A distributed resource real-time access optimization method based on edge computing

Abstract

This invention discloses a method for real-time access optimization of distributed resources based on edge computing, comprising: constructing a distributed sensing network and sensing datasets for each node; calculating operational stability indices for each resource to be accessed; constructing an objective function by combining unit power access cost, planned access power, access time window parameters, and carbon trading cost factors, and calculating global and local objective function values ​​for each node; calculating a consistency index based on the global objective function value, according to the local objective function values ​​of each node and data heterogeneity, and determining whether to initiate a collaborative optimization mechanism to adjust the optimization results of each node; distributing the collaboratively optimized access scheme to each node, recalculating the consistency index based on node feedback, and re-optimizing if the consistency index exceeds a threshold. This method can comprehensively consider operational stability, economic cost, and low-carbon goals before distributed resource access, and can effectively coordinate decision-making differences among edge nodes to generate a globally optimal access scheme.

Description

Technical Field

[0001] This invention relates to the field of distributed resource technology, and in particular to a method for optimizing real-time access to distributed resources based on edge computing. Background Technology

[0002] With the construction of new power systems, the large-scale integration of distributed resources such as photovoltaics and energy storage has become an inevitable trend. However, distributed resources are characterized by their diverse types, strong output fluctuations, and geographical dispersion, posing significant challenges to the stable, economical, and low-carbon operation of the power grid.

[0003] Existing resource access or scheduling optimization schemes mainly suffer from the following shortcomings: a) Most schemes focus on power scheduling after distributed resource access, lacking an effective mechanism for multi-dimensional quantitative evaluation and collaborative optimization of resources before access; b) The evaluation dimensions are singular, often only considering economic efficiency or stability, failing to internalize carbon costs under the "dual carbon" objective as a decision factor; c) In a distributed edge computing architecture, due to data heterogeneity (differences in acquisition accuracy and latency), conflicts between local optimization and global objectives are easily generated among nodes, and existing technologies lack quantitative collaborative consistency judgment and adaptive adjustment mechanisms; d) Insufficient quantification of key characteristics such as the elastic adjustment potential of resources and equipment health status leads to inaccurate access decisions.

[0004] Therefore, there is an urgent need for a solution that can comprehensively consider operational stability, economic costs, and low-carbon goals before distributed resource access, and can effectively coordinate decision-making differences among edge nodes to obtain the globally optimal access solution. Summary of the Invention

[0005] To address the shortcomings of existing technologies in the background section, this invention provides a method for optimizing real-time access to distributed resources based on edge computing. This method can comprehensively consider operational stability, economic costs, and low-carbon goals before accessing distributed resources, and can effectively coordinate decision-making differences among edge nodes to generate a globally optimal access scheme.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] This invention provides a method for optimizing real-time access to distributed resources based on edge computing, comprising:

[0008] S1: Construct a distributed sensing network based on the location information and topological relationship of edge nodes, obtain the real-time active power parameters of each distributed resource to be accessed and perform preprocessing, and construct the sensing dataset corresponding to each edge node.

[0009] S2: Calculate the operational stability index of each distributed resource to be accessed based on the data in the perception dataset;

[0010] S3: Using operational stability indicators, unit power access cost, planned access power, access time window parameters, and carbon trading cost factors, an objective function is constructed, and then the global objective function value of the distributed sensing network and the local objective function value of each edge node are calculated; among them, the optimization scheme in the collaborative optimization mechanism takes minimizing the global objective function value as the optimization objective of the optimization algorithm, and iteratively optimizes to generate the optimization results of the planned access power value of each resource within the jurisdiction of each edge node.

[0011] S4: Calculate the consistency index based on the local objective function value and data heterogeneity of each edge node, and determine whether the collaborative optimization mechanism needs to be activated to adjust the optimization results of each edge node based on the consistency index;

[0012] S5: Distribute the optimized access scheme to each edge node, recalculate the consistency index based on the node feedback, and return to S4 if the consistency index exceeds the threshold.

[0013] Furthermore, the specific process of S2 is as follows:

[0014] S21: For the first A distributed resource to be connected, from Each edge node acquires its corresponding perceptual dataset; among which... This represents the total number of edge nodes participating in data collection and computation in the current scenario.

[0015] S2.2: For each node Computing distributed resources Normalized value of power fluctuation ;

[0016]

[0017] in, For nodes Distributed resources Real-time power; For nodes Distributed resources Average power; For nodes Distributed resources Minimum power; For nodes Distributed resources Maximum power;

[0018] S2.3: Weights based on each edge node ,right The power fluctuation normalized values ​​of each node are weighted and fused to calculate the distributed resource. Comprehensive power fluctuation value ;

[0019] S2.4: Based on distributed resources Comprehensive power fluctuation value Combined with the first Device health parameters of a distributed resource and elasticity index Calculate and obtain the distributed resources to be accessed operational stability indicators .

[0020] Furthermore, the weights of each edge node The calculation formula is

[0021]

[0022] in, For edge nodes The shortest path length to each of the other nodes.

[0023] Furthermore, the distributed resources to be accessed operational stability indicators The calculation formula is:

[0024] ;

[0025] ;

[0026] in, and All are weighting coefficients, and ; Adjustable power for distributed resources; Rated power for distributed resources; This refers to the load response sensitivity parameter.

[0027] Furthermore, the specific process of S3 is as follows:

[0028] S3.1: Obtain distributed resources in the target region operational stability indicators Unit power access cost Planned access power Access time window parameters Target weight coefficient Carbon price coefficient ;

[0029] S3.2: For each distributed resource in the target region Based on its projected carbon emissions and industry benchmark emissions The corresponding carbon trading cost factor is calculated. :

[0030] S3.3: Based on target weight coefficients Unit power access cost Planned access power Access time window parameters and the corresponding carbon trading cost factor Construct the economic cost in the objective function ;

[0031] S3.4: Based on target weight coefficients and the operational stability indicators of each distributed resource Constructing the stability benefit in the objective function :

[0032] S3.5: Overall stability benefits Economic cost The final objective function is obtained. ; where the objective function value The calculation formula is:

[0033] .

[0034] Furthermore, carbon trading cost factors The calculation formula is:

[0035] .

[0036] Furthermore, the economic cost in the objective function The calculation formula is:

[0037] ;

[0038] Stability gains in the global objective function :

[0039] ;

[0040] in, This represents the total number of distributed resources in the target region.

[0041] Furthermore, the specific process of S4 is as follows:

[0042] S4.1: Get edge nodes The optimized local objective function value obtained based on local computation ,in, , This represents the total number of edge nodes.

[0043] S4.2: Calculate any two edge nodes and Data heterogeneity between ;

[0044] S4.3: Based on the local objective function values ​​corresponding to any two edge nodes , and the corresponding data heterogeneity Calculate the network-wide consistency index ;

[0045] S4.4: The consistency index Consistency threshold with preset If a comparison is made, If the condition is met, it is determined that the collaborative optimization mechanism needs to be activated; otherwise, it is determined that it does not need to be activated.

[0046] S4.5: If it is determined that a collaborative optimization mechanism needs to be activated, then each edge node should move to its neighbor nodes. The nodes in the process exchange their respective optimization schemes. ;in, For one A dimensional vector, whose corresponding vector components are for nodes The planned values ​​of each distributed resource under its jurisdiction;

[0047] S4.6: Each edge node calculates the adjustment amount of its own optimization scheme based on the optimization schemes received from its neighboring nodes. And update the local optimization scheme based on the adjustment amount; The formula for calculating the adjustment amount is as follows:

[0048] ;

[0049]

[0050] in, For coordinated step size; For nodes The set of neighboring nodes; Neighboring nodes Optimization scheme; For nodes With nodes Communication quality factor between them; For edge nodes with neighboring nodes Local communication timeliness factor between them; The decay rate; For nodes With nodes Last successful communication time; This is the current time of successful communication;

[0051] S4.7: Based on the updated optimization scheme, repeat S4.1 to S4.4, recalculate the consistency index and make judgments until a new consistency index is reached. satisfy Or it may reach the preset maximum number of iterations.

[0052] Furthermore, edge nodes and Data heterogeneity between The calculation formula is as follows:

[0053]

[0054] in, For edge nodes For the first Preprocessing delay of class-aware parameters; For edge nodes For the first Preprocessing delay of class-aware parameters; The number of types of sensing parameters ( ).

[0055] Furthermore, network-wide consistency indicators The calculation formula is as follows:

[0056] .

[0057] This invention proposes a distributed resource real-time access optimization method based on edge computing. The method, encompassing the entire process of distributed resource access to a power system, constructs a multi-dimensional, collaborative optimization decision-making approach: In resource characteristic quantification and evaluation, edge node weights are calculated to accurately reflect node importance through comprehensive network topology, providing a scientific basis for data fusion; the resource elasticity index considers adjustment potential and response characteristics, accurately distinguishing the elasticity differences of different resources to aid in selection and scheduling; the operational stability index integrates power fluctuations and equipment characteristics, evaluating from both real-time stability and long-term potential dimensions to screen suitable resources and reduce system fluctuation risks; the objective function construction achieves multi-objective collaboration, flexibly adapting to complex scenarios such as grid peak and user off-peak periods using weight coefficients, accurately quantifying the entire lifecycle cost, covering implicit factors such as carbon costs, avoiding decision bias, and combining intelligent algorithms for automatic optimization, improving decision-making efficiency and accuracy, while also possessing flexible scalability to adapt to future technology policy iterations; consistency indicators and collaborative optimization mechanisms consider data heterogeneity and node collaboration, accurately identifying discrepancies, triggering adaptive adjustments, ensuring system decision consistency and stability, and improving edge computing collaborative efficiency. From resource assessment and decision optimization to coordinated regulation, the goal of achieving stability, economy, and low carbon integration is realized, providing scientific, intelligent, and sustainable technical support for the construction of new power systems, promoting the upgrade of power systems from experience-driven to data-driven, and helping to efficiently absorb clean energy and ensure the safe and economical operation of the power grid. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart of a distributed resource real-time access optimization method based on edge computing provided in an embodiment of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0061] Example 1

[0062] This embodiment achieves efficient and adaptive access management of distributed resources through a complete closed-loop process, from resource information collection to dynamic optimization and adjustment, based on edge nodes.

[0063] like Figure 1 As shown, this embodiment provides a method for optimizing real-time access to distributed resources based on edge computing, including:

[0064] S1: Construct a distributed sensing network based on the location information and topological relationship of edge nodes, collect sensing parameters of each distributed resource to be accessed and preprocess them to construct the sensing dataset corresponding to each edge node.

[0065] Specifically, a distributed sensing network is first constructed using the location information and topological relationships of each edge node. This network enables the collection of real-time operational status sensing parameters of each distributed resource to be accessed at close range with low latency. This allows for the acquisition of sensing parameters for each distributed resource within the jurisdiction of each edge node, ultimately constructing a sensing dataset corresponding to each edge node. Taking an edge node as an example, its corresponding perception dataset is denoted as . .

[0066] More specifically, perceptual datasets The active power-related parameters refer to the sensing parameters obtained by edge nodes through hardware acquisition and software processing, which digitize the active power and related parameters of distributed resources (such as photovoltaics, energy storage, and loads) during operation. These sensing parameters include power parameters, equipment operating status parameters, and basic grid operating parameters.

[0067] Power phase parameters include, but are not limited to, real-time active power, average power, and historical power extremes of distributed resources; equipment operating status parameters include, but are not limited to, equipment health and the adjustable power range corresponding to the elasticity index; and basic grid operating parameters include, but are not limited to, node voltage and line current.

[0068] In practice, edge nodes use sensors such as current transformers and voltage transformers to collect electrical signals such as current and voltage during the operation of distributed resources in real time. These analog signals are then converted into digital signals by analog-to-digital converter chips, and finally processed using built-in algorithms (such as those based on...). The power calculation model calculates the raw value of active power. Based on this, the edge nodes perform filtering and noise reduction (eliminating fluctuations and glitches caused by electromagnetic interference), error correction (compensating for hardware deviations by combining sensor calibration coefficients), and per-unit processing (unifying the dimensions based on rated power) on the raw power data, ultimately forming sensing parameters that can be directly used for system analysis. These sensing parameters are the basic data for calculating the operational stability index of computing resources.

[0069] S2: Calculate the operational stability index of each distributed resource to be accessed based on the data in the perception dataset.

[0070] S2.1: For the first A distributed resource to be connected, from Each edge node acquires its corresponding perceptual dataset; among which... This represents the total number of edge nodes participating in data acquisition and computation in the current scenario; the sensing dataset includes, but is not limited to, each node. Distributed resources Real-time power Average power Minimum power Minimum sensing parameter .

[0071] Real-time power Within the current or preset sampling period, the node Collected distributed resources The active power; Average power, i.e., node power Within a preset historical time window, for distributed resources The average value of the collected active power; For distributed resources within the pre-time window The minimum value of the active power collected; For distributed resources within a preset time window The maximum value of the active power collected.

[0072] S2.2: For each node Computing distributed resources Normalized value of power fluctuation :

[0073]

[0074] S2.3: Weights based on each edge node ,right The power fluctuation normalized values ​​of each node are weighted and fused to calculate the distributed resource. Comprehensive power fluctuation value :

[0075]

[0076] The weights of the edge nodes are determined based on their topological importance in the distributed sensing network, and the calculation formula is as follows:

[0077]

[0078] in, For edge nodes The shortest path length to each of the other nodes.

[0079] Calculating weights based on the shortest path between nodes aligns with the topological characteristics of edge computing networks, accurately reflecting the status of nodes in actual communication and data interaction networks. For example, in distributed resource access scenarios, nodes with good connectivity and located in key network positions have higher data collection value for system decision-making; this method assigns them reasonable high weights, improving the scientific utilization of data. Without relying on centralized configuration, it calculates weights solely based on network topology relationships (path length). In multi-node distributed architectures, nodes do not require additional negotiation and automatically obtain weights based on their own topological contributions, ensuring computational fairness and adapting to the decentralized needs of edge computing.

[0080] It provides basic node weights for upper-level calculations such as operational stability indicators, enabling subsequent analyses involving multi-node data fusion (such as resource stability assessment) to better align with network realities, improve the system's accuracy in judging resource access and operational status, and assist in the optimized management and stable operation of distributed resources in scenarios such as power systems.

[0081] S2.4: Based on distributed resources Comprehensive power fluctuation value Combined with the first Device health parameters of a distributed resource and elasticity index Calculate and obtain the distributed resources to be accessed operational stability indicators .

[0082] More specifically, equipment health parameters It combines distributed resources Real-time operational data and historical status records, based on edge nodes Local sensing networks collect distributed resources Real-time operational data (such as equipment temperature, component runtime, and fault alarm records); retrieve edge node data. The historical health data of the resource stored locally (such as the number of historical failures, maintenance cycle, and performance degradation trend) is used to obtain the equipment health parameters through a preset health assessment model.

[0083] The detailed steps for establishing this health assessment model can be divided into four stages:

[0084] The first stage is data preprocessing. Real-time operational data of distributed resources is collected based on the local sensing network of edge nodes, including equipment temperature, component runtime, and fault alarm records. Simultaneously, historical health data of the resource stored locally on the nodes is retrieved, covering historical fault counts, maintenance cycles, and performance degradation trends. Both types of data are cleaned, outliers are removed, the data format is standardized, records from different time dimensions are standardized to the same time granularity, and missing data is interpolated to complete the data, ultimately forming a structured health data sample set.

[0085] Next, the feature engineering phase is initiated, extracting key health features from the preprocessed dataset. Real-time equipment temperature is converted into a temperature risk coefficient, component runtime is converted into a fatigue wear index, and historical failure counts and maintenance frequency are integrated into a cumulative failure weight. Simultaneously, the performance degradation rate is fitted using performance decay trends. These features are then normalized, scaling them to the 0-1 range to eliminate dimensional differences between features.

[0086] The next stage is model building and training. A suitable model for lightweight computation at edge nodes is selected, such as a weighted fusion model or a lightweight decision tree model. With device health parameters as the output target, the extracted normalized features are used as input, combined with the quantified values ​​of device health status marked in actual operation and maintenance. Model training is then completed locally at the edge node. During the process, the weight coefficients of each feature are adjusted to ensure that the fitting error between the model output and the actual health status is below a preset threshold.

[0087] Finally, in the model validation and deployment phase, health data samples from devices not involved in training are selected for validation to test the accuracy and stability of the model's output health parameters under different operating conditions. If the validation passes, the model is deployed to the local computing module of the edge node, and a periodic update mechanism is set up so that when the resource's operating data accumulates to a certain scale, the model is retrained to adapt to long-term changes in device performance.

[0088] The resilience index measures the resilience of distributed resources in terms of power regulation, reflecting their response potential to system changes (such as load fluctuations), and is used in the calculation of higher-level indicators such as operational stability. The calculation formula is:

[0089]

[0090] in, Adjustable power for distributed resources; Rated power for distributed resources; This refers to the load response sensitivity parameter.

[0091] Based on the adjustable power of distributed resources (i.e., the difference between the upper and lower limits of power that distributed resources can flexibly adjust, reflecting the adjustment potential), and the allocated power. (That is, the ratio of the rated output of a distributed resource to its power consumption), the ratio of the two This reflects the "relative adjustment capability" (e.g., a large rated power but a small adjustable range indicates weak relative adjustment capability). The load response sensitivity parameter is then introduced. (Related to the response characteristics of distributed resources to load changes, different resource types and equipment characteristics correspond to different...) By combining relative adjustment capability with response characteristics, the resource elasticity index is finally obtained. This quantifies the overall performance of resources in power elasticity adjustment.

[0092] By linking the difference between the adjustable power limits, rated power, and load response characteristics of distributed resources, resource elasticity can be precisely quantified from two dimensions: "adjustment potential" and "response sensitivity," thus overcoming the limitations of measuring only power range. For example, comparing different types of energy storage can clearly distinguish between lithium batteries (high-power...) ,medium ) and lead-acid batteries (low ,Low The elasticity differences provide a detailed basis for resource selection and scheduling.

[0093] In scenarios such as power systems, systems need to cope with the uncertainties of load fluctuations and renewable energy integration. The resource resilience index provides quantitative support for the system's "flexible dispatch" and "stable operation." By screening highly resilient resources through this index, the system's ability to respond to changes can be improved, promoting renewable energy consumption, ensuring grid stability, and assisting in optimized system operation. Therefore, the resilience index is used as a key parameter in the calculation of operational stability indicators, providing core data reflecting resource resilience capabilities for higher-level applications such as resource integration assessment and dispatch strategy formulation. This makes system decisions more scientific and aligned with actual operational needs, driving the power system towards intelligent and flexible regulation.

[0094] Distributed resources to be accessed operational stability indicators The calculation formula is:

[0095] ;

[0096] The calculation of the operational stability indicators of the resources to be connected is based on the perception parameters of the distributed resources collected by the edge nodes. For each resource to be connected... This distributed resource requires comprehensive data analysis, considering both real-time power sensing data collected from different edge nodes and the inherent health and resilience characteristics of the distributed resource itself. The calculation involves a weighted summation of two parts: the first part focuses on power fluctuations, starting with each edge node... Calculate the number of samples collected. Real-time active power of distributed resources The average active power of the distributed resource collected by the edge node The difference is then divided by the maximum active power of the resource collected by that node. and minimum value The difference is used to obtain the normalized value of the power fluctuation of a single node, which is then multiplied by the weight of that edge node. And for all Sum the results of each node, and then divide by the number of nodes. Then multiply by the weighting factor The first part quantifies the impact of resource power fluctuations on stability; the second part focuses on the characteristics of distributed resources themselves, including device health parameters. and elasticity index Multiply by the weighting factor. ( This is used to measure the impact of equipment health and resilience on stability. Through this calculation, factors such as power fluctuations of resources and the inherent characteristics of the equipment are integrated to ultimately obtain an operational stability index. This indicator reflects the stability of the distributed resources to be connected, providing a basis for subsequent distributed resource access optimization decisions. It helps to select more stable and system-compatible distributed resources for access, ensuring the safe and efficient operation of the power system. Therefore, by comprehensively considering factors such as power data fluctuations, the health status of the distributed resource equipment itself, and the potential impact on the system after access, the stability characteristics of distributed resource operation are quantified. Characteristics such as whether the output power of distributed resources is stable and whether the equipment has continuous and reliable operation capabilities are transformed into stability indicators that can be used for evaluation and comparison. This allows for the assessment of the potential impact of different resources on the stable operation of the system, providing a crucial basis for subsequent selection of suitable access resources and planning of access strategies.

[0097] S3: Using operational stability indicators, unit power access cost, planned access power, access time window parameters, and carbon trading cost factors, an objective function is constructed, and then the global objective function value of the distributed sensing network and the local objective function value of each edge node are calculated. The optimization scheme in the collaborative optimization mechanism uses minimizing the global objective function value as the optimization objective of the algorithm, iteratively optimizing to generate the optimized results of the planned access power values ​​of each resource within the jurisdiction of each edge node.

[0098] By combining the actual situation of the distributed resources to be integrated, such as the number, type, capacity limitations, and layout location of the distributed resources, and the previously calculated operational stability indicators, the objective function value is constructed. The specific process is as follows:

[0099] S3.1: Obtaining Distributed Resources operational stability indicators Unit power access cost Planned access power Access time window parameters Target weight coefficient Carbon price coefficient ;

[0100] S3.2: For each distributed resource Based on its projected carbon emissions and the benchmark emissions of the industry The corresponding carbon trading cost factor is calculated. :

[0101]

[0102] The logic of "quantifying carbon trading costs" aims to transform the differences in carbon emissions resulting from the integration of distributed resources into economic cost factors that can be used for decision-making. Specifically, it is based on industry benchmark emissions for specific resource types. As a benchmark, calculate the estimated carbon emissions of the resources to be connected. The ratio to this baseline emission level reflects the relative extent of resource-based carbon emissions compared to the industry average. Subsequently, a carbon price coefficient is introduced. This represents the economic cost per unit of carbon emission (such as the price of one ton of carbon in the carbon trading market). Multiplying the relative carbon emission level by the carbon price coefficient yields the additional cost incurred due to deviations from the benchmark. Finally, using "1" as the benchmark state with no additional carbon cost, the formula... Integrate the economic impact of carbon emissions into carbon trading cost factors. This is used for cost accounting in subsequent resource integration decisions. It ensures that low-carbon resources (such as solar and wind power) have carbon emissions far below industry benchmarks. Resources with carbon emissions approaching 1 or even less have a natural advantage in competition for resource access; high-carbon emission resources (such as traditional coal-fired power) are... This significantly increases the likelihood of the power system prioritizing low-carbon resources from a technological decision-making perspective. Regarding cost suitability, the carbon price coefficient... It can be flexibly adjusted according to regional carbon market price fluctuations and policy guidance, increasing the weight of high-carbon resource costs when carbon prices rise; for emerging low-carbon technologies, it can create cost advantages by adjusting coefficients or benchmark values, adapting to carbon management needs at different stages. Simultaneously, it works in conjunction with the global objective function of resource access to construct a full-cost closed loop of "stability-economy-carbon constraints," allowing decision-making to move beyond hardware costs and stability, incorporating long-term implicit carbon costs, avoiding the problem of short-term economic optimization but long-term uncontrolled carbon risks, and ensuring scientific and sustainable decision-making. Furthermore, industry benchmark emissions... Customization is available based on specific industry sectors and equipment types.

[0103] S3.3: Based on target weight coefficients Unit power access cost Planned access power Access time window parameters and the corresponding carbon trading cost factor Construct the economic cost in the objective function ;

[0104]

[0105] in, This represents the total number of distributed resources.

[0106] S3.4: Based on target weight coefficients and the operational stability indicators of each distributed resource Constructing the stability benefit in the objective function :

[0107] ;

[0108] In practical implementation, the target weight coefficient These are empirical values, preset based on the current operating state of the power grid: when the power grid is in stability-priority mode, a larger value is set. Value; when the power grid is in an economy-first mode, a smaller value is set. value.

[0109] S3.5: Overall stability benefits Economic cost The final objective function is obtained, and the calculation formula is:

[0110] .

[0111] In practical implementation, the global objective function of the distributed sensing network... The total number of distributed resources represented is the number of distributed resources in the entire distributed sensing network, and the local objective function of each edge node is... The total number of distributed resources represented is the number of distributed resources under the jurisdiction of each edge node.

[0112] Based on operational stability indicators (This reflects the impact of resource access on system stability.) (The smaller the value, the more stable it is), take its reciprocal. Amplify the positive contribution of "stable resources". Through (Target weighting coefficient, the proportion of stability control in the overall target), transforming the requirement of "prioritizing access to stable resources" into mathematical constraints— The smaller, The larger the value, the better for the objective function. The stronger the positive contribution, the more likely the system will be to access stable resources.

[0113] when At times, the system prioritizes "stability," making it suitable for scenarios with extremely high stability requirements, such as peak grid loads and high penetration of new energy sources; when... At this time, the system prioritizes "economic efficiency" and is suitable for scenarios with surplus resources and cost sensitivity.

[0114] The intermediate value dynamically balances the two, and through mathematical expression, the strategy requirements of "stability priority / cost priority / collaborative optimization" are transformed into a calculable and iterative global objective function, supporting the system to automatically select the optimal access solution.

[0115] Target weight coefficient As a core variable for balancing stability and economy, the solution allows for flexible adaptation to complex and diverse application scenarios. In scenarios with stringent requirements for system stability, such as peak grid loads and high penetration of renewable energy, increasing... This allows the objective function to prioritize operational stability indicators. (because (Increased share), guiding the system to prioritize access Small, stable resources (such as energy storage devices and high-quality photovoltaic power plants with stable output) effectively mitigate the impact of resource access on the main grid, ensuring the reliable operation of the power system; while in scenarios with resource surplus and greater sensitivity to cost control (such as off-peak hours on the user side), reducing... This highlights the economic dimension ( Increased weighting helps in selecting low-cost, low-carbon emission resources (such as idle small-capacity wind power) and reducing electricity costs. Simultaneously, a carbon trading cost factor is introduced. This can enhance the adaptability to low-carbon goals. Under the guidance of the "dual carbon" policy, the weight of this factor can be adjusted to promote the integration of a high proportion of clean energy (photovoltaics, wind power, etc.) and help achieve the goal of green energy transformation.

[0116] Traditional resource access decisions often focus on direct hardware costs, neglecting implicit costs. This solution considers the cost per unit power for access. Access power Access time window parameters Carbon trading cost factor Comprehensive integration enables precise quantification of the entire lifecycle cost of resource access (from basic construction and commissioning costs to scheduling costs incurred during operation due to access timing, and emission costs under carbon constraints). Taking high-carbon-emission small coal-fired power resources as an example, even their direct access costs... Low, but because High values ​​will be "penalized" in economic calculations (the corresponding value increases, lowering the overall economic efficiency), avoiding the neglect of long-term carbon risks and resource waste due to a one-sided focus on direct costs, and helping the power system achieve more scientific and sustainable cost control in the resource access process.

[0117] in the formula , These parameters are open and can be flexibly adjusted with technological advancements and policy iterations. In the future, if new resource assessment indicators such as "voltage support capacity" and "frequency response speed" need to be introduced (e.g., to address the needs of new power systems for coordinated regulation of power generation, grid, load, and storage), new calculation dimensions can be easily extended based on the existing global objective function (e.g., adding calculation terms and corresponding weight coefficients associated with the new indicators). This eliminates the need for fundamental restructuring of the solution, ensuring that the technical solution can adapt to the long-term needs of power system evolution. It provides flexible and scalable architectural support for continuous system optimization and upgrading, and facilitates efficient handling of complex decision-making needs in the construction of new power systems.

[0118] In practice, the final objective function value This is the fitness value used by subsequent optimization algorithms for iterative optimization; the optimization objective is to find the fitness value that makes the algorithm more suitable for optimization. Minimize the power combination of distributed resource access , thus obtaining the optimized solution , The total number of edge nodes, where, , For edge nodes Distributed resources within jurisdiction Access power, For edge nodes The total number of distributed resources under management. More specifically, optimization algorithms include those employing genetic algorithms, particle swarm optimization, etc.

[0119] The construction of the global objective function addresses the system's diverse needs for resource access, including energy efficiency improvement, cost control, and security. It transforms complex resource access optimization requirements into a mathematical function, solving for the global objective function value. This clarifies how resource access should be configured to better align with the overall system optimization direction, providing clear quantitative guidance for subsequent optimization decisions. Qualitative decision-making strategies such as "prioritizing stability" and "controlling access costs" are transformed into... This clear quantitative calculation model lays the foundation for the application of intelligent algorithms (such as genetic algorithms and particle swarm optimization). Only the resources to be accessed need to be input. , , With these parameters, the system can automatically iterate and optimize using algorithms to quickly output the "optimal resource access combination." Compared to the traditional model that relies on human experience and judgment, this significantly improves decision-making efficiency, avoids insufficient human experience or subjective judgment bias, significantly enhances decision-making accuracy, reduces reliance on human expertise and experience, and promotes the intelligent and automated upgrading of power system resource access decision-making.

[0120] S4: Calculate the consistency index based on the number of edge nodes and the local objective function value of the nodes, and determine whether the collaborative optimization mechanism needs to be activated to adjust the optimization results of each edge node based on the consistency index.

[0121] S4.1: Obtain the optimized local objective function value of each edge node based on local computation. ,in, , This represents the total number of edge nodes.

[0122] ;

[0123] in, For edge nodes Local weighting coefficients; For edge nodes Under the jurisdiction of the Local operational stability metrics for distributed resources; For edge nodes Under the jurisdiction of the Local unit power access cost of a distributed resource; For edge nodes Under the jurisdiction of the Local real-time access power of a distributed resource; For edge nodes Under the jurisdiction of the Local carbon trading costs for distributed resources; For edge nodes Under the jurisdiction of the Local carbon trading costs for distributed resources; For edge nodes The total number of distributed resources under its jurisdiction;

[0124] S4.2: Calculate any two edge nodes and Data heterogeneity between The data heterogeneity is used to quantify the difference in data processing capabilities between two nodes.

[0125] ;

[0126] in, For edge nodes For the first Preprocessing delay of class-aware parameters; For edge nodes For the first Preprocessing delay of class-aware parameters; To determine the number of types of sensing parameters.

[0127] Data heterogeneity It measures two edge nodes ( and The degree of difference in data processing characteristics between different types of sensing parameters. First, based on the number of types of sensing parameters... and edge nodes , For the first Preprocessing latency of class data and For each type of perception parameter (from...) arrive ), calculate the absolute difference in preprocessing latency between the two nodes. Then calculate the sum of the preprocessing delays of the two. The ratio of the absolute difference to the sum is used as the heterogeneous contribution between nodes under this type of sensing parameter. Subsequently, for all... Summation of heterogeneous contributions of class-aware parameters, through Obtain edge nodes and Data heterogeneity This allows for the quantification of differences in data preprocessing latency characteristics among nodes.

[0128] From a data characteristic quantification perspective, it provides a precise means to quantify the data heterogeneity among edge nodes. By focusing on preprocessing latency, a key data processing characteristic, and combining it with the comprehensive calculation of multiple sensing parameters, it can meticulously reflect the differences in data processing capabilities and efficiency among different nodes. This allows the system to clearly understand the degree of data heterogeneity among nodes, providing a foundation for subsequent heterogeneity-based decision optimization. In terms of collaborative decision optimization, data heterogeneity... It can incorporate higher-level calculations such as consistency metrics, enabling the system to fully consider the impact of data processing differences when evaluating node collaboration levels, and avoiding decision-making biases caused by ignoring heterogeneity. For example, in the consistency assessment of resource access decisions, it can be combined with... It can more accurately determine whether the disagreement between nodes stems from the decision itself or from differences in data processing, thereby guiding nodes to perform operations such as data synchronization and processing flow optimization, improving the overall collaborative efficiency and decision quality of the edge computing system, and ensuring the stable and efficient operation of the system in distributed application scenarios.

[0129] S4.3: Based on the local objective function values ​​corresponding to any two edge nodes , and the corresponding data heterogeneity Calculate the network-wide consistency index :

[0130] .

[0131] Consistency Indicators It is used to measure the consistency of local objective function values ​​among edge nodes. It is based on the number of edge nodes. and the local objective function values ​​corresponding to any two edge nodes. , There is also data heterogeneity, which reflects the differences in data between nodes. During the calculation, for each pair of edge nodes ( and First, calculate the absolute difference between the local objective function values ​​of the two. Combine it with data heterogeneity Add them together and take the reciprocal to get the consistency contribution value for this pair of nodes. Then, for all node pairs (from... arrive , arrive Sum the contribution values ​​of ( ).

[0132] From the perspective of consistency measurement, an effective method for quantifying the consistency of local objective functions among edge nodes is constructed by incorporating data heterogeneity. This approach not only considers differences in objective function values ​​but also takes into account the heterogeneous characteristics of node data itself. This makes consistency assessment more aligned with real-world scenarios (such as differences in decision-making foundations caused by variations in the data types and precision collected by different nodes). It accurately reflects the level of collaboration among multiple edge nodes in resource access and other decision-making processes, providing a quantitative basis for judging the overall consistency of system decisions and identifying disagreements between nodes. Regarding system optimization, consistency indicators... It can be used to guide collaborative optimization between edge nodes, if If the value is too low, it indicates that there are large differences in the local objective functions between nodes and insufficient coordination. This can trigger data interaction and parameter adjustment between nodes, promote the overall decision-making of the system to develop in a more consistent and better direction, improve the coordination efficiency and decision-making quality of the edge computing system in tasks such as resource allocation and operation control, and ensure the stability and efficiency of the system operation in application scenarios such as distributed resource access.

[0133] S4.4: The consistency index Consistency threshold with preset If a comparison is made, If the condition is met, it is determined that the collaborative optimization mechanism needs to be activated; otherwise, it is determined that it does not need to be activated.

[0134] This is used to determine whether a collaborative optimization mechanism needs to be activated to ensure the consistency of optimization results among edge nodes. First, a consistency index reflecting the degree of consistency of the objective function among edge nodes is calculated. Next, this indicator is compared with a pre-set indicator threshold. Compare them. If the consistency index Less than the preset threshold This indicates that the objective function values ​​among the edge nodes differ significantly and the degree of collaboration is insufficient. In this case, a collaborative optimization mechanism is activated to adjust the optimization results of each edge node based on a certain adjustment amount, thereby promoting greater consistency in the optimization results among the nodes. Greater than or equal to If the current consistency between nodes is considered good, then collaborative optimization is not required. From the perspective of ensuring decision consistency, by setting thresholds and comparing consistency indicators, a clear criterion for "whether to collaborate" is provided. This avoids overall decision failure or system instability caused by excessive differences in optimization results between nodes, ensuring that the optimization results of each node in edge computing systems are consistent in resource allocation, operation control, and other tasks, thus improving the reliability of system decisions. Regarding system adaptive optimization, when the collaborative optimization mechanism is activated, the optimization results of nodes can be adjusted specifically, enabling the system to have adaptive adjustment capabilities. It can dynamically optimize decisions based on the actual consistency between nodes, enhancing the edge computing system's ability to cope with complex scenarios and dynamic changes, and ensuring continuous, efficient, and stable operation of the system in application scenarios such as distributed resource access.

[0135] S4.5: If it is determined that a collaborative optimization mechanism needs to be activated, then each edge node should move to its neighbor nodes. The nodes in the process exchange their respective optimization schemes. ;in, For one A dimensional vector, whose corresponding vector components are for nodes Planned values ​​for each resource under jurisdiction;

[0136] In specific implementation, the global objective function value The fitness value, which serves as the basis for iterative optimization, is the optimization objective to find a value that makes the optimization process more efficient and efficient. Minimize the power combination of distributed resource access That is, to obtain the optimized solution. , The total number of edge nodes, where, , For edge nodes Distributed resources within jurisdiction Access power, For edge nodes The total number of distributed resources under management. More specifically, optimization algorithms include those employing genetic algorithms, particle swarm optimization, etc.

[0137] S4.6: Each edge node calculates the adjustment amount of its own optimization scheme based on the optimization schemes received from its neighboring nodes. The local optimization scheme is then updated based on this adjustment; the formula for calculating the adjustment is as follows:

[0138] ;

[0139]

[0140] in, For coordinated step size; For nodes The set of neighboring nodes; Neighboring nodes Optimization scheme; For nodes With nodes Communication quality factor between them; For edge nodes with neighboring nodes The local communication timeliness factor is a parameter used to measure the timeliness of the communication link between an edge node and its neighboring nodes. It reflects the real-time response capability of communication interaction. The longer the time interval between two communications, the lower the value of this factor becomes with the decay rate, thus reflecting the degree of decay of the communication link timeliness over time. The shorter the time interval, the closer the factor value is to the standard level, which means that the real-time performance and response efficiency of the communication are better. The decay rate; For nodes With nodes Last successful communication time; This is the current time of successful communication.

[0141] Taking into account factors such as differences in neighboring node optimization schemes, communication quality, and communication time differences, an adjustment calculation model is constructed. This model accurately reflects the impact of inter-node collaboration needs and communication characteristics on optimization, ensuring that the adjustment reflects both the complementarity of optimization schemes between nodes and adapts to the actual communication environment, thus improving the accuracy of collaborative optimization. Regarding system adaptability, a communication time difference-related factor is introduced. Considering the role of communication timeliness in collaboration, the system can adapt to dynamically changing communication environments (such as communication delays between nodes, interruption recovery, etc.), ensuring that edge nodes can still effectively collaborate and optimize under complex communication conditions, enhancing the edge computing system's ability to cope with dynamic scenarios, promoting the collaboration of optimization results among nodes in applications such as distributed resource access, and improving the overall operating efficiency and stability of the system.

[0142] S4.7: Based on the updated optimization scheme, repeat S4.1 to S4.4, recalculate the consistency index and make judgments until a new consistency index is reached. satisfy Or it may reach the preset maximum number of iterations.

[0143] In terms of collaborative accuracy, a consistency index is calculated based on the number of edge nodes and the local objective function values ​​calculated by each node. Because edge nodes in a distributed system collect data and have different computational perspectives, local optimizations may occur, but the overall system may be inconsistent. The consistency index is a key metric used to measure whether the optimization directions of edge nodes in a distributed network are coordinated. If the index shows significant differences in the optimization schemes of each node, a collaborative optimization mechanism is activated. This allows the dispersed edge nodes to exchange optimization information and adjust their own schemes while preserving the characteristics of their local data. Through multiple iterative calculations, the objective function values ​​of each node tend to converge, achieving a globally unified access scheme and avoiding conflicts between local optimizations and the overall system objectives.

[0144] S5: Distribute the optimized access scheme to each edge node, recalculate the consistency index based on the node feedback, and return to S4 if the consistency index exceeds the threshold.

[0145] In practice, the optimized access scheme is distributed to each edge node. The edge nodes act as execution terminals, controlling the access operations of corresponding resources according to the scheme, such as adjusting the output of distributed power sources and managing the charging and discharging of energy storage devices. Simultaneously, the system's operational status is monitored in real time after resource access. If a consistency indicator exceeds a preset threshold, indicating a change in the resource access environment or system requirements, the process from data collection to optimization decision-making is restarted, forming a closed loop of "perception-analysis-decision-execution-monitoring-re-optimization." This ensures that distributed resource access always adapts to the dynamic needs of the system, achieving continuous and efficient optimization.

[0146] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0147] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for optimizing real-time access to distributed resources based on edge computing, characterized in that, include: S1: Construct a distributed sensing network based on the location information and topological relationship of edge nodes, obtain the sensing parameters of each distributed resource to be accessed and preprocess them, and construct the sensing dataset corresponding to each edge node. S2: Calculate the operational stability index of each distributed resource to be accessed based on the data in the perception dataset; S3: Using operational stability indicators, unit power access cost, planned access power, access time window parameters, and carbon trading cost factors, an objective function is constructed, and then the global objective function value of the distributed sensing network and the local objective function value of each edge node are calculated; among them, the optimization scheme in the collaborative optimization mechanism takes minimizing the global objective function value as the optimization objective of the optimization algorithm, and iteratively optimizes to generate the optimization results of the planned access power value of each resource within the jurisdiction of each edge node. S4: Calculate the consistency index based on the local objective function value and data heterogeneity of each edge node, and determine whether the collaborative optimization mechanism needs to be activated to adjust the optimization results of each edge node based on the consistency index; S5: Distribute the optimized access scheme to each edge node, recalculate the consistency index based on the node feedback, and return to S4 if the consistency index exceeds the threshold.

2. The method according to claim 1, characterized in that, The specific process of S2 is as follows: S21: For the first A distributed resource to be connected, from Each edge node acquires its corresponding perceptual dataset; among which... This represents the total number of edge nodes participating in data collection and computation in the current scenario. S2.2: For each node Computing distributed resources Normalized value of power fluctuation ; ； in, For nodes Distributed resources Real-time power; For nodes Distributed resources Average power; For nodes Distributed resources Minimum power; For nodes Distributed resources Maximum power; S2.3: Weights based on each edge node ,right The power fluctuation normalized values ​​of each node are weighted and fused to calculate the distributed resource. Comprehensive power fluctuation value ; S2.4: Based on distributed resources Comprehensive power fluctuation value Combined with the first Device health parameters of a distributed resource and elasticity index Calculate and obtain the distributed resources to be accessed operational stability indicators .

3. The method according to claim 2, characterized in that, Weights of each edge node The calculation formula is ； in, For edge nodes The shortest path length to each of the other nodes.

4. The method according to claim 2, characterized in that, Distributed resources to be accessed operational stability indicators The calculation formula is: ； ； in, and All are weighting coefficients, and ; Adjustable power for distributed resources; Rated power for distributed resources; This refers to the load response sensitivity parameter.

5. The method according to claim 1, characterized in that, The specific process of S3 is as follows: S3.1: Obtain distributed resources in the target region operational stability indicators Unit power access cost Planned access power Access time window parameters Target weight coefficient Carbon price coefficient ; S3.2: For each distributed resource in the target region Based on its projected carbon emissions and industry benchmark emissions The corresponding carbon trading cost factor is calculated. : S3.3: Based on target weight coefficients Unit power access cost Planned access power Access time window parameters and the corresponding carbon trading cost factor Construct the economic cost in the objective function ; S3.4: Based on target weight coefficients and the operational stability indicators of each distributed resource Constructing the stability benefit in the objective function : S3.5: Overall stability benefits Economic cost Thus, the final objective function is obtained; where the formula for calculating the objective function is: 。 6. The method according to claim 5, characterized in that, Carbon trading cost factor The calculation formula is: 。 7. The method according to claim 5, characterized in that, Economic cost in the objective function The calculation formula is: ； Stability gains in the objective function : ； in, The total number of distributed resources within the target area; where the target area refers to the global area of ​​the distributed sensing network or the jurisdiction area of ​​each edge node.

8. The method according to claim 1, characterized in that, The specific process of S4 is as follows: S4.1: Get edge nodes The optimized local objective function value obtained based on local computation ,in, , This represents the total number of edge nodes. S4.2: Calculate any two edge nodes and Data heterogeneity between ; S4.3: Based on the local objective function values ​​corresponding to any two edge nodes , and the corresponding data heterogeneity Calculate the network-wide consistency index ; S4.4: The consistency index Consistency threshold with preset If a comparison is made, If the condition is met, it is determined that the collaborative optimization mechanism needs to be activated; otherwise, it is determined that it does not need to be activated. S4.5: If it is determined that a collaborative optimization mechanism needs to be activated, then each edge node should move to its neighbor nodes. The nodes in the process exchange their respective optimization schemes. ;in, For one A dimensional vector, whose corresponding vector components are for nodes The planned value of each distributed resource under its jurisdiction; S4.6: Each edge node calculates the adjustment amount of its own optimization scheme based on the optimization schemes received from its neighboring nodes. And update the local optimization scheme based on the adjustment amount; The formula for calculating the adjustment amount is as follows: ； ； in, For coordinated step size; For nodes The set of neighboring nodes; Neighboring nodes Optimization scheme; For nodes With nodes Communication quality factor between them; For edge nodes with neighboring nodes Local communication timeliness factor between them; The decay rate; For nodes With nodes Last successful communication time; This is the current time of successful communication; S4.7: Based on the updated optimization scheme, repeat S4.1 to S4.4, recalculate the consistency index and make judgments until a new consistency index is reached. satisfy Or it may reach the preset maximum number of iterations.

9. The method according to claim 8, characterized in that, edge nodes and Data heterogeneity between The calculation formula is as follows: ； in, For edge nodes For the first Preprocessing delay of class-aware parameters; For edge nodes For the first Preprocessing delay of class-aware parameters; To determine the number of types of sensing parameters.

10. The method according to claim 8, characterized in that, Network-wide consistency indicators The calculation formula is as follows: 。

Images