A distributed resource control policy generation method and system

Abstract

The application discloses a kind of distributed resource control strategy generation method and system, it is related to power distribution network operation control technical field, method includes obtaining power distribution network multi-source operation data, calculates the information value degree of each type of operation data to preset control target, filters out low-value data to form data set;Combining network frame topology calculates node electrical coupling strength and cluster division. With cluster as unit modeling, optimize voltage qualification rate;Load balancing degree, issue aggregation control instruction. With the minimum adjustment cost in cluster, solve distributed resource active power;Reactive power output, integrate output result to generate collaborative control strategy.The application reduces the model scale, greatly improves the operation efficiency, and can quickly generate large-scale distributed resource collaborative control scheme.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation control technology, and in particular to a method and system for generating distributed resource control strategies. Background Technology

[0002] With the deepening of the construction of new power systems, the scale of distributed resources such as distributed photovoltaic, wind power, energy storage, and electric vehicle charging piles in the distribution network continues to expand, and the distribution network is gradually evolving from a traditional passive radial network to an active multi-terminal complex network. Traditional dispatching modes are difficult to adapt to its flexible control requirements, and there is an urgent need to develop efficient and coordinated control technologies for large-scale distributed resources.

[0003] Currently, for the generation of distributed resource collaborative control strategies, the global centralized optimization method is an important technical approach. This method collects measurement data from the entire distribution network, constructs a unified optimization model, and uses intelligent optimization algorithms to solve for the output scheme of each distributed resource with objectives such as voltage compliance and minimum network loss. However, this method faces significant challenges in practical engineering applications: the value density of various operational data in the distribution network varies greatly for control decisions, and the measurement data of some remote nodes contributes very little to the local control objectives, belonging to low-value redundant information. However, existing methods do not distinguish between all measurement data; including all of them in the optimization model not only increases the model size and computational complexity, but the additional noise introduced by low-value data may also interfere with the judgment of the optimization direction. At the same time, when the number of distributed resources reaches hundreds or even thousands, the dimensionality of decision variables in the centralized optimization model increases sharply, facing serious convergence difficulties and computational time problems, and the generation efficiency of control strategies cannot meet the real-time requirements of online applications. Summary of the Invention

[0004] To address the problems of existing technologies that indiscriminately include all measurement data in the optimization model, leading to increased computational burden from low-value redundant data and excessively high dimensionality of decision variables in large-scale centralized optimization, thus making solution difficult, this invention provides a method and system for generating distributed resource control strategies. By filtering low-value data through data value selection to simplify the model input size, and by using partitioned aggregation to decouple large-scale optimization into parallel solvable subproblems to compress the dimensionality of decision variables, this method reduces model size, improves solution efficiency, and enables efficient generation of large-scale distributed resource control strategies. The specific technical solution is as follows: In a first aspect, the present invention provides a method for generating a distributed resource control strategy, comprising the following steps: Acquire multi-source operation data of the power distribution network, calculate the information value of various types of operation data to the preset control target, and remove data categories whose information value is lower than the preset threshold to form a dataset; Based on the dataset and distribution network topology, the electrical coupling strength between nodes is identified and clusters are divided. Taking each cluster as a decision unit, the aggregation control model with the goal of maximizing the cluster voltage qualification rate and the optimal load balance is solved, and aggregation control instructions for each cluster are generated. The aggregated control command is issued as a constraint to each cluster; within each cluster, the active and / or reactive power output allocation values ​​of each distributed resource are solved with the minimum adjustment cost of the distributed resources as the sub-objective. Summarize the output allocation values ​​of each distributed resource within each cluster, generate and output the collaborative control strategy of the target output curve of each distributed resource.

[0005] Preferably, the information value is calculated based on mutual information theory, specifically including: For each type of operational data and each control target indicator, calculate the mutual information value; The mutual information values ​​are normalized to obtain the information value of each type of operational data.

[0006] Preferably, the formula for calculating the mutual information value is: in, Indicates the first Time series of class running data, Represents the time series of regulatory target indicators; The joint probability distribution of the two types of data. and Each of these represents a marginal probability distribution, wherein both the joint probability distribution and the marginal probability distribution are estimated from the time series using a kernel density estimation method.

[0007] Preferably, the calculation of the information value of various types of operational data to the preset control target specifically includes: The preset control objectives include the voltage deviation minimization objective and the load balance optimization objective; the information value of each type of operating data in the voltage deviation dimension and the load balance dimension are calculated respectively, and the information value of the voltage deviation dimension and the load balance dimension are weighted and summed according to preset weight coefficients to obtain the comprehensive information value of each type of operating data.

[0008] Preferably, the identification of electrical coupling strength between nodes and the division of the cluster includes: Based on the network topology and line parameters of the distribution network, establish the node admittance matrix; Inverting the node admittance matrix yields the node impedance matrix, which in turn allows us to derive the voltage sensitivity matrix to reactive power. An electrical distance metric between nodes is constructed based on the elements of the sensitivity matrix. Based on the electrical distance matrix, the distribution network nodes are automatically clustered to obtain the cluster partitioning results.

[0009] Preferably, the optimization objective of the aggregation control model is: in, For clusters Voltage at the point of common coupling, Rated voltage, For clusters The equivalent load rate, The average load rate of the entire cluster. and These are the weighting coefficients; The decision variables of the aggregation control model are the total active power regulation and the total reactive power regulation of each cluster. The constraints include the upper and lower limits of the voltage at the common coupling point of the clusters; the thermal stability constraints of the power of the interconnect lines between clusters; and the total range constraints of the adjustable power of each cluster.

[0010] Preferably, the adjustment cost of each distributed resource is determined according to a preset weighting coefficient based on the type of the distributed resource.

[0011] Preferably, the step of minimizing the adjustment cost of distributed resources and solving for the active and / or reactive power output allocation values ​​of each distributed resource specifically includes: Using the aggregation control command as a power balance equation constraint, the current output value and type weight coefficient of each distributed resource in the cluster are obtained; Each distributed resource in the cluster is configured with a proxy node, and each proxy node calculates the initial marginal adjustment cost based on the type weight coefficient and the current output level. Each agent node communicates with its neighboring nodes, taking the minimization of the adjustment cost as the convergence direction, and iteratively updates its own marginal adjustment cost estimate and corresponding power allocation value by exchanging marginal adjustment cost information. The power allocation value satisfies the power balance equation constraint and individual operation constraint. The iteration stops when the estimated marginal adjustment costs of each agent node become consistent, and the current output allocation value is output as the output allocation value.

[0012] Preferably, the collaborative control strategy further includes triggering correction conditions; When the deviation between the voltage of the critical node of the distribution network monitored in real time and the expected voltage value of the main strategy exceeds the preset dead zone threshold, and / or the deviation between the actual load rate of the critical line and the expected load rate of the main strategy exceeds the preset dead zone threshold, the strategy correction procedure is triggered. The correction procedure maintains the aggregation control instructions of each cluster unchanged, re-executes the autonomous allocation process within each cluster, calculates the incremental correction amount of the output of each distributed resource, and generates additional adjustment instructions to be superimposed on the currently executed control instructions.

[0013] Secondly, the present invention also provides a distributed resource control strategy generation system, which applies the aforementioned method and includes: The data value filtering module is used to acquire multi-source operation data of the distribution network, calculate the information value of various types of operation data to preset control targets, and form a dataset based on the information value. The partitioned aggregation solution module is used to identify the electrical coupling strength between nodes based on the dataset and the distribution network topology, and to divide the nodes into clusters accordingly. It also uses each cluster as a decision unit to solve the aggregation control model with the goal of achieving the highest cluster voltage qualification rate and the best load balance, and generates aggregation control instructions for each cluster. The autonomous allocation module within the area is used to issue the aggregated control command as a constraint to each cluster, and within each cluster, with the minimum adjustment cost of distributed resources as the sub-objective, solve for the active and / or reactive power output allocation values ​​of each distributed resource. The strategy generation and output module is used to summarize the output allocation values ​​of each distributed resource in each cluster, generate a collaborative control strategy containing the target output curves of each distributed resource, and output it.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention introduces a data value filtering mechanism to eliminate low-value, redundant data based on information value, thereby reducing data dimensionality at the source. It also eliminates invalid noise, effectively reducing the input size of the optimization model. Furthermore, by using partitioned aggregation to solve the large-scale global optimization problem, it decouples the problem into several parallel-computable subproblems, compressing the hundreds or thousands of decision variables in the original centralized optimization to a cluster-level number of decision variables, significantly reducing the difficulty and computation time. The synergistic effect of data filtering and partitioned dimensionality reduction jointly solves the problem of low optimization efficiency caused by massive amounts of low-value data, greatly improving the real-time performance of control strategy generation and the quality of the solution. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0016] Figure 1 A flowchart of a distributed resource control strategy generation method provided in an embodiment of the present invention.

[0017] Figure 2 The flowchart of the distributed algorithm for autonomous allocation within a region is provided for an embodiment of the present invention.

[0018] Figure 3 This is an overall flowchart of a distributed resource control strategy generation method provided in an embodiment of the present invention.

[0019] Figure 4 This is a schematic diagram of a distributed resource control strategy generation system provided in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be understood that, when used in this specification, the terms “comprising” and “including” indicate the presence of the described feature; whole; step; operation; element and / or component, but do not exclude the presence or addition of one or more other features; whole; step; operation; element; component and / or collection thereof.

[0022] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” are intended to include the plural forms.

[0023] It should also be further understood that the term "and / or" as used in this specification refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0024] Please refer to the following examples. Figures 1 to 4 .

[0025] This invention provides a method for generating distributed resource control strategies, applicable to the coordinated control strategy generation scenario of active distribution networks containing high-penetration distributed photovoltaic, wind power, energy storage, and flexible loads. This method relies on a hierarchical collaborative distributed resource control architecture, which includes four core layers: a data value filtering layer; a partitioned aggregation solution layer; an intra-regional autonomous allocation layer; and a strategy generation output layer. Please refer to [link to relevant documentation]. Figure 1 The method provided in this embodiment of the invention includes the following steps: Step S1: Obtain multi-source operation data of the distribution network, calculate the information value of various types of operation data to the preset control target, and remove data categories whose information value is lower than the preset threshold to form a dataset; First, the distribution network dispatch and control system accesses multi-source operational data from the entire network through a data acquisition and monitoring system. This data originates from feeder terminal units deployed at each node and mainly includes the following types: voltage amplitude measurements at each node, reflecting the voltage level at various points in the distribution network; active and reactive power measurements injected at each node, reflecting the distribution of load consumption and power output; current active and reactive power output measurements of each distributed resource, specifically covering the output of distributed photovoltaic inverters; wind turbine output; charging and discharging power of energy storage systems; load power of electric vehicle charging piles; and output of small hydropower stations; and switch position status information for each branch, used to determine the current topological connections of the distribution network.

[0026] It should be noted that not all of the aforementioned multi-source operational data has equal value for the subsequent generation of collaborative control strategies. Measurement data from some remote nodes may contribute very little to the local control objectives, and the correlation between power measurement information from certain branches and voltage control objectives is also weak. If all data are used indiscriminately as input to the optimization model, it will not only increase the model size and computational burden but may also introduce additional measurement noise, interfering with the determination of the optimization direction. Therefore, this embodiment evaluates the value of each type of operational data. Information value measures the statistical correlation strength between a certain type of operational data and the preset control objectives.

[0027] The preset control objectives in this embodiment are to minimize feeder voltage deviation and optimize line load balance. Minimizing voltage deviation means making the voltage of each node in the distribution network as close as possible to the rated value to avoid overvoltage or undervoltage problems; optimizing load balance means making the load rate of each feeder and each transformer tend to be balanced to avoid situations where some equipment is heavily loaded or overloaded while other equipment is lightly loaded.

[0028] For the two control objectives mentioned above, the mutual information value between each type of operational data and the target indicator is calculated. Mutual information, derived from information theory, is used to measure the degree of interdependence between two random variables, and its calculation formula is as follows: in, Indicates the first Time series of class running data, The time series of the control target indicators (voltage deviation indicators or load balancing indicators); The joint probability distribution of the two types of data. and These represent their respective marginal probability distributions. This formula quantifies the amount of information shared between the two variables by statistically analyzing the difference between their joint and independent distributions. A higher mutual information value indicates a stronger statistical correlation between the operational data and the regulatory target, and thus a greater guiding role in regulatory decisions.

[0029] For continuous measurement time series data and Its probability distribution is nonparametrically estimated using the kernel density estimation method. Specifically, for a length of... Time series Its probability density function is estimated as follows: in, For Gaussian kernel function, , This is the bandwidth parameter.

[0030] bandwidth Determined using Silverman's rule: ,in The standard deviation of the sample. Joint probability density. The estimation uses a two-dimensional kernel density estimation: Sampling window length The number of measurement data points within a complete control cycle preceding the current moment is taken, where the control cycle is selected as 15 minutes or 30 minutes. The specific duration is determined based on the sampling cycle of the data acquisition and monitoring system, and 100 to 500 continuous sampling points are selected.

[0031] The calculated mutual information values ​​are normalized to eliminate the influence of differences in the units of measurement between different data categories. Normalization can be performed using the maximum value normalization method. in, This represents the maximum mutual information value across all data categories. After normalization, this represents the information value of each category of running data. The value range is from zero to one.

[0032] Furthermore, the information value of the voltage deviation dimension and the load balancing dimension are weighted and summed according to preset weights to obtain the comprehensive information value: in, For the first The informational value of this type of data for voltage deviation targets. For the first The informational value of this type of data to the load balancing objective; and Let be the weighting coefficient, satisfying It can be adjusted according to the emphasis on voltage quality and load balancing in actual engineering projects.

[0033] In this embodiment, an information value threshold is set. The information value threshold can be set based on the distribution of the value of various data types in historical data, using the elbow rule or expert experience, and is typically set between 0.2 and 0.5. In this embodiment, it is set to 0.3. For operational data categories with a comprehensive information value lower than the information value threshold, they are marked as low-value redundant data and deleted; only high-value data categories with a comprehensive information value higher than or equal to the threshold are retained. These retained high-value data collectively constitute the dataset. The dataset has a significantly lower data dimensionality than the original data, but retains the core information most relevant to control decisions, which can significantly reduce the computational complexity of subsequent optimization solutions while ensuring the quality of the control strategy.

[0034] Step S2: Based on the dataset and the distribution network topology, identify the electrical coupling strength between nodes and divide them into clusters; using each cluster as a decision unit, solve the aggregation control model with the goal of maximizing the cluster voltage qualification rate and the load balance, and generate aggregation control instructions for each cluster. Based on the topology and node measurement information in the dataset generated in step S1, the distribution network is clustered according to electrical distance. Electrical distance is an important indicator for measuring the degree of electrical coupling between two nodes in a distribution network. In a distribution network, even if two nodes are geographically adjacent, if the electrical connection impedance between them is large, the mutual influence of their voltage and power is relatively weak. Therefore, using electrical distance rather than physical distance as the basis for clustering is more scientific.

[0035] The specific calculation process for electrical distance is as follows: Based on the current distribution network topology and line parameters, establish the node admittance matrix. The dimension of the node admittance matrix is ​​the total number of nodes in the distribution network. Its diagonal elements are the sum of the admittances of all branches connected to that node, and the off-diagonal elements are the negative values ​​of the branch admittances between two nodes, which directly reflects the electrical connection characteristics of the network.

[0036] By inverting the nodal admittance matrix, the nodal impedance matrix is ​​obtained, and then the voltage sensitivity matrix to reactive power is derived. in, Indicates at node When injecting a unit reactive power change, the node The change in voltage amplitude. This sensitivity matrix can be derived from the inverse of the nodal admittance matrix combined with the power flow Jacobian matrix. The larger the absolute value, the tighter the voltage-reactive coupling between the two nodes and the closer the electrical distance.

[0037] It should be noted that the voltage sensitivity matrix to reactive power can be calculated by combining the imaginary part of the node impedance matrix with the system reference voltage, or obtained by calculating the inverse of the Jacobian matrix from the power flow calculation. The specific derivation is well known in the field and will not be elaborated here.

[0038] Based on the elements of the sensitivity matrix, construct an electrical distance metric between nodes: in, The value ranges from zero to one, and the smaller the value, the tighter the electrical coupling between the two nodes.

[0039] After calculating the electrical distance for all node pairs, an electrical distance matrix is ​​formed. Based on this electrical distance matrix, a community detection algorithm is used to automatically cluster distribution network nodes.

[0040] In this embodiment, the Louvain algorithm is used to iteratively optimize the modularity index, enabling the automatic discovery of node clusters in a network that are internally tightly connected and externally sparsely connected, without requiring a pre-specified number of clusters. The formula for calculating modularity is: in, For nodes The weighting degree, The sum of weighted degrees across the entire network. For nodes The cluster number assigned to you. when The value is 1 if it is true; otherwise, the value is 0.

[0041] During the iteration process, the Louvain algorithm gradually maximizes the modularity by repeatedly executing two phases: node movement and community aggregation. until No further additions are made, and the final cluster partitioning result is obtained.

[0042] The Louvain algorithm increases modularity by continuously moving nodes between different clusters. The value continues to increase until convergence, ultimately yielding the cluster partitioning result. After clustering, all nodes of the distribution network are divided into... The system consists of several clusters, each with tightly coupled nodes and relatively weaker electrical coupling between clusters. Secondly, each cluster is treated as an independent aggregation decision-making unit, and an aggregation control model is constructed and solved. The design philosophy of the aggregation control model is to treat the cluster as an equivalent "virtual power source," determining the total power regulation target for each cluster from a global coordination perspective, rather than focusing on the individual allocation schemes of distributed resources within the cluster. The optimization objectives of the aggregation control model are to maximize the voltage compliance rate at the cluster's point of common coupling (PCC) and achieve optimal load balance between clusters. The voltage compliance rate is measured by the degree to which the PCC voltage deviates from the rated value; the load balance is measured by the difference in equivalent load rates between clusters. The two objectives are transformed into a comprehensive objective function using a weighted method: in, For clusters Voltage at the point of common coupling, Rated voltage, For clusters The equivalent load rate, The average load rate of the entire cluster. and These are the weighting coefficients. The weighting coefficients satisfy... Preferably, the voltage deviation weighting The value ranges from 0.4 to 0.8, representing the load balancing weight. The specific value is set by the operators based on the risk of voltage exceeding limits in the distribution network or equipment overload risk. The decision variable of the model is the total active power regulation of each cluster. Total reactive power regulation The constraints include: upper and lower limits for the voltage at the cluster's common junction point; thermal stability constraints for the power of the interconnect lines between clusters; and the total range of adjustable power for each cluster. Due to the number of clusters... The number of decision variables in the aggregation control model is significantly reduced compared to the total number of distributed resources, resulting in a substantial decrease in solution difficulty and time. The solution outputs the aggregation control commands for each cluster. .

[0043] Since the number of clusters is much smaller than the total number of distributed resources, the dimensionality of decision variables in the aggregation control model is significantly reduced, resulting in a substantial decrease in solution difficulty and time. It should be noted that the objective function of the aggregation control model includes the average load rate of all clusters, and the constraints include thermal stability constraints on the power of inter-cluster tie lines. These are all cross-cluster coupling variables; therefore, the aggregation control model is solved as a unified optimization problem, and each cluster cannot be completely independent at this stage. However, since the dimensionality of decision variables has been compressed from hundreds or thousands to the cluster level, the computational burden of the overall solution remains within an acceptable range. After completing the aggregation control instruction solution, in the intra-regional autonomous allocation stage of step S3, each cluster can independently execute its own aggregation control instructions in parallel. At this point, the optimization allocation problems within each cluster are decoupled, allowing full utilization of parallel computing resources to further improve overall efficiency.

[0044] This step is the first stage of a two-stage nested solution framework, which uses cluster-level aggregation to lead the solution in order to solve the dimensionality curse problem faced by large-scale global optimization.

[0045] Step S3: The aggregation control command is issued as a constraint to each cluster; within each cluster, the active and / or reactive power output allocation values ​​of each distributed resource are solved with the minimum adjustment cost of the distributed resources as the sub-objective. First, the cluster aggregation control commands generated in step S2 are used as equality constraints and distributed to the corresponding allocation calculation modules of each cluster through the scheduling communication network. Within each cluster, an optimal allocation model is constructed with the goal of minimizing the adjustment cost of distributed resources within the cluster. The adjustment cost reflects the economic cost and scheduling priority of different distributed resources participating in power regulation. The optimal allocation scheme should, under the premise of satisfying the constraints of the cluster aggregation control commands, call upon distributed resources with lower adjustment costs as much as possible.

[0046] The adjustment cost of each distributed resource is determined based on a preset weight coefficient for its resource type. The weight coefficient reflects the scheduling priority of different resource types. In this embodiment, the adjustment weight coefficient for controllable loads is set relatively low because power adjustment of interruptible or shiftable loads has minimal impact on users and is relatively inexpensive, making them suitable for priority use. The adjustment weight coefficient for energy storage devices is next, as while energy storage systems offer fast response and flexible adjustment, energy conversion losses and battery cycle life degradation occur during charging and discharging. The adjustment weight coefficient for photovoltaic inverters is set relatively high because reducing photovoltaic output directly leads to the waste of clean energy and economic losses. If priority is needed to ensure the consumption of green energy, the adjustment weight coefficient for photovoltaic inverters can be set to the highest level to minimize the reduction in photovoltaic output during optimization.

[0047] The mathematical expression of the cluster internal optimal allocation model is as follows: The constraints include: (1) Power balance equation constraint (aggregate control command constraint): (2) Upper and lower limits of active power output of each distributed resource: (3) Upper and lower limits of reactive power output of each distributed resource: (4) Apparent power constraint of inverter capacity: in, For resources Type weighting coefficients, and Resources The adjustment of active and reactive power output, This is the conversion factor for reactive power regulation costs relative to active power regulation costs; and Resources The current contribution and contribution without merit, ; The upper and lower limits are defined by merit and effort. ; The upper and lower limits of no-effort output. This is the rated capacity of the inverter. The above constraints ensure the physical feasibility of the allocation scheme and the safety of the equipment.

[0048] It should be noted that the autonomous allocation step within the zone can employ a centralized computing scheme, where the cluster controller uniformly solves the problem and issues allocation instructions; alternatively, it can be implemented using a distributed algorithm based on a consensus protocol. In the distributed algorithm scheme, the distributed resource agent nodes within the cluster constitute a distributed computing network. Each agent node communicates only with its neighboring nodes, exchanging its own marginal adjustment cost information. The marginal adjustment cost refers to the cost change caused by a slight adjustment of one unit in resource output. Through multiple rounds of iterative information exchange and local optimization calculations, the marginal adjustment costs of each agent node gradually converge, at which point the optimal allocation state that satisfies the constraints of the cluster aggregation control instructions is reached.

[0049] This step is the second stage in a two-stage nested solution framework, involving autonomous allocation within resource-level zones. After determining the aggregation and control objectives for each cluster in the first stage, this step is responsible for specifically decomposing the cluster-level aggregation and control instructions to the various distributed resources within the cluster.

[0050] Step S4: Summarize the output allocation values ​​of each distributed resource in each cluster, generate and output the collaborative control strategy of the target output curve of each distributed resource.

[0051] The output allocation values ​​of all distributed resources in the distribution network obtained from step S3 are summarized. Information such as the identifier of each distributed resource, its cluster number, target active power output value, target reactive power output value, and corresponding start and end times are arranged according to a preset data format to form a coordinated control strategy. The control strategy file can use a structured text format, such as JSON or XML, to facilitate parsing by the dispatch control system and execution by the terminal controller. The content of the strategy file is the target output curve of each distributed resource, that is, the sequence of active and reactive power output setpoints that each distributed resource should achieve at each control time point within a complete control cycle.

[0052] Finally, the generated executable control strategy file is sent to the corresponding terminal controllers one by one through the distribution network dispatch communication network, based on the communication addresses of each distributed resource. After receiving the instructions, the terminal controllers adjust the output of the controlled resources according to the instructions. Throughout the entire process of strategy execution, the dispatch control system continuously collects real-time measurement data, monitors the execution effect, and forms a complete control system.

[0053] Specifically, in a preferred embodiment of this application, the step of minimizing the adjustment cost of distributed resources and solving for the active and / or reactive power output allocation values ​​of each distributed resource is achieved by employing a distributed algorithm based on a consensus protocol. Please refer to [link to relevant documentation]. Figure 2 Specifically, it includes: Step S301: Using the aggregation control command as a power balance equation constraint, obtain the current output value and type weight coefficient of each distributed resource in the cluster; Each cluster receives the aggregated control command generated in step S2, and uses the total active power adjustment and total reactive power adjustment of the cluster in the aggregated control command as an equality constraint that the cluster must satisfy. That is, the sum of the output adjustments of all resources within the cluster must be exactly equal to the aggregated control command of that cluster. At the same time, each agent node needs to obtain its current active power output, current reactive power output, and weighting coefficients determined by the resource type. This information is the basic data for calculating the marginal adjustment cost.

[0054] Step S302: Each distributed resource in the cluster is configured with a proxy node, and each proxy node calculates the initial marginal adjustment cost based on the type weight coefficient and the current output level. Marginal adjustment cost is the change in adjustment cost caused by a slight adjustment of one unit in the output of distributed resources. Its mathematical expression is: The adjustment cost function is: The initial marginal adjustment cost of each proxy node is determined by its own type weight coefficient. (Reflecting the scheduling priority of the resource) and calculated. In this embodiment, the adjustment cost weight coefficients of different types of resources are ranked as follows: adjustment weight coefficient of controllable load < adjustment weight coefficient of energy storage device < adjustment weight coefficient of photovoltaic inverter.

[0055] Step S303: Each agent node communicates with its neighboring nodes, taking the minimization of the adjustment cost as the convergence direction, and iteratively updates its own marginal adjustment cost estimate and corresponding power allocation value by exchanging marginal adjustment cost information. The power allocation value satisfies the power balance equation constraint and individual operation constraint. Each agent node communicates only with its neighboring agent nodes and receives marginal adjustment cost estimates sent by its neighboring nodes. At the same time, it estimates its own marginal adjustment cost. Send to neighboring nodes. Each proxy node updates its own marginal adjustment cost estimate according to the following consensus protocol update rules: in, Let step size be the iteration step size. Satisfy 0 < <1 / , For nodes The set of adjacent nodes, This represents the communication weighting coefficient between adjacent nodes.

[0056] In a distributed algorithm based on a consensus protocol, each agent node achieves global consensus by communicating with its neighboring nodes to bring their marginal adjustment costs into a consistent manner, through local information exchange. Based on the updated estimated marginal adjustment costs, each agent node adjusts its active and reactive power output allocation values, provided that constraints are met.

[0057] The constraints include: (1) Power balance equation constraint (aggregate control command constraint): (2) Upper and lower limits of active power output of each distributed resource: (3) Upper and lower limits of reactive power output of each distributed resource: (4) Apparent power constraint of inverter capacity: Step S304: Stop the iteration when the estimated marginal adjustment cost of each agent node tends to be consistent, and output the current output allocation value as the output allocation value.

[0058] After multiple iterations, the estimated marginal adjustment costs of all agent nodes gradually converge (i.e., At this point, the optimal allocation state is reached, satisfying the constraints of the cluster aggregation control command. In engineering practice, a preset convergence threshold can be set. ,when Stop iterating when the time comes.

[0059] At this point, the current output allocation value of each agent node is the optimal output allocation scheme that minimizes the cluster adjustment cost. This is output as the solution result of step S3 to step S4 for strategy summarization.

[0060] Through a marginal adjustment cost consistency iteration mechanism, each agent node communicates and exchanges marginal adjustment cost information only with its neighboring nodes, achieving autonomous optimization allocation of resources within the cluster without a central node. This achieves cost-optimal allocation with minimal adjustment costs while satisfying cluster aggregation control commands and various physical constraints.

[0061] Specifically, in a preferred embodiment of this application, please refer to... Figure 3 The collaborative control strategy also includes triggering correction conditions; When the deviation between the voltage of a critical node in the distribution network monitored in real time and the expected voltage value of the main strategy exceeds the preset dead zone threshold, and / or the deviation between the actual load rate of a critical line and the expected load rate of the main strategy exceeds the preset dead zone threshold, the strategy correction procedure is triggered; the critical node refers to a voltage-weak node or a feeder end node, which can be pre-determined based on voltage sensitivity ranking or historical voltage over-limit frequency; the critical line refers to a line with a small rated capacity or a long-term high load rate, which is selected based on the line's thermal stability limit and historical load data.

[0062] Specifically, the triggering correction condition includes at least one of the following: When the absolute value of the deviation between the voltage amplitude of a key node monitored in real time by the distribution network dispatch and control system and the expected voltage value of that node in the main strategy exceeds a preset voltage deviation dead zone threshold, the strategy correction procedure is triggered. The key node refers to a hub node in the distribution network that has a significant impact on voltage stability or a sensitive node located at the end of a feeder. These nodes can be manually designated in advance based on the network topology and historical operating data, or automatically selected and determined through a voltage vulnerability assessment algorithm. The voltage deviation dead zone threshold ranges from 1% to 5% of the rated voltage, and the specific value can be adjusted according to the voltage level of the distribution network and actual operating requirements. For example, for a 10kV distribution network, it can be set to 2% of the rated voltage (i.e., 0.2kV); for a 400V low-voltage distribution network, it can be set to 3% of the rated voltage (i.e., 12V). The dead zone threshold is set to ensure that correction is only initiated when the deviation has a substantial impact, avoiding frequent activation of the correction procedure due to measurement noise or minor load fluctuations.

[0063] When the absolute value of the deviation between the actual load rate of a critical line monitored in real time by the distribution network dispatch control system and the expected load rate of that line in the main strategy exceeds a preset load rate deviation dead zone threshold, the strategy correction procedure is triggered. The critical line refers to an important transmission corridor with a high load rate or a significant impact on power supply reliability, and can be predetermined based on the line's rated capacity and historical load levels. The load rate deviation dead zone threshold ranges from 5% to 15%, and the specific value can be adjusted according to the line's rated capacity and actual operating requirements. For example, for a line with a rated capacity of 10MVA, it can be set to 10%, meaning that correction is triggered when the deviation between the actual load rate and the expected load rate exceeds 10 percentage points.

[0064] The voltage deviation dead zone threshold and the load rate deviation dead zone threshold can be set independently or combined using an "OR" logic, meaning that the correction procedure is triggered when either condition is met. Preferably, when both conditions are met simultaneously, the correction is executed first according to the voltage deviation trigger condition, because voltage exceeding the limit poses a more direct threat to the safe operation of the distribution network.

[0065] The correction procedure maintains the aggregation control instructions of each cluster unchanged, re-executes the autonomous allocation process within each cluster, calculates the incremental correction amount of the output of each distributed resource, and generates additional adjustment instructions to be superimposed on the currently executed control instructions.

[0066] Specifically, once the correction procedure is triggered, each cluster will use its current real-time measurement status (including the current actual output value of each distributed resource) to determine its status. ; Using the current actual voltage value of each node as the new starting point, and with the aggregated control command as the power balance equation constraint, a distributed iterative solution based on the consistency of marginal adjustment costs is re-executed within each cluster. This re-solution only calculates the incremental correction amount of each distributed resource relative to the current actual output. , Instead of recalculating the full output value, the iterative initial value is closer to the actual running point, and the aggregate amount in the constraints remains unchanged. Therefore, the convergence speed of the corrected procedure is significantly faster than the initial solution in step S3. The calculation of the incremental correction amount can be mathematically expressed as: The constraints are: This also includes the individual operational constraints. Since the total aggregate adjustment on the right-hand side of the equation is zero, the essence of the corrective allocation is to rebalance and adjust the output of each resource within the cluster, without changing the total power characteristics presented by the cluster to the outside world.

[0067] Each agent node generates an additional adjustment instruction based on the incremental correction amount. The additional adjustment instruction serves as a supplement to the main strategy and is superimposed on the currently executing control instruction to form a corrected execution instruction.

[0068] It should be noted that during the execution of the correction procedure, if the calculated incremental correction exceeds the sum of the adjustable margins of all distributed resources within the cluster, causing the correction procedure to fail to converge under the premise of satisfying all constraints, then it is determined that the current deviation exceeds the correction capability range. In this case, instead of performing partial correction, a full-process re-optimization is triggered—that is, returning to step S1 to re-execute the data value screening; the entire process of partition aggregation solution and intra-region autonomous allocation, and regenerating the complete collaborative control strategy with the current real-time state as the initial state.

[0069] This preferred embodiment constructs trigger correction conditions that include dead-zone thresholds for voltage deviation and load rate deviation. When the deviation exceeds the tolerance range, the correction procedure is initiated, avoiding unnecessary global recalculations due to frequent minor random fluctuations and effectively reducing the computational burden on the control system. The correction procedure maintains the aggregated control commands for each cluster unchanged, only autonomously allocating the incremental correction amount within the re-execution zone of each cluster. This fully utilizes the natural decoupling characteristics of the two-stage solution framework, achieving hierarchical adaptive control that maintains global precision in coarse adjustments and adapts to local conditions in fine adjustments. This significantly improves the accuracy and dynamic control of the control strategy in response to load fluctuations and equipment response deviations.

[0070] Please see Figure 4 This invention also provides a distributed resource control policy generation system, which applies the aforementioned method and includes: The data value filtering module is used to acquire multi-source operation data of the distribution network, calculate the information value of various types of operation data to preset control targets, and form a dataset based on the information value. The partitioned aggregation solution module is used to identify the electrical coupling strength between nodes based on the dataset and the distribution network topology, and to divide the nodes into clusters accordingly. It also uses each cluster as a decision unit to solve the aggregation control model with the goal of achieving the highest cluster voltage qualification rate and the best load balance, and generates aggregation control instructions for each cluster. The autonomous allocation module within the area is used to issue the aggregated control command as a constraint to each cluster, and within each cluster, with the minimum adjustment cost of distributed resources as the sub-objective, solve for the active and / or reactive power output allocation values ​​of each distributed resource. The strategy generation and output module is used to summarize the output allocation values ​​of each distributed resource in each cluster, generate a collaborative control strategy containing the target output curves of each distributed resource, and output it.

[0071] The functional explanations of each module in this embodiment are the same as those of a distributed resource control strategy generation method, and the technical effects are the same, so they will not be repeated here.

[0072] Those skilled in the art will recognize that the units of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0073] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0074] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0075] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drive; read-only memory (ROM); random access memory (RAM); portable hard disk; magnetic disk or optical disk, and other media capable of storing program code.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the specification of the present invention.

Claims

1. A method for generating distributed resource control strategies, characterized in that, Includes the following steps: Acquire multi-source operation data of the power distribution network, calculate the information value of various types of operation data to the preset control target, and remove data categories whose information value is lower than the preset threshold to form a dataset; Based on the dataset and distribution network topology, the electrical coupling strength between nodes is identified and clusters are divided. Taking each cluster as a decision unit, the aggregation control model with the goal of maximizing the cluster voltage qualification rate and the optimal load balance is solved, and aggregation control instructions for each cluster are generated. The aggregated control command is issued as a constraint to each cluster; within each cluster, the active and / or reactive power output allocation values ​​of each distributed resource are solved with the minimum adjustment cost of the distributed resources as the sub-objective. Summarize the output allocation values ​​of each distributed resource within each cluster, generate and output the collaborative control strategy of the target output curve of each distributed resource.

2. The method for generating a distributed resource control strategy according to claim 1, characterized in that, The information value is calculated based on mutual information theory, and specifically includes: For each type of operational data and each control target indicator, calculate the mutual information value; The mutual information values ​​are normalized to obtain the information value of each type of operational data.

3. The method for generating a distributed resource control strategy according to claim 2, characterized in that, The formula for calculating the mutual information value is: in, Indicates the first Time series of class running data, Represents the time series of regulatory target indicators; The joint probability distribution of the two types of data. and Each of these represents a marginal probability distribution, wherein both the joint probability distribution and the marginal probability distribution are estimated from the time series using a kernel density estimation method.

4. The method for generating a distributed resource control strategy according to claim 2, characterized in that, The calculation of the information value of various types of operational data to the preset control target specifically includes: The preset control objectives include the voltage deviation minimization objective and the load balance optimization objective; the information value of each type of operating data in the voltage deviation dimension and the load balance dimension are calculated respectively, and the information value of the voltage deviation dimension and the load balance dimension are weighted and summed according to preset weight coefficients to obtain the comprehensive information value of each type of operating data.

5. The method for generating a distributed resource control strategy according to claim 1, characterized in that, The process of identifying the electrical coupling strength between nodes and dividing the cluster includes: Based on the network topology and line parameters of the distribution network, establish the node admittance matrix; Inverting the node admittance matrix yields the node impedance matrix, which in turn allows us to derive the voltage sensitivity matrix to reactive power. An electrical distance metric between nodes is constructed based on the elements of the sensitivity matrix. Based on the electrical distance matrix, the distribution network nodes are automatically clustered to obtain the cluster partitioning results.

6. The method for generating a distributed resource control strategy according to claim 1, characterized in that, The optimization objective of the aggregation control model is: in, For clusters Voltage at the point of common coupling, Rated voltage, For clusters The equivalent load rate, The average load rate of the entire cluster. and These are the weighting coefficients; The decision variables of the aggregation control model are the total active power regulation and the total reactive power regulation of each cluster. The constraints include the upper and lower limits of the voltage at the common coupling point of the clusters; the thermal stability constraints of the power of the interconnect lines between clusters; and the total range constraints of the adjustable power of each cluster.

7. The method for generating a distributed resource control strategy according to claim 1, characterized in that, The adjustment cost of each distributed resource is determined based on the weighting coefficient preset according to the type of distributed resource.

8. The method for generating a distributed resource control strategy according to claim 1, characterized in that, The sub-objective of minimizing the adjustment cost of distributed resources, specifically includes solving for the active and / or reactive power output allocation values ​​of each distributed resource: Using the aggregation control command as a power balance equation constraint, the current output value and type weight coefficient of each distributed resource in the cluster are obtained; Each distributed resource in the cluster is configured with a proxy node, and each proxy node calculates the initial marginal adjustment cost based on the type weight coefficient and the current output level. Each agent node communicates with its neighboring nodes, taking the minimization of the adjustment cost as the convergence direction, and iteratively updates its own marginal adjustment cost estimate and corresponding power allocation value by exchanging marginal adjustment cost information. The power allocation value satisfies the power balance equation constraint and individual operation constraint. The iteration stops when the estimated marginal adjustment costs of each agent node become consistent, and the current output allocation value is output as the output allocation value.

9. A method for generating a distributed resource control strategy according to any one of claims 1-8, characterized in that, The collaborative control strategy also includes trigger correction conditions; When the deviation between the voltage of the critical node of the distribution network monitored in real time and the expected voltage value of the main strategy exceeds the preset dead zone threshold, and / or the deviation between the actual load rate of the critical line and the expected load rate of the main strategy exceeds the preset dead zone threshold, the strategy correction procedure is triggered. The correction procedure maintains the aggregation control instructions of each cluster unchanged, re-executes the autonomous allocation process within each cluster, calculates the incremental correction amount of the output of each distributed resource, and generates additional adjustment instructions to be superimposed on the currently executed control instructions.

10. A distributed resource control strategy generation system, characterized in that, The method described by any one of claims 1-9 comprises: The data value filtering module is used to acquire multi-source operation data of the distribution network, calculate the information value of various types of operation data to preset control targets, and form a dataset based on the information value. The partitioned aggregation solution module is used to identify the electrical coupling strength between nodes based on the dataset and the distribution network topology, and to divide the nodes into clusters accordingly. It also uses each cluster as a decision unit to solve the aggregation control model with the goal of achieving the highest cluster voltage qualification rate and the best load balance, and generates aggregation control instructions for each cluster. The autonomous allocation module within the area is used to issue the aggregated control command as a constraint to each cluster, and within each cluster, with the minimum adjustment cost of distributed resources as the sub-objective, solve for the active and / or reactive power output allocation values ​​of each distributed resource. The strategy generation and output module is used to summarize the output allocation values ​​of each distributed resource in each cluster, generate a collaborative control strategy containing the target output curves of each distributed resource, and output it.

Images