after reaching the maximum number of iterations and output the current solution as an approximate optimal solution. Simultaneously, the system introduces a communication timeout handling mechanism, setting a timeout period (e.g., 200 milliseconds) for each inter-node communication. If the master coordinating node does not receive a response from a participating node within the timeout period, a predictive compensation strategy is adopted. This strategy uses the node's state data from the previous iteration to estimate and continue the optimization process. When communication is restored, the actual received data is used for correction.

[0068] S106. Generate control commands for each controllable device in the edge collaborative node group based on the collaborative optimal control solution and issue them for execution.

[0069] Among them, controllable equipment refers to electrical equipment in the distribution network whose parameters can be adjusted by the system to improve the operating status of the distribution network, including photovoltaic inverters, energy storage converters, active filters, load controllers, etc.; control instructions refer to instruction information generated by the system based on the cooperative optimal control solution, which includes controllable equipment adjustment parameters and execution requirements.

[0070] Specifically, the system collects the optimal control solution for each associated node in the edge collaborative node group. Based on the jurisdiction of each node, the control solution is broken down into specific adjustment parameters for each controllable device, including active power adjustment, reactive power adjustment, filter compensation capacity, load adjustment ratio, etc. Then, standardized control instructions are generated for each controllable device. These instructions include a unique device identifier, control type, target setpoint, execution priority, and timestamp. For multiple devices requiring coordinated action, a synchronization identifier is added to ensure timing consistency.

[0071] Next, the system establishes a communication channel to send control commands from the main coordinating node to the corresponding controllable devices via the edge gateway. After receiving the commands, the devices parse the command content and verify its legality and timeliness. They then execute the adjustment actions according to the target settings and time requirements in the commands, adjust the device operating parameters, and simultaneously feed back the execution status and adjusted operating data to the system. The system monitors the control effect in real time.

[0072] Based on steps S101 to S106, this application further achieves precise hierarchical matching of tasks and nodes through steps S201 to S209, generates a scientific initial task scheduling scheme, classifies alarm levels based on the deviation rate of abnormal nodes and executes differentiated control strategies, and combines cloud computing power to achieve global collaborative prevention and control of severe over-limit, thereby improving the efficiency and security of power big data edge collaborative processing from the source of task scheduling and the entire process of abnormal handling.

[0073] Please refer to the details. Figure 2 This is a schematic diagram of a system performing layered collaborative processing of power big data in an embodiment of this application; S201. Measure the performance parameters corresponding to each distribution network edge node, and sum the performance parameters according to the weight coefficient of the first dimension to obtain the comprehensive capability index corresponding to each distribution network edge node.

[0074] The first dimension weight coefficient refers to the weight value preset by the system for different dimension performance parameters of the distribution network edge node, which is used to calculate the comprehensive capability index. It is set according to the importance of each performance parameter to the node's processing task.

[0075] This step is performed when the system has not yet generated an initial task scheduling plan and a comprehensive assessment of the processing capabilities of all edge nodes in the distribution network is required. The scenario is the preliminary resource assessment for the system to carry out power big data processing tasks, providing a quantitative basis for the matching of subsequent tasks and nodes.

[0076] Specifically, the system comprehensively measures the multi-dimensional performance parameters of each edge node in the distribution network, acquiring the node's computing power, storage performance, and communication capabilities. Simultaneously, it combines historical operational data of each node to calculate historical failure rates and historical task success rates, thereby obtaining the node's reliability index and forming a complete node performance parameter system. The system retrieves preset first-dimensional weighting coefficients, which assign differentiated weights to different performance parameters such as computing power, storage, communication, and reliability. After standardizing each node's performance parameters, they are weighted according to their corresponding weighting coefficients. The weighted results of each parameter are summed to obtain a unique comprehensive capability index for each distribution network edge node. A higher index value indicates a stronger overall processing capability of the node.

[0077] S202. Obtain the task feature vector corresponding to each task in the task request queue to be processed, and perform hierarchical matching between tasks and distribution network edge nodes based on the task feature vector and comprehensive capability index to obtain the initial task scheduling scheme.

[0078] Among them, the task request queue refers to the ordered set of all power big data processing tasks collected by the system and waiting to be assigned to various edge nodes for processing; the task feature vector refers to the vector used to characterize the core processing attributes of the task, and its dimensions include key features such as latency sensitivity, computational complexity and spatial correlation; hierarchical matching refers to the matching method by which the system assigns tasks with different characteristics to different processing levels and different edge nodes according to the task feature vector and the node comprehensive capability index.

[0079] Specifically, the system sorts through the collected queue of pending task requests, extracting the core features of each task, including latency sensitivity, computational complexity, and spatial correlation. These features are quantified to construct a unique task feature vector for each task. The system establishes hierarchical matching rules between tasks and nodes, determining the processing level requirements of tasks based on their feature vectors. For example, tasks with high latency sensitivity and low computational complexity are suited for edge layer processing, while tasks with low latency sensitivity and high computational complexity are suited for cloud processing. Simultaneously, the system combines the comprehensive capability index of each edge node to match tasks at corresponding processing levels to nodes whose capability index matches the task's processing requirements. For regional tasks, they are matched to core edge nodes with strong communication capabilities and high comprehensive capability indices. After matching all tasks with nodes, the system clarifies the execution node, execution level, estimated execution time, and task priority for each task, integrating these information to form a complete initial task scheduling scheme.

[0080] S203. During the execution of the initial task scheduling scheme, the target node is determined, and the current alarm level is determined based on the deviation rate of the target node's running data relative to the parameter limit threshold.

[0081] Among them, the deviation rate refers to the difference between the abnormal operating data of the target node and the parameter deviation threshold, which is a percentage relative to the parameter deviation threshold and is used to quantitatively characterize the degree of abnormality of the node parameter; the current alarm level refers to the alarm level determined by the system based on the deviation rate of the target node and corresponding to the degree of parameter abnormality, including the early warning level, the collaborative control trigger level, and the severe deviation level.

[0082] Specifically, during the allocation and execution of various power big data processing tasks according to the initial task scheduling scheme, the system continuously monitors the real-time operating data of each distribution network edge node, compares the operating data with preset parameter over-limit thresholds, and identifies target nodes whose operating data exceeds the thresholds. For each abnormal operating data of the target node, the system calculates the difference between it and the corresponding parameter over-limit threshold, divides the difference by the parameter over-limit threshold to obtain the over-limit deviation rate, and matches the system's preset alarm level classification standard based on the magnitude of the over-limit deviation rate. If the over-limit deviation rate is small and does not reach the coordinated control triggering standard, it is judged as a warning level; if the over-limit deviation rate is moderate and reaches the coordinated control triggering standard but does not reach the severe over-limit standard, it is judged as a coordinated control triggering level; if the over-limit deviation rate is large and exceeds the severe over-limit standard, it is judged as a severe over-limit level, and finally determines the current alarm level corresponding to the target node.

[0083] In one specific embodiment, the warning level can be set with the following conditions: a warning is triggered when the voltage deviation rate exceeds +5% or is lower than -5%, the frequency deviation exceeds ±0.2Hz, and the harmonic distortion rate exceeds 5%; the coordinated control trigger level can be set with the following conditions: a general over-limit alarm is triggered when the voltage deviation rate exceeds +7% or is lower than -7%, the frequency deviation exceeds ±0.5Hz, and the harmonic distortion rate exceeds 8%; the severe over-limit level can be set with the following conditions: a severe over-limit alarm is triggered when the voltage deviation rate exceeds +10% or is lower than -10%, the frequency deviation exceeds ±1Hz, and the harmonic distortion rate exceeds 12%. S204. Within the jurisdiction of the target node, invoke locally controllable devices to perform local control.

[0084] Among them, the jurisdiction area refers to the specific area in the distribution network where the target node is responsible for data collection, monitoring and equipment control; locally controllable equipment refers to power equipment deployed within the jurisdiction area of ​​the target node, directly controlled by the target node, and capable of adjusting operating parameters; local control refers to the control method that restores the abnormal operating data of the target node to the normal range only within the jurisdiction area of ​​the target node by adjusting the operating parameters of the locally controllable equipment.

[0085] Specifically, after determining that the current alarm level of the target node is a warning level, the system initiates a local control process. The target node retrieves information on all locally controllable devices within its jurisdiction, including device type, operating status, and regulation capacity. Based on the type of abnormal operating data of the target node, the system selects the corresponding locally controllable device; for example, if the voltage deviation is abnormal, it selects the energy storage system and reactive power compensation device; if the frequency deviation is abnormal, it selects the photovoltaic inverter and energy storage converter. The target node issues control commands to the selected locally controllable devices, specifying the device's regulation parameters, regulation range, and execution time. Upon receiving the commands, the controllable devices execute the regulation actions. The target node continuously monitors changes in operating data within its jurisdiction until the abnormal data returns to the normal range, completing the local control. If the data still does not recover after control, the system escalates the alarm level and initiates subsequent collaborative control processes.

[0086] S205. Construct edge collaborative node groups for coordination and control.

[0087] Specifically, after the system initiates the multi-node collaborative control process, the upstream fog node or the associated node with the highest comprehensive capability index in the edge collaborative node group of the target node serves as the main coordinating node. The main coordinating node retrieves a preset multi-parameter sensitivity matrix and real-time distribution network topology data from local storage. Based on the abnormal parameter type of the target node (such as voltage, frequency, harmonic anomalies), the corresponding sensitivity matrix is ​​selected, and the elements in each column of the corresponding row of the target node in the matrix are searched. Strongly coupled nodes with sensitivity coefficients exceeding preset thresholds are selected, and the adjustment actions of these nodes have a significant improvement effect on the abnormal parameters of the target node. At the same time, combined with the radial topology of the distribution network, the upstream and downstream adjacent nodes of the target node are included in the collaborative scope. The selected nodes are checked for fault status and online status, and invalid nodes are removed. Together with the target node, they form an edge collaborative node group, and the main coordinating node is responsible for coordinating and controlling all aspects of the group's work.

[0088] S206. Based on the multi-parameter sensitivity matrix of each associated node in the edge collaborative node group and the adjustment capability of the local controllable device corresponding to each associated node, calculate the preliminary adjustment amount of each associated node to the target node, and summarize to obtain the preliminary collaborative control scheme.

[0089] Specifically, after a system response severely exceeds limits, each associated node in the edge collaborative node group retrieves its locally cached multi-parameter sensitivity matrix. Based on the type of abnormal parameters of the target node and its own electrical coupling relationship with the target node, it determines the influence coefficient of its local controllable devices on the abnormal parameters of the target node. Each associated node, considering the maximum adjustment capacity of its local controllable devices and their current operating status, calculates the initial adjustment amount that can improve the abnormal parameters of the target node, clarifying the adjustment direction, adjustment range, and execution priority of each controllable device. Each associated node reports the calculated initial adjustment amount to the main coordination node. The main coordination node summarizes and organizes all initial adjustment amounts, verifies the rationality of the adjustment amounts and their matching with the device's adjustment capacity, eliminates unreasonable adjustment amounts exceeding the device's adjustment range, and classifies and integrates the effective adjustment amounts by node and by device, clarifying the control responsibilities and control sequence of each associated node, thus forming a preliminary collaborative control scheme.

[0090] S207. After sending the target node's operating data, preliminary coordinated control scheme, and global distribution network topology parameters to the cloud platform, the cloud platform performs a full-network power flow simulation based on the current operating data of the entire network to obtain the predicted distribution of network parameters after the execution of the preliminary coordinated control scheme.

[0091] Specifically, the main coordinating node collects real-time operational data from target nodes, the generated preliminary coordinated control scheme, and global distribution network topology parameters. This data is then packaged and sent to the cloud platform via a secure communication channel. Upon receiving the data, the cloud platform retrieves the current operational data of the entire distribution network, including parameters such as voltage, current, power, and frequency for all nodes. Based on the global distribution network topology parameters, it constructs a complete power grid simulation model. The various adjustment commands from the preliminary coordinated control scheme are used as input conditions and substituted into the simulation model to execute a full-network power flow simulation. During the simulation, the cloud platform simulates the controllable devices of each associated node executing adjustment actions according to the scheme, calculates the changes in electrical parameters of each node in real time, and after the simulation, compiles the predicted distribution of parameters for the entire network after implementing the preliminary coordinated control scheme, clarifying the predicted values, trends, and potential risks of exceeding limits for each node's parameters.

[0092] S208. Based on the predicted distribution of parameters across the entire network, identify potential cascading risk nodes whose operating data exceeds the corresponding parameter threshold due to the execution of the preliminary collaborative control scheme, and include the potential cascading risk nodes in the edge collaborative node group.

[0093] Among them, potential cascading risk nodes refer to nodes identified through the network-wide parameter prediction distribution that, after the execution of the initial collaborative control scheme, will have their operating data exceed the corresponding parameter over-limit threshold. The anomaly of such nodes may trigger a chain reaction, causing more nodes to exceed the limit. Cascading risk refers to the chain risk that, after the parameters of a single or partial node are abnormal, other nodes will subsequently become abnormal.

[0094] Specifically, the cloud platform, based on the predicted distribution of parameters across the entire network, compares the predicted operating data of each node with the corresponding parameter exceedance thresholds, filtering out nodes whose predicted data exceeds the thresholds. These nodes are identified as potential cascading risk nodes that may arise from the execution of the initial collaborative control scheme. The cloud platform assesses the risk level of these potential cascading risk nodes, determining their risk level based on factors such as their electrical distance from the target node and the degree of parameter exceedance. The cloud platform sends the identification information, risk level, predicted exceedance parameters, and degree of exceedance of the potential cascading risk nodes to the main coordinating node. Upon receiving this information, the main coordinating node reviews the online status and controllable equipment configuration of the potential cascading risk nodes. After confirming that the nodes possess collaborative control capabilities, they are incorporated into the original edge collaborative node group, expanding the scope of collaborative control.

[0095] S209. Based on the results of the whole network power flow simulation, generate preventive constraint margins for each associated node in the expanded edge collaborative node group and distribute them to each associated node.

[0096] Among them, the preventive constraint margin refers to the safety constraint threshold set for each associated node in the extended edge collaborative node group based on the full network power flow simulation results. It is added to the local subproblem to limit the adjustment range of the equipment and avoid new parameter over-limitation caused by over-adjustment.

[0097] Specifically, based on the network-wide power flow simulation results, the cloud platform analyzes the parameter variation range and equipment adjustment limits of each associated node in the expanded edge collaborative node group. Combining this with the risk characteristics of potentially cascading risk nodes, it generates a corresponding preventative constraint margin for each associated node. This margin is set separately for different controllable devices and adjustment parameters of the node; for example, a maximum adjustment amplitude limit is set for voltage regulation devices, and an adjustment rate constraint is set for power regulation devices, ensuring that the device adjustment actions do not exceed safe limits. The cloud platform packages the preventative constraint margins of each associated node and sends them to the main coordinating node through a secure communication channel. The main coordinating node then forwards the margins to each associated node in the expanded edge collaborative node group. Upon receiving the margins, each associated node incorporates the preventative constraint margins as additional safety constraints into its own local subproblem construction process.

[0098] To further ensure the efficient and stable operation of distributed optimization algorithms in complex communication environments, this application employs a method involving steps S301 to S311. Specifically, as follows... Figure 3 The diagram shown is another flowchart illustrating an efficient edge-collaborative processing method for power big data in this application embodiment. S301. Each associated node in the edge collaborative node group sends probe data packets to each other to establish a communication link probe connection between node pairs.

[0099] Among them, the probe data packet refers to the dedicated data packet generated by the system for testing the quality of the communication link between nodes. It has a uniform size and format and does not contain actual business data. The communication link probe connection refers to the temporary communication connection established between each associated node for transmitting probe data packets, which is used to obtain communication quality parameters.

[0100] This step is performed after the edge collaborative node group is constructed and before iteratively solving the local subproblems. The scenario is that the system needs to obtain the communication status data between nodes in advance to provide a basis for adjusting the iteration step size coefficient in subsequent iterations.

[0101] Specifically, the system issues probe commands to all associated nodes within the edge collaboration node group, specifying parameters such as the size of the probe data packet, the sending frequency, and the sending duration. Each associated node, following the commands, sends probe data packets to all other associated nodes in the group one by one, recording the sending timestamp during the process. Upon receiving the probe data packet, the receiving associated node records the receiving timestamp and sends a reception confirmation message back to the sender. Through this process, a temporary communication link probe connection is established between each pair of associated nodes.

[0102] S302. Within a sliding time window of a preset window length, obtain the communication quality parameters between each pair of associated nodes, and sum the communication quality parameters according to the weight coefficient of the second dimension to obtain the communication quality score between each pair of associated nodes.

[0103] Among them, the preset window length refers to the duration of the time window set by the system for collecting and statistically analyzing communication quality parameters; the sliding time window refers to the time window that moves continuously at fixed time intervals for continuously collecting communication quality data; the second dimension weight coefficient refers to the weight value preset by the system for different communication quality parameters for weighted calculation of communication quality scores, which is set according to the degree of influence of each parameter on communication quality; the communication quality score refers to the numerical value that quantifies the quality of the communication link between node pairs, obtained by weighted summation of communication quality parameters, and the value range is usually 0-1, with higher values ​​indicating better communication quality.

[0104] This step is executed after the communication link detection connection is established between each pair of associated nodes. The scenario involves the system continuously collecting and quantifying communication quality data to provide a quantitative basis for adjusting the iteration step size coefficient. Specifically, the system starts a sliding time window of a preset window length. Within each time window, each associated node continuously records communication quality parameters with other nodes, including average communication latency, latency jitter standard deviation, packet loss rate, and available bandwidth. After each time window ends, each associated node reports the communication quality parameters within that window to the main coordinating node. The main coordinating node retrieves the preset second-dimensional weighting coefficient, which assigns corresponding weights to average communication latency, latency jitter standard deviation, packet loss rate, and available bandwidth. The main coordinating node performs a weighted summation of the communication quality parameters of each node pair according to the weighting coefficients. Each parameter is first standardized to eliminate dimensional differences before the weighted sum is calculated, ultimately yielding a unique communication quality score for each pair of associated nodes.

[0105] Optionally, the system can first have each associated node locally calculate the communication quality parameters for communicating with other nodes, calculate the communication quality score locally according to the weight coefficient of the second dimension, and then report the score to the main coordinating node. The main coordinating node verifies the reported score, removes outliers, and determines the final score.

[0106] S303. Obtain the standard deviation of delay jitter between each pair of associated nodes from the communication quality parameters, and adjust the preset benchmark step size by attenuation according to the standard deviation of delay jitter between each pair of associated nodes to obtain the initial iteration coefficient.

[0107] Among them, the reference step size refers to the basic coefficient preset by the system for adjusting the iteration step size, providing an initial step size reference for iterative solution; the attenuation adjustment refers to the adjustment method of reducing the reference step size proportionally according to the magnitude of the delay jitter standard deviation; the initial iteration coefficient refers to the iteration step size coefficient obtained after attenuation adjustment of the delay jitter standard deviation, which has not yet undergone gain adjustment for communication quality scoring.

[0108] Specifically, the system extracts the standard deviation of latency jitter between each associated node pair from the collected communication quality parameters. This parameter reflects the degree of fluctuation in communication latency; a larger value indicates a more unstable communication link. The system retrieves a preset baseline step size and establishes a mapping relationship between the standard deviation of latency jitter and the attenuation coefficient. The larger the standard deviation of latency jitter, the smaller the corresponding attenuation coefficient. According to this mapping relationship, the corresponding attenuation coefficient is determined for each node pair. The baseline step size is multiplied by the attenuation coefficient to obtain the initial iteration coefficient for each node pair. This achieves preliminary adjustment of the step size based on communication stability, ensuring that the iteration step size is appropriate when communication is unstable and avoiding algorithm oscillations.

[0109] Optionally, the system can first divide the standard deviation of latency jitter into multiple intervals, set a fixed attenuation coefficient for each interval, determine the attenuation coefficient according to the interval to which the standard deviation of latency jitter of each node pair belongs, and multiply it with the baseline step size to obtain the initial iteration coefficient. Optionally, the system can also establish a functional relationship between the standard deviation of latency jitter and the attenuation coefficient. Substituting the standard deviation of latency jitter for each node pair into the function, the corresponding attenuation coefficient is calculated. This coefficient is then multiplied by the baseline step size to obtain the initial iteration coefficient. The functional relationship can be dynamically optimized based on historical communication data and algorithm performance. It is understood that other methods can also be used to obtain the initial iteration coefficient, such as modifying the attenuation coefficient based on historical iteration results; this is not limited here.

[0110] In one specific embodiment, the formula for calculating the initial step size coefficient is: ,in, Let be the step size coefficient of the k-th cooperative group. This is the baseline step size (default value 0.1). Let $\frac{k}{k}$ be the standard deviation of the delay jitter in the k-th coordination group. This is the jitter sensitivity factor (default value 2.0). The communication quality score (range 0-1) for the kth collaborative group. This is the mass gain coefficient (default value 0.5).

[0111] S304. Adjust the gain of the initial iteration coefficients based on the communication quality scores between each pair of associated nodes to obtain the corrected iteration coefficients.

[0112] Among them, gain adjustment refers to the adjustment method of increasing the initial iteration coefficient proportionally according to the communication quality score; the corrected iteration coefficient refers to the iteration step size coefficient finally obtained after attenuation adjustment and gain adjustment, which is used to guide the iterative solution of local subproblems.

[0113] Specifically, the system obtains the communication quality scores between each pair of associated nodes and establishes a mapping relationship between the communication quality scores and gain coefficients. A higher communication quality score corresponds to a larger gain coefficient, and the corrected iterative coefficients are positively correlated with the communication quality scores. Based on this mapping relationship, a corresponding gain coefficient is determined for each pair of nodes. The initial iterative coefficients are multiplied by the gain coefficients to obtain the corrected iterative coefficients for each pair of nodes. For pair of nodes with high communication quality scores, the corrected iterative coefficients are larger, which accelerates the iterative solution; for pair of nodes with low communication quality scores, the corrected iterative coefficients are relatively smaller, ensuring the stability of the iterative solution.

[0114] Optionally, the system can divide the communication quality score into four levels: excellent, good, medium, and poor, and set a corresponding gain coefficient for each level. The gain coefficient is determined according to the score level of each node pair, and multiplied by the initial iteration coefficient to obtain the corrected iteration coefficient.

[0115] S305. During the iterative process of solving the local subproblems of each associated node based on the corrected iterative coefficients, monitor the changing trend of the original residuals between adjacent iteration steps.

[0116] Among them, the original residual refers to the degree of inconsistency between the local solutions of each associated node in adjacent iteration steps, which is a key indicator for measuring the convergence of iteration; the trend of change refers to the change of the original residual in consecutive iteration steps, including continuous decrease, stability, oscillation, increase, etc.

[0117] Specifically, after each iteration, the system collects the local solutions reported by each associated node for the current iteration step, compares them with the local solutions of the previous iteration, and calculates the original residuals between adjacent iteration steps. The system records the original residual values ​​for each iteration, forming a residual sequence. By analyzing the residual sequence, the system determines the trend of the original residuals. For example, if the residuals continuously decrease over multiple iterations, it indicates positive convergence; if the residuals fluctuate in different iteration steps, it indicates oscillation; if the residual values ​​remain basically unchanged or increase, it indicates slow convergence or divergence.

[0118] Optionally, the system can collect the local solutions of each associated node by the main coordinating node, calculate and record the original residuals, and intuitively judge the trend of residual change by drawing residual change curves. At the same time, trend judgment rules can be set, such as considering a continuous decrease in residuals for 3 consecutive rounds as a continuous decrease, and considering an oscillation if the residual fluctuation amplitude exceeds a preset value for 3 adjacent rounds.

[0119] S306. If the original residual continues to decrease in a preset number of iterations, then increase the corrected iteration coefficient.

[0120] Among them, the continuous preset number of iterations refers to the minimum number of iterations that the system pre-sets to determine whether the residual continues to decrease. This number is set according to the characteristics of the algorithm and the actual application scenario. Increasing the corrected iteration coefficient means increasing the coefficient value by a preset ratio or a fixed magnitude on the basis of the original corrected iteration coefficient in order to accelerate the iteration convergence speed.

[0121] Specifically, the system continuously monitors the changing trend of the original residual during the iteration process. When it detects that the original residual shows a decreasing trend in consecutive preset number of iterations (e.g., 3 or 5 times), and the decrease meets the preset requirements, it determines that the current iteration convergence state is good. The system increases the corrected iteration coefficients by a preset increase ratio (e.g., 10% or 20%) or a fixed amount to obtain new corrected iteration coefficients. The new coefficients are distributed to each associated node, and each node uses the new coefficients to solve local subproblems in subsequent iterations, accelerating the iteration convergence speed and shortening the overall solution time.

[0122] Optionally, the system can also dynamically adjust the increase ratio according to the decrease in residuals. The greater the decrease in residuals, the higher the increase ratio. For example, when the residuals decrease by more than 30%, the increase ratio is 20%, and when the decrease is 10%-30%, the increase ratio is 10%, ensuring that the coefficient adjustment is compatible with the convergence speed.

[0123] S307. If the original residual oscillates, reduce the corrected iteration coefficient.

[0124] Among them, oscillation refers to the irregular up-and-down fluctuation of the original residual in adjacent iteration steps, without a continuous downward or upward trend, which is used to indicate that the consistency of the local solution is in an unstable state during the iterative solution process; reducing the modified iteration coefficient refers to reducing the coefficient value according to a preset rule on the basis of the existing modified iteration coefficient, which is an adjustment method to suppress algorithm oscillation and improve the stability of iterative solution.

[0125] Specifically, during the iteration process, the system continuously tracks the numerical changes of the original residuals and forms a continuous residual sequence. When analysis reveals that the residual sequence has no fixed pattern of change, and the residual values ​​of adjacent iteration steps show significant fluctuations, exceeding the system's preset oscillation threshold, the system determines that the original residuals are oscillating. This state indicates that the current corrected iteration coefficients are mismatched with the communication status and solution progress between nodes. Excessively large coefficients lead to excessively high local solution update amplitudes, causing instability in the solution process. At this point, the system initiates a coefficient reduction process, adjusting the corrected iteration coefficients according to preset adjustment rules. The reduced coefficients decrease the update amplitude of local solutions in each iteration, gradually stabilizing the iterative solution process, suppressing residual oscillations, and driving the algorithm towards convergence. Simultaneously, the system sends the adjusted corrected iteration coefficients to all associated nodes in the edge collaborative node group in real time, and each node immediately adopts the new coefficients to carry out subsequent local subproblem solving work.

[0126] Optionally, the system can first preset a fixed coefficient reduction ratio. When residual oscillation is detected, the current corrected iteration coefficient is directly reduced according to this ratio. Then, the reasonableness of the reduced coefficient is checked to ensure that the coefficient is not lower than the minimum iteration coefficient threshold preset by the system. After the check is passed, the new coefficient is sent to each associated node.

[0127] S308. After each iteration step is completed, calculate the change in local solution of each associated node between the current iteration step and the previous iteration step.

[0128] In this context, an iteration step refers to the complete solution cycle in a distributed collaborative optimization algorithm, in which each associated node completes a local subproblem solution, performs a data interaction between nodes, and updates the global coordination variable. The change in local solution refers to the numerical difference between the local solution obtained by the same associated node in the current iteration step and the local solution obtained in the previous iteration step, which is used to characterize the update magnitude and trend of the local solution of a single node.

[0129] Specifically, after each iteration step, the main coordinating node collects the local solutions for the current iteration step reported by all associated nodes in the edge collaborative node group, and retrieves the historical local solutions for the previous iteration step from local storage. For each associated node, the system calculates the numerical difference between the local solution of the current iteration step and the local solution of the previous iteration step, dimension by dimension, based on the parameter type and dimension of the local solution. Then, normalization is used to eliminate the dimensional differences between different parameter dimensions, resulting in a unique local solution change for each associated node. For local solutions with multiple parameters, the system also calculates a weighted average of the differences in each dimension as the overall local solution change for that node, comprehensively reflecting the update status of the node's local solutions. The main coordinating node records and stores the local solution changes of all nodes, forming a sequence of node solution change values, providing continuous data support for subsequent iteration state judgments.

[0130] S309. If there is a target associated node whose local solution change is lower than the preset marginal contribution threshold in a series of preset iteration steps, then the iteration participation state of the target associated node is switched from active state to locked state.

[0131] Among them, the marginal contribution threshold refers to the system's preset critical value for the change in local solution, which is used to determine whether the update of the local solution of the associated node makes an effective contribution to the convergence of the global optimal solution. If it is lower than the threshold, it means that the local solution of the node has basically stabilized. The target associated node refers to the associated node whose change in local solution is lower than the marginal contribution threshold in a consecutive preset number of iterations. The active state refers to the state in which the associated node participates in the iterative solution normally, completes the solution of the local subproblem, uploads the local solution and receives the update of the global coordination variable. The locked state refers to the state in which the associated node suspends participation in the core process of iterative solution, no longer updates the local solution, does not need to upload solution data, and does not participate in data interaction between nodes.

[0132] Specifically, after calculating the local solution change in each iteration step, the system analyzes the solution change sequence of each associated node, checking one by one whether any node has a local solution change that is consistently below a preset marginal contribution threshold for a preset number of consecutive iteration steps. If such a node exists, the system marks it as a target associated node, determining that its local solution has essentially converged to a stable value. Continuing to participate in the iteration would contribute almost nothing to the convergence of the global optimal solution, and would instead consume its own computing power and inter-node communication resources. At this point, the system sends a state switching command to the target associated node, officially switching its iteration participation state from active to locked. Simultaneously, the main coordinating node updates the node state ledger of the edge collaborative node group. In subsequent iterations, the system no longer waits for the locked target associated node to send back local solution updates, but still broadcasts the updated data of the global coordination variables to it so that it can monitor its own coupling constraint residuals.

[0133] S310. Receive the broadcast update of the global coordination variable and calculate the coupling constraint residuals corresponding to the target associated node based on the received global coordination variable.

[0134] Among them, the global coordination variable refers to the global variable used to coordinate the local solutions of each associated node in the edge collaborative node group so that the local solutions of each node satisfy the coupling constraints between nodes. It is calculated and broadcast by the main coordination node based on the local solutions of all active nodes. The coupling constraint residual refers to the deviation value between the local solution of the target associated node and the global coupling constraint, calculated based on the latest global coordination variable when the target associated node is in the locked state. It is used to characterize the degree of matching between the locked local solution and the current global solution state.

[0135] Specifically, after the target associated node enters the locked state, it still retains the ability to receive broadcast information from the main coordinating node. When the main coordinating node updates the global coordination variables based on the latest local solutions of the active nodes in the edge coordination node group and broadcasts this update to all nodes in the group, the target associated node will receive the broadcast update data of the global coordination variables in real time. After receiving the update, the target associated node uses its currently locked local solution as a basis, substitutes it into the inter-node coupling constraints of the distribution network, and calculates the deviation between its local solution and the coupling constraints in conjunction with the latest global coordination variables. This deviation value is the coupling constraint residual corresponding to the target associated node. The larger the value of the coupling constraint residual, the greater the deviation between the local solution locked by the target node and the current global solution state, indicating that it can no longer meet the coupling constraint requirements between nodes; the smaller the value, the more the locked local solution still matches the global solution state. The target associated node will record the calculated coupling constraint residual locally and report some key residual data to the main coordinating node as needed.

[0136] S311. If the coupling constraint residual exceeds the preset recovery threshold, the iteration participation state of the target associated node will be restored from the locked state to the active state, and the node will rejoin the iteration solution with the currently locked local solution as the initial value.

[0137] Among them, the recovery threshold refers to the system's preset coupling constraint residual critical value, which is used to determine whether the target associated node in the locked state needs to be restored to the active state. If the threshold is exceeded, it means that the locked local solution can no longer meet the global coupling constraint requirements. The initial value refers to the initial local solution value used when a new round of local subproblem solving is carried out after the target associated node is restored to the active state. It is directly used by the currently locked local solution.

[0138] Specifically, after the target associated node completes the calculation of the coupling constraint residuals, the calculation result is compared with the system's preset recovery threshold. If the coupling constraint residuals exceed the recovery threshold, the system determines that the local solution locked by the node is no longer suitable for the current global solution state. Continuing to lock it will destroy the coupling constraint relationship between nodes and affect the solution effect of the global optimization model. At this time, the system sends a state recovery command to the target associated node, officially restoring its iteration participation state from the locked state to the active state. At the same time, the main coordinating node updates the node state ledger of the edge collaborative node group, adding the node back to the active node list. After the target associated node is restored to the active state, it no longer reinitializes the local solution. Instead, it directly uses the currently locked local solution as the initial value, combines the latest global coordination variables and the current corrected iteration coefficients, and immediately starts a new round of local subproblem solving work, allowing the local solution to quickly catch up with the progress of the global solution and once again make an effective contribution to the convergence of the global optimal solution.

[0139] To facilitate understanding, the method of this application is illustrated below through a specific application scenario. Assume a distribution network area is equipped with several edge computing nodes and intelligent monitoring terminals, involving distributed photovoltaic (PV) installations. During peak PV power generation periods, the system detects that the voltage deviation rate of a certain feeder terminal node reaches +9.2%, exceeding the general over-limit threshold (i.e., parameter over-limit threshold) of +7%, with an over-limit deviation rate of approximately 31.4%. The system determines that this over-limit deviation rate reaches the collaborative control trigger level. Based on the sensitivity matrix and topology, the system selects eight associated nodes with electrical coupling influence with the target node, constructing an edge collaborative node group. By detecting the communication quality between each node and calculating the corrected iteration step size coefficient, the system performs distributed iterative solutions. After several rounds of iteration, the optimal collaborative control solution is obtained. After regulation, the maximum voltage deviation rate of the node is reduced to +2.1%, and all over-limit nodes return to normal operation.

[0140] The power big data edge collaborative high-efficiency processing system of this invention is applied to electronic devices. Figure 4 A schematic diagram of the architecture of an electronic device suitable for implementing embodiments of the present invention is shown.

[0141] It should be noted that, Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0142] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions (computer programs), or by instructions (computer programs) controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of this embodiment includes a storage medium and a processor, wherein the storage medium stores multiple instructions that can be loaded by the processor to execute any step of the method provided in the embodiments of the present invention.

[0143] Specifically, the storage medium and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more signal lines. The storage medium stores computer-executable instructions that implement data access control methods, including at least one software functional module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software program and module stored in the storage medium. The storage medium can be, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The storage medium stores the program, and the processor executes the program after receiving the execution instructions.

[0144] Furthermore, the software programs and modules within the aforementioned storage medium may also include an operating system, which may include various software components and / or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.) and can communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip with signal processing capabilities. The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc., which can implement or execute the methods, steps, and logic flowcharts disclosed in this embodiment. The general-purpose processor may be a microprocessor or any conventional processor.

[0145] Since the instructions stored in the storage medium can execute the steps in any of the methods provided in the embodiments of the present invention, the beneficial effects of any of the methods provided in the embodiments of the present invention can be achieved, as detailed in the preceding embodiments, and will not be repeated here.

[0146] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for efficient edge collaborative processing of power big data, characterized in that, The method includes: During the execution of the initial task scheduling scheme, the operating data of the distribution network edge nodes are monitored in real time to determine the target nodes whose operating data exceeds the preset parameter limit threshold, and the corresponding limit deviation rate of the target nodes. If the detected out-of-limit deviation rate meets the preset collaborative control trigger level, then based on the preset multi-parameter sensitivity matrix and distribution network topology, the associated nodes that have electrical coupling influence on the target node are selected from all distribution network edge nodes to construct an edge collaborative node group. A global optimization model is constructed with the goal of restoring the normal operation of the target node, and the global optimization model is decomposed into local sub-problems corresponding to each associated node in the edge collaborative node group based on a distributed collaborative optimization algorithm; The iteration step size coefficient is adjusted according to the communication quality score between each associated node in the edge collaborative node group to obtain the corrected iteration coefficient. The corrected iteration coefficient is positively correlated with the communication quality score. The communication quality score is obtained by weighted calculation based on the multi-dimensional communication quality parameters of the communication link between associated nodes. The local subproblems of each associated node in the edge collaborative node group are solved based on the corrected iteration coefficients, and the optimal collaborative control solution of each associated node is obtained through data interaction between nodes. Based on the cooperative optimal control solution, control commands are generated for each controllable device in the edge cooperative node group and then issued for execution.

2. The method according to claim 1, characterized in that, Before the step of real-time monitoring of the operating data of the distribution network edge nodes during the execution of the initial task scheduling scheme, determining the target node whose operating data exceeds a preset parameter limit threshold, and the corresponding limit deviation rate of the target node, the method further includes: Measure the performance parameters corresponding to each distribution network edge node. The performance parameters include computing power indicators, storage performance indicators, communication capability indicators, and reliability indicators calculated based on historical failure rate and historical task success rate. The performance parameters are weighted and summed according to the preset first dimension weighting coefficient to obtain the comprehensive capability index corresponding to each distribution network edge node; Obtain the task feature vector corresponding to each task in the task request queue to be processed. The task feature vector represents the latency sensitivity, computational complexity and spatial correlation. Based on the task feature vector and the comprehensive capability index, hierarchical matching between tasks and distribution network edge nodes is performed to obtain an initial task scheduling scheme.

3. The method according to claim 1, characterized in that, After the step of real-time monitoring of the operating data of the distribution network edge nodes during the execution of the initial task scheduling scheme, determining the target node whose operating data exceeds a preset parameter limit threshold, and the corresponding limit deviation rate of the target node, the method further includes: Based on the deviation rate of the target node's operating data relative to the parameter limit threshold, the current alarm level corresponding to the target node is determined. The current alarm level includes a warning level, a coordinated control trigger level, and a severe limit violation level. In response to the warning level, local controllable devices within the jurisdiction of the target node are invoked to perform local control. In response to the aforementioned severe limit violation level, the joint cloud platform makes global collaborative decisions on the edge collaborative node group.

4. The method according to claim 1, characterized in that, Before the step of adjusting the iteration step size coefficient based on the communication quality score between each associated node in the edge collaborative node group to obtain the corrected iteration coefficient, the method further includes: Each associated node in the edge collaborative node group sends probe data packets to each other to establish a communication link probe connection between node pairs; Within a sliding time window of a preset window length, the communication quality parameters between each pair of associated nodes are obtained. The communication quality parameters include average communication latency, latency jitter standard deviation, packet loss rate, and available bandwidth. The communication quality parameters are weighted and summed according to the preset second-dimensional weighting coefficients to obtain the communication quality score between each pair of associated nodes.

5. The method according to claim 4, characterized in that, The step of adjusting the iteration step size coefficient based on the communication quality score between each associated node in the edge collaborative node group to obtain the corrected iteration coefficient specifically includes: Obtain the standard deviation of delay jitter between each pair of associated nodes from the communication quality parameters; The preset baseline step size is attenuated and adjusted based on the standard deviation of the time delay jitter between each pair of associated nodes to obtain the initial iteration coefficient, which is negatively correlated with the standard deviation of the time delay jitter. The initial iteration coefficients are adjusted by gain based on the communication quality scores between each pair of associated nodes to obtain the corrected iteration coefficients. During the iterative process of solving the local subproblems of each associated node based on the corrected iterative coefficients, the changing trend of the original residuals between adjacent iteration steps is monitored. If the original residual continues to decrease in a predetermined number of iterations, then the corrected iteration coefficient is increased. If the original residual oscillates, then the corrected iteration coefficient is reduced.

6. The method according to claim 3, characterized in that, The steps for the joint cloud platform to make global collaborative decisions on the edge collaborative node group in response to the severe limit violation level specifically include: In response to the severe limit violation level, based on the multi-parameter sensitivity matrix locally cached by each associated node in the edge collaborative node group and the adjustment capability of the local controllable device corresponding to each associated node, the preliminary adjustment amount of each associated node to the target node is calculated, and the preliminary collaborative control scheme is obtained by summarizing. After sending the target node's operating data, the preliminary coordinated control scheme, and the global topology parameters of the distribution network to the cloud platform, the cloud platform performs a full-network power flow simulation on the preliminary coordinated control scheme based on the current operating data of the entire network, and obtains the predicted distribution of the network parameters after the execution of the preliminary coordinated control scheme. Based on the predicted distribution of parameters across the entire network, potential cascading risk nodes that generate operating data exceeding the corresponding parameter threshold due to the execution of the preliminary collaborative control scheme are identified, and these potential cascading risk nodes are included in the edge collaborative node group. Based on the results of the full network power flow simulation, preventive constraint margins are generated for each associated node in the extended edge collaborative node group and distributed to each associated node. The preventive constraint margins are used as additional safety constraints in the local subproblems corresponding to each associated node when constructing the global optimization model.

7. The method according to claim 1, characterized in that, The step of solving the local subproblems of each associated node in the edge collaborative node group based on the corrected iteration coefficients, and iterating through data interaction between nodes to obtain the collaborative optimal control solution of each associated node, specifically includes: After each iteration step is completed, the change in local solution of each associated node between the current iteration step and the previous iteration step is calculated; If there is a target associated node whose local solution change is lower than a preset marginal contribution threshold in a consecutive preset number of iteration steps, then the iteration participation state of the target associated node is switched from active state to locked state. In the locked state, the node no longer waits for the target associated node to send back local solution updates. Receive broadcast updates of global coordination variables, and calculate the coupling constraint residuals corresponding to the target associated node based on the received global coordination variables; If the coupling constraint residual exceeds the preset recovery threshold, the iteration participation state of the target associated node is restored from the locked state to the active state, and it re-participates in the iterative solution with the currently locked local solution as the initial value.

8. A high-efficiency edge collaborative processing system for power big data, characterized in that, The system includes: one or more processors and memory; The memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is run on the power big data edge collaborative high-efficiency processing system, the system performs the method as described in any one of claims 1-7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is run on the power big data edge collaborative high-efficiency processing system, the system performs the method as described in any one of claims 1-7.

Images