Feeder load and distributed flexible resource cooperative direct control method and system

Abstract

The invention provides a feeder load and distributed flexible resource cooperative direct control method and system, and relates to the technical field of power system power distribution network automatic control. The method comprises the following steps: carrying out adjustment characteristic modeling on multi-type distributed resources; calculating regulation and control performance indexes of all types of distributed resources, and configuring index weights by adopting an analytic hierarchy process; performing clustering aggregation on the distributed resources based on index weights and a clustering algorithm, and constructing a two-stage control architecture of a virtual power plant main control layer and a resource cluster inner control layer; a voltage sensitivity matrix based on a linearized power flow model is introduced, a cooperative control optimization model considering a feeder load and distributed resource coupling effect is constructed, and revenue maximization is realized. According to the invention, high-frequency, large-amplitude and low-cost flexible load regulation and control of the power distribution network are realized, and the economical efficiency and responsiveness of the virtual power plant in orderly power utilization and frequency modulation auxiliary services are improved.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for power distribution networks, specifically to a method and system for coordinated direct control of feeder loads and distributed flexible resources in distribution networks. Background Technology

[0002] Currently, with the continuous expansion of smart grid and new energy access scale, the large-scale access of distributed power sources (such as photovoltaic and wind power) and various types of adjustable loads (such as electric vehicles and energy storage devices) in the distribution network has led to the power system exhibiting characteristics of high source-load coupling, enhanced volatility, and significantly increased control complexity.

[0003] Traditional centralized control center-based unified scheduling strategies or regulation mechanisms relying on single controllable resources are limited by control lag, untimely response, and low resource utilization efficiency. These limitations make them insufficient to meet the rapid response and high-precision control requirements of current power distribution systems for peak shaving, frequency regulation, load migration, and orderly power consumption. In actual operation, feeder load power is highly sensitive to node voltage, while the operating status of distributed resources in the power distribution system (such as photovoltaic power generation and energy storage charging and discharging behavior) directly or indirectly causes local voltage fluctuations, thus affecting the real-time response of feeder loads. A significant coupling and interaction exists between these two factors. Without precise modeling and coordinated control during the control process, local control conflicts, system oscillations, and even unnecessary energy waste and power quality degradation can easily occur.

[0004] Currently, the distribution network lacks a collaborative control methodology capable of dynamically modeling and identifying the response characteristics of feeder loads, while simultaneously integrating the characteristic constraints of multiple types of distributed resources to construct a collaborative control system with hierarchical decoupling capabilities and a control priority allocation mechanism. Especially at the real-time scale, how to tap the adjustment potential of feeder loads and coordinate responses with flexible resources while ensuring system operational safety is crucial. Therefore, a novel collaborative direct control technology is urgently needed that can integrate the characteristics of feeder loads and multiple types of distributed flexible loads, establish a unified control framework, and achieve real-time response and dynamic optimization. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for coordinated direct control of feeder loads and distributed flexible resources, so as to achieve the goal of fast, accurate and low-cost intelligent regulation and control, which is applicable to practical scenarios such as virtual power plants and automatic power control; it realizes flexible load regulation of distribution networks with high frequency, large amplitude and low cost, and improves the economy and responsiveness of virtual power plants in orderly power consumption and frequency regulation ancillary services.

[0006] To achieve the above objectives, in a first aspect, the present invention provides a method for coordinated direct control of feeder loads and distributed flexible resources, comprising: The regulation characteristics of various types of distributed resources are modeled, and regulation characteristic models and cost models are constructed for feeder loads, distributed photovoltaics, energy storage and electric vehicles. Calculate the regulation performance indicators for various types of distributed resources, including controllable capacity, regulation accuracy, and unit regulation cost, and use the analytic hierarchy process (AHP) to configure the indicator weights. Based on index weights and clustering algorithms, distributed resources are grouped and aggregated to construct a two-level control architecture of virtual power plant main control layer and resource cluster internal control layer. By introducing a voltage sensitivity matrix based on a linearized power flow model, a collaborative control optimization model considering the coupling effect of feeder load and distributed resources is constructed to maximize benefits.

[0007] According to the present invention, a method for coordinated direct control of feeder load and distributed flexible resources is provided. The regulation characteristic model of the feeder load is constructed based on the ZIP load model and CVR coefficient, and the cost model of the feeder load is constructed based on the voltage level and real-time load power. According to the present invention, a method for coordinated direct control of feeder loads and distributed flexible resources is provided. The regulation characteristic model of distributed photovoltaic (PV) is constructed based on the active power operating constraints of PV; the regulation characteristic model of energy storage is constructed based on the operating power constraints, energy constraints, and dynamic characteristics of energy storage; and the regulation characteristic model of electric vehicles is constructed based on the operating power constraints, energy constraints, dynamic characteristics, and user electricity demand of electric vehicles. The cost models for distributed PV, energy storage, and electric vehicles are all fixed compensation, and are constants. , and . According to the method for coordinated direct control of feeder load and distributed flexible resources provided by the present invention, the calculation of the control performance index includes: The adjustable capacity of distributed photovoltaic power is 10% of its current operating power. The adjustable capacity of energy storage is calculated based on its state of charge, charge / discharge power, and charge / discharge efficiency. The adjustable capacity of electric vehicles is calculated based on user electricity demand and battery state of charge constraints. According to the method for coordinated direct control of feeder load and distributed flexible resources provided by the present invention, the control accuracy of distributed resources is: (2-11) in, and These represent the actual operating power and reference operating power of the distributed resource during a certain historical period, respectively. The allowable error for adjusting this distributed resource; This represents the total number of historical periods involved in the statistical calculations.

[0008] According to the present invention, a method for coordinated direct control of feeder load and distributed flexible resources is provided, which uses the analytic hierarchy process (AHP) to configure index weights, including: The target layer is the regulation performance of distributed resources in the distribution network, the criterion layer is the technical and economic efficiency, and the indicator layer is based on the controllable capacity, regulation accuracy and unit regulation cost. Construct an indicator importance judgment matrix: (2-12) Where A represents the indicator importance judgment matrix, The matrix A represents the comparison result of the importance between indicator i and indicator j, where n represents the number of indicators; the elements of the indicator importance judgment matrix A satisfy... ,like i = j ,but ; Calculate the nth root of the product of indicators in each row of the indicator importance judgment matrix, and then normalize the result to obtain the indicator weights. and the eigenvector W composed of the weights of each indicator; Calculate the largest eigenvalue of the importance judgment matrix. Then calculate the consistency index. And combine the order of the random judgment matrix and the random consistency index Calculate the consistency ratio The calculation formula is: (2-15) (2-16) (2-17) If the calculated consistency ratio If the value is less than or equal to the preset threshold, the importance judgment matrix A of the indicator is deemed reasonable, and the weight of the indicator is determined. According to the present invention, a method for coordinated direct control of feeder loads and distributed flexible resources is provided. Based on index weights and clustering algorithms, distributed resources are grouped and aggregated to construct a two-level control architecture consisting of a virtual power plant main control layer and a resource cluster internal control layer. The method includes: An improved K-means++ clustering algorithm is used for resource clustering, including index normalization, selection of the optimal number of clusters, initialization of cluster centers, Euclidean distance calculation considering index weights, sample clustering and iterative update of cluster centers, until the location of the cluster centers is determined to complete the clustering and aggregation. According to the present invention, a method for coordinated direct control of feeder load and distributed flexible resources is provided, comprising a two-level control architecture including: The virtual power plant's main control layer is responsible for receiving automatic power control commands and distributing them to various resource aggregates. The resource cluster internal control layer completes the automatic power control command splitting and minimum cost allocation based on the cluster's internal resource constraints. According to the present invention, a method for coordinated direct control of feeder load and distributed flexible resources is provided. The voltage sensitivity matrix is ​​derived based on a linearized power flow model and is used to characterize the relationship between node power injection and voltage response. It is also updated in real time to provide feedback on the control effect. The objective function of the collaborative control optimization model is to maximize the net revenue of the virtual power plant. The constraints include regional tie-line power constraints, node voltage constraints, distributed resource cluster operation constraints, static var compensator operation constraints, and voltage regulation and load power regulation constraints. Secondly, the present invention provides a system for coordinated direct control of feeder load and distributed flexible resources, comprising: The module is used to model the regulation characteristics of various types of distributed resources, including regulation characteristic models and cost models for feeder loads, distributed photovoltaics, energy storage, and electric vehicles. The calculation module is used to calculate the control performance indicators of various types of distributed resources, including controllable capacity, control accuracy and unit control cost, and uses the analytic hierarchy process to configure the indicator weights. The control module is used to group and aggregate distributed resources based on index weights and clustering algorithms, and to build a two-level control architecture of the virtual power plant main control layer and the resource cluster internal control layer. The coordination module is used to introduce a voltage sensitivity matrix based on a linearized power flow model to construct a coordinated control optimization model that considers the coupling effect between feeder load and distributed resources, thereby maximizing benefits.

[0009] This invention has at least the following technical effects: This invention provides a method and system for coordinated direct control of feeder loads and distributed flexible resources. The method includes: modeling the regulation characteristics of multiple types of distributed resources, constructing regulation characteristic models and cost models for feeder loads, distributed photovoltaics, energy storage, and electric vehicles; calculating the regulation performance indicators of each type of distributed resource and configuring indicator weights using the analytic hierarchy process (AHP); clustering and aggregating distributed resources based on indicator weights and a clustering algorithm to construct a two-level control architecture consisting of a virtual power plant main control layer and a resource cluster internal control layer; and introducing a voltage sensitivity matrix based on a linearized power flow model to construct a coordinated control optimization model considering the coupling effect between feeder loads and distributed resources, thereby maximizing revenue. This invention achieves high-frequency, large-amplitude, and low-cost flexible load regulation in distribution networks, improving the economy and responsiveness of virtual power plants in orderly power consumption and frequency regulation ancillary services. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0011] In the attached diagram: Figure 1a This is a schematic diagram of the feasible operating domain of the single electric vehicle power according to the present invention; Figure 1b This is a schematic diagram of the feasible energy operation domain for a single electric vehicle according to the present invention. Figure 2 This is a schematic diagram of the distributed resource regulation performance index system of the present invention; Figure 3 This is a flowchart of the K-means++ algorithm based on index weighting, as described in this invention. Figure 4 This is a flowchart of the method for coordinated direct control of feeder load and distributed flexible resources according to the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this 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 this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0013] The following detailed description of some embodiments of the present invention will be provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0014] Please see Figure 4 This invention provides a method for coordinated direct control of feeder loads and distributed flexible resources. It is a method applicable to the joint modeling, coordinated control, and optimized scheduling of feeder loads and various types of distributed flexible resources (such as distributed photovoltaics, energy storage, and electric vehicles) in distribution networks. It can be applied to scenarios such as automatic power control, frequency regulation, and power quality management in distribution networks, and includes the following steps: Step 1: Modeling the Regulation Characteristics of Multiple Types of Distributed Resources. Model the regulation characteristics of multiple types of distributed resources, constructing regulation characteristic models and cost models for feeder loads, distributed photovoltaics, energy storage, and electric vehicles.

[0015] Specifically, a unified regulation characteristic model is constructed for various types of distributed resources, such as photovoltaics, energy storage, and electric vehicles. The constraints include maximum output, state of charge and discharge, state of charge (SOC), and user-expected charging time. All distributed resources are treated as controllable power sources (constant power model) to achieve horizontal alignment of resources. A regulation cost model consistent with actual operation is constructed, incorporating power quality compensation and electricity price sensitivity into the regulation costs.

[0016] In some embodiments, step 1 includes the following specific content: Step 1.1: Feeder Load Regulation Characteristics and Cost Model Different types of loads and equipment in a power distribution network generally exhibit the characteristic that the active power of the load is sensitive to voltage changes. Feeder load power control is a direct power control technology based on the active power-voltage coupling characteristic. It adjusts the feeder voltage through parallel reactive power compensation equipment or series voltage regulating devices, thereby affecting the active power of the feeder load. Therefore, the feeder load power is a function of voltage, and its steady-state characteristics can be represented using the ZIP load model, as shown in equations (1-1)-(1-3). , and These are the proportionality coefficients for the active components of constant impedance loads, constant current loads, and constant power loads, respectively. b Z , b I and b P These are the proportionality coefficients for the reactive components of constant impedance load, constant current load, and constant power load, respectively.

[0017] (1-1) (1-2) (1-3) In the formula, and They are nodes i Active and reactive loads, and These are the initial values ​​for active load and reactive load, respectively; and These are the actual node voltage and the rated voltage, respectively.

[0018] The steady-state regulation characteristics of feeder loads are currently often described based on step-down energy-saving technologies, and the percentage change in feeder load power is expressed as a percentage. Percentage of voltage change The ratio is defined as the CVR coefficient, which characterizes the sensitivity of the load's active power to voltage changes, as shown in equation (1-4): (1-4) Furthermore, the CVR coefficient has a certain quantitative relationship with the model parameters of the ZIP load model, as shown in equation (1-5). Therefore, the CVR coefficient of the feeder can be quickly calculated based on the feeder load model, and the feeder load regulation characteristic relationship can be established.

[0019] (1-5) On the other hand, to facilitate the coordinated control of multiple feeder loads and distributed resources by the virtual power plant, a corresponding cost model needs to be established to quantify the regulation costs of feeder load power control. The characteristics and principles of feeder load power control inevitably affect the power quality and electricity demand on the user side. The standard "Economic Assessment of Power Quality Part 1: Methods for Economic Assessment of Electricity Users" (GB / Z 32880.1-2016) stipulates that the party responsible for power quality problems needs to provide corresponding compensation. Therefore, this invention, from the perspective of power quality, determines the power quality level based on voltage deviation. L {1 (Good), 2 (Average), 3 (Poor)}, and provide economic compensation to load users whose power quality level has declined based on real-time electricity prices, as a kind of incentive behavior for the load side.

[0020] According to the national standard "Power Quality - Supply Voltage Deviation" (GB / T 12325-2008), voltage deviation is defined as the relative value of the deviation between the actual operating voltage of a node and the nominal voltage of the system. Considering the basic requirements of the distribution network for voltage level (deviation within ±7%), a piecewise power quality function is set, and the specific segmentation rules are as follows: (1-6) Among them, D V This indicates the deviation of the power supply voltage in the distribution network.

[0021] Based on this, and according to the node power quality level after the control is implemented, the unit power compensation coefficient for providing economic compensation to users in each time period is determined. The feeder load control cost model is modeled as a function of voltage level and real-time load power, thus obtaining the feeder load power control cost model shown in equation (1-7): (1-7) in, This indicates the cost of feeder load power control. Indicates the feeder load power. This indicates the time interval during which the feeder load is applied.

[0022] Step 1.2: Adjustment characteristics and cost model of distributed resources This invention considers distributed resources involved in constructing a virtual power plant, including distributed photovoltaics, energy storage, and electric vehicles, and treats all three as controllable power sources. This means that power control is achieved by providing a reference power, corresponding to the constant power load component in the ZIP load model. Specifically, this step analyzes the regulation characteristics and methods of the three distributed resources based on their operational constraints.

[0023] Distributed photovoltaic (PV) regulation characteristics. "Current curtailment" is a fundamental measure to address PV fluctuations. Currently, PV inverters mainly include two types: PQ control and PV control. Both can regulate active power output by issuing active power control reference values, thereby meeting the active power regulation needs of the distribution network.

[0024] In real-world environments, distributed photovoltaic (PV) systems typically operate using Maximum Power Point Tracking (MPPT), thus they are considered to only possess active power curtailment capabilities. This invention primarily considers their active power operation constraints to limit the amount of curtailed solar power in different time periods. : (1-8) (1-9) In the formula, For the first t Time period node i The planned output value of photovoltaic power generation For the first t Time period node i The predicted output value of the photovoltaic system.

[0025] Energy storage regulation characteristics. This invention only considers electrochemical energy storage as a controllable distributed resource. The virtual power plant, during its decision-making process, considers its day-ahead charge and discharge plans to determine its optimal quantity and price bidding strategy. Since the number of distributed energy storage units connected to each node is fixed after the energy storage is put into operation, for the sake of simplicity and to facilitate the virtual power plant's day-ahead quantity and price bidding decisions, this invention describes its regulation characteristics on a node-by-node basis. Specifically, it can be characterized using operating power, energy constraints, and dynamic characteristics, as shown in the following expressions: (1-10) (1-11) (1-12) (1-13) (1-14) In the formula, and The first tTime period node i The charging and discharging flags of the energy storage are set to 1. When they are set to 1, it means that the energy storage is in the corresponding working state. The energy storage on the same node is only allowed to be in one of the working states. and These are the charging power and discharging power of the energy storage, respectively. and These are its maximum charging power and maximum discharging power, respectively. For the first t Time period node i The energy stored in the battery. and These are their charging and discharging efficiencies, respectively. and They are nodes i The maximum and minimum allowable capacity of the energy storage.

[0026] Electric vehicle regulation characteristics. For a single electric vehicle, its physical regulation characteristics are the same as those of electrochemical energy storage, both needing to meet operating power constraints, energy constraints, and dynamic characteristics. However, as a user load resource, electric vehicles, compared to electrochemical energy storage, also need to further consider the user's electricity demand, i.e., the off-grid conditions. To achieve the user's desired energy state As shown in equation (1-15): (1-15) Furthermore, it can be achieved through, for example Figure 1a and Figure 1b The power and energy operating feasible regions shown intuitively characterize their regulation characteristics and adjustability.

[0027] Finally, as mentioned above, the virtual power plant needs to conduct technical consultations and compensation agreements with users beforehand to obtain control over resources, equipment, and devices. In this regard, the present invention assumes that the virtual power plant provides fixed compensation per unit of regulated power according to the signed compensation agreement, using parameters... , and All figures are in yuan per kilowatt-hour. This means that the cost models for distributed photovoltaic power, energy storage, and electric vehicles are all based on fixed compensation, which are constants. , and .

[0028] Step 2: Calculation and Weight Configuration of Regulation Performance Indicators. Calculate the regulation performance indicators for each type of distributed resource, including controllable capacity, regulation accuracy, and unit regulation cost, and configure the indicator weights using the Analytic Hierarchy Process (AHP).

[0029] Specifically, three key performance indicators are introduced: controllable capacity, control accuracy, and unit control cost. The relative weights of the performance indicators are calculated using the Analytic Hierarchy Process (AHP) to guide subsequent resource aggregation and priority ranking, thus achieving a unified consideration of resource technical capabilities and economic objectives.

[0030] In step 2, the calculation and weighting of regulation performance indicators include the following steps: introducing three key performance indicators: adjustable capacity, regulation accuracy, and unit regulation cost; using the Analytic Hierarchy Process (AHP) to calculate the relative weights of the performance indicators to guide subsequent resource aggregation and priority ranking; and achieving a unified consideration of resource technical capabilities and economic objectives.

[0031] Step 2.1: Definition and Calculation of Adjustable Resource Regulation Performance Indicators The issuance and response cycle of automatic power control commands is typically on the order of seconds or minutes. Therefore, virtual power plants participating in automatic power control require resource regulation methods with strong timeliness. This invention proposes a collaborative direct control technology for feeder loads and multiple types of distributed flexible loads to improve the efficiency of automatic power control in virtual power plants, quantitatively characterize the aggregation and regulation capabilities of virtual power plants on a real-time scale, and clarify the principles of control command decomposition.

[0032] The regulation performance of controllable resources within a distribution network can be characterized from multiple aspects. This invention considers defining resource characteristic indicators from both technical and economic perspectives, specifically including controllable capacity. Accuracy of regulation With unit control cost ,like Figure 2 As shown, the calculation of regulation performance indicators can be divided into three categories: controllable capacity, regulation accuracy, and unit regulation cost.

[0033] (1) Adjustable capacity index Controllable capacity index of resources It is an important indicator characterizing the adjustability margin of distributed resources. In real-time automatic power control in virtual power plants, it is necessary to monitor and collect the state parameters of distributed resource units in real time and calculate the adjustability capacity. Due to differences in adjustment characteristics, the definition and calculation method of this indicator also differ. The definitions and calculation methods of the real-time adjustability capacity indicators for different types of distributed resources are given below.

[0034] Calculation of the real-time adjustable capacity of distributed photovoltaic (PV) units. Distributed PV generally operates in maximum power point tracking (MPPT) mode, therefore it is assumed that it only has the capability to adjust active power. According to the requirements of the "Technical Regulations for Grid Connection of Photovoltaic Power Plants" (Q / GDW 1617—2015) and the "Technical Regulations for Power System Connection of Photovoltaic Power Plants" (GB / T 19964-2012), the power fluctuation rate of PV under normal operating conditions is limited to 10% / min. Therefore, the adjustable capacity of distributed PV units is considered to be 10% of their current operating power, as shown in equation (2-1).

[0035] (2-1) Real-time adjustable capacity calculation of energy storage units. Based on the analysis of electrochemical energy storage regulation characteristics, its operational constraints are generally described using charge / discharge power and state of charge. Building upon this, this invention combines the time scale of APC (Automatic Power Control) signals to determine the adjustable state of the energy storage unit. First, the minimum up-adjustment time of the energy storage unit is calculated. With the reduction of time , which represents the length of time that can be controlled when maximizing the controllable capacity under different scenarios, and the specific calculation method is shown in equations (2-2)-(2-3).

[0036] (2-2) (2-3) in, Let be the state of charge of energy storage unit s at time t. and These are the rated charging and discharging powers of energy storage unit S, respectively. and To improve its charging and discharging efficiency, and These represent the maximum and minimum states of charge allowed for the energy storage unit, respectively. Its rated capacity.

[0037] Secondly, the adjustable state is determined based on the resource aggregation and APC cycle, and then the adjustable capacity of its output is calculated using equations (2-4)-(2-5). and adjustable capacity .

[0038] (2-4) (2-5) Real-time adjustable capacity calculation for electric vehicle units. Compared to distributed photovoltaic and energy storage, electric vehicles, as a flexible load, require analysis of their technically adjustable capacity, prioritizing the fulfillment of user electricity demands, in order to obtain control authorization from electric vehicle users. Specifically, this invention assumes that electric vehicle charging stations have both charging and discharging functions, and that users will provide their expected travel times based on their electricity needs. and the expected state of charge at this moment. Therefore, the state of charge that the electric vehicle needs to reach in the next cycle can be calculated according to equation (2-6). .

[0039] (2-6) In the formula: Rated charging power for electric vehicles, and These are its charging efficiency and rated capacity, respectively.

[0040] For scenarios requiring a reduction in load power, if the battery charge at time t is greater than... If the user's electricity demand can still be met, the charging power can be reduced to zero or the vehicle can be discharged within the current control cycle; otherwise, the electric vehicle needs to be charged to its maximum capacity within the current control cycle. If it operates at the rated charging power, then the lower power limit can be expressed as: (2-7) in: , Let t be the state of charge of the nth electric vehicle.

[0041] If it is necessary to increase the load power of electric vehicles, this demand tends to meet the user's electricity needs. Therefore, it is only necessary to consider the battery's state of charge (SOC) constraint, ensuring that the SOC at the beginning of the next control cycle is not greater than the maximum SOC. Then the upper boundary of the power can be expressed as: (2-8) at this time, Combined with real-time measured electric vehicle charging power It can calculate the adjustable capacity of the nth electric vehicle. With adjustable capacity .

[0042] (2-9) (2-10) (2) Regulation accuracy index The control accuracy directly reflects the precision with which each distributed resource unit responds to APC commands, and its calculation method is defined as follows: (2-11) In the formula, and These represent the actual operating power and reference operating power of the distributed resource unit within a certain historical period, respectively. The allowable adjustment error for this unit is generally taken as a certain percentage of its maximum capacity; This refers to the total number of historical time periods involved in the statistical calculations. This invention uses... , and This represents the control accuracy of three types of distributed resource units.

[0043] (3) Unit control cost indicators In automatic power control of virtual power plants, control cost is a key influencing factor in their resource regulation strategy. For distributed resources, the unit regulation cost of a single resource can fully characterize the economic efficiency of distributed resource regulation. Furthermore, different distributed resource units differ in type, regulation characteristics, and importance, thus their unit regulation costs also vary. This invention uses... , and This represents the unit control cost of three types of distributed resource units.

[0044] Step 2.2: Use the Analytic Hierarchy Process (AHP) to determine the weights of the indicators and form a resource evaluation system. Based on the selected performance indicators for distributed resource regulation, the practical significance of each indicator should be further considered, the differences in the importance of different indicators should be analyzed, and weights should be allocated to the indicators as the basis for distributed resource clustering and aggregation. A feasible approach is to analyze this based on professional knowledge, experience, and practical needs, subjectively judging the relative importance of different indicators, and then assigning weights to them. Commonly used subjective weighting methods include the Delphi method, binomial coefficient method, order relation analysis, and analytic hierarchy process (AHP). These weighting methods, to a large extent based on experience, can produce decision schemes that meet expected goals and needs.

[0045] The weight allocation method used in this invention is the Analytic Hierarchy Process (AHP), which is widely used in the power system field. This is a decision-making method that decomposes decision-related elements into hierarchical levels such as objectives, criteria, and indicators, and then performs qualitative and quantitative analysis based on these levels. The specific process is as follows: (1) Establish a hierarchical structure model. Specifically, the regulation performance of control resources in the distribution network is taken as the target layer, technicality and economic efficiency are taken as the criterion layer, and the three regulation performance indicators selected in the previous section are taken as the indicator layer.

[0046] (2) Construct an indicator importance judgment matrix.

[0047] (2-12) The above formula is the constructed indicator importance judgment matrix A, where... This represents the comparison result of the importance between indicator i and indicator j, where n represents the dimension of the judgment matrix, i.e., the number of indicators, and the elements of the judgment matrix should satisfy... ,like i = j ,but . The value should be based on experience and actual needs, and the importance of the indicator should be quantitatively rated according to the proportional scale table, as shown in Table 1.

[0048] Table 1. Scale Table for Analytic Hierarchy Process (AHP)

[0049] (3) Calculate the indicator weights. In the analytic hierarchy process (AHP), there are three specific methods for calculating indicator weights: eigenvalue method, geometric mean method, and arithmetic mean method. This invention uses the relatively simple geometric mean method (square root method) to calculate the indicator weights. As shown in equations (2-13) and (2-14), firstly, the nth root of the product of the indicators in each row of the judgment matrix is ​​calculated, and then the calculation results are normalized to obtain the indicator weights. And the eigenvector W composed of the weights of each indicator.

[0050] (2-13) (2-14) (4) Perform a consistency check. Perform a consistency check on the judgment matrix to determine whether the matrix has satisfactory consistency. First, calculate the largest eigenvalue of the judgment matrix. The calculation method is as shown in equation (2-15). Next, the consistency index is calculated using equation (2-16). And combined with the order of the random judgment matrix and the randomness consistency index Calculate the consistency ratio If the calculated consistency ratio Less than or equal to a preset threshold, such as If the judgment matrix is ​​correct, then the weight of the indicator can be determined.

[0051] (2-15) (2-16) (2-17) It should be noted that R I R is the average consistency index corresponding to the "randomly generated judgment matrix," serving as a reference standard for consistency testing. I No calculation is needed; it's a preset fixed value. It needs to be obtained by looking up a table based on the order n of the judgment matrix (common matrices are third and fourth order; when n=3, R...). I =0.58, R when n=4 I =0.90, different orders correspond to different R values. I value).

[0052] Step 3: Clustering and Aggregation, and Two-Tier Control Strategy. Based on index weights and clustering algorithms, distributed resources are clustered and aggregated to construct a two-tier control architecture consisting of a virtual power plant main control layer and a resource cluster internal control layer.

[0053] Specifically, an improved K-means++ clustering algorithm is used, combined with weighted indicators, to achieve resource clustering and avoid decision-making dimension explosion; a two-level scheduling structure is divided into a "virtual power plant main control layer" and a "resource cluster internal control layer" to achieve a closed-loop control from global to local; the main control layer is responsible for issuing automatic power control commands to each aggregate; the cluster layer completes command splitting and minimum cost allocation according to internal resource constraints.

[0054] This invention addresses the collaborative direct control of a virtual power plant for distributed resource regulation and feeder load power control. It quantifies the regulation capability by defining the regulation performance index of distributed resources, realizes cluster aggregation based on data-driven algorithms, and finally uses voltage sensitivity to describe the coupling relationship between distributed resources and feeder load regulation, thereby achieving real-time aggregation regulation capacity assessment and command decomposition of the virtual power plant.

[0055] Based on the preceding analysis, automatic power control in a virtual power plant requires the clustering of distributed resources. The fundamental purpose is to achieve hierarchical control of massive distributed resources. The virtual power plant uses the clustered aggregates as direct decision-making objects to participate in the first-level automatic power control command decomposition. Subsequently, the cluster performs internal second-level control command decomposition, reducing the computational burden on the virtual power plant's control system. For the first level, this invention first employs a K-means++ algorithm based on improved index weights to cluster and aggregate distributed resources. Simultaneously, utilizing the control performance index data defined and calculated in step 2.2 above, along with the configured index weights, it obtains resource clustering results that consider the physical meaning of the indicators. The specific steps of this method are as follows: Step 3.1, Indicator Normalization Clustering indicators have different meanings, magnitudes, and dimensions, requiring normalization. Considering the actual meaning of the indicators, the three selected clustering indicators are divided into two categories: positive and inverse indicators. The former includes adjustable capacity. and control accuracy The latter includes unit control costs The normalization methods for the two types of indicators are shown below, where "+" and "-" represent the positive and negative indicators of normalization, respectively. Represents the original data value. and These represent the maximum and minimum values ​​of the indicator data, respectively.

[0056] (3-1) (3-2) Step 3.2: Selection of the optimal number of clusters This invention uses the silhouette coefficient method to evaluate and select clustering results. By calculating the average silhouette coefficient of the clustering results, the cohesion and separation of each cluster are determined. After comparing the average silhouette coefficients at different K values, the optimal number of clusters is selected. The specific process of this method is as follows: (1) Calculate the sample The average distance to other samples in the same cluster This is called a sample. Intra-cluster dissimilarity; (2) Calculate the sample To other clusters The average distance of all samples within the range This is called a sample. with cluster Inter-cluster dissimilarity; (3) Calculate the silhouette coefficient of the sample. The values ​​are all between [-1, 1], and the closer the value is to 1, the more reasonable the current clustering result of the sample is. The calculation formula is as shown in equation (3-3): (3-3) (4) Take the mean of the silhouette coefficients of all samples as the silhouette coefficient S of the clustering result, draw a silhouette coefficient line graph, and select the K value corresponding to the maximum silhouette coefficient value as the optimal number of clusters.

[0057] Step 3.3, Cluster Center Initialization The K-means++ algorithm improves clustering efficiency by adjusting the probability of different locations serving as cluster centers. The specific cluster center generation process is as follows: (1) Randomly select a sample from the input sample set as the first cluster center; (2) Calculate the other samples in the sample set Euclidean distance to the nearest cluster center ; (3) Add a cluster center, and when selecting, add those with greater cluster size. The probability of a sample value; (4) Repeat steps (2) and (3) until K cluster centers are selected.

[0058] Step 3.4: Calculation of the Euclidean distance matrix considering index weights Based on the K-means++ algorithm, the influence of index weights is further considered, and Euclidean distance from distributed unit samples to each cluster center is calculated using equation (3-4).

[0059] (3-4) in, , Represents the k-th cluster center. This is a collective term for all indicators of sample Ui. Let j be the weight of the j-th indicator. For the k-th cluster center A collective term for all indicators.

[0060] Step 3.5: Sample Clustering and Cluster Center Update Based on the Euclidean distance matrix calculation results, the feature vector of each sample is assigned to the cluster corresponding to the nearest cluster center, and after the clustering is completed, the cluster center set is updated in the manner shown in equation (3-5): (3-5) in For clusters The total number of distributed resources in the system. For clusters The i-th sample in the dataset.

[0061] Step 3.6: Determine the cluster center location and complete the aggregation. Repeat steps 3.4 and 3.5 until the updated cluster center set is exactly the same as the cluster centers of the previous iteration, then stop the iterative solution and complete the cluster aggregation.

[0062] Based on the above method, it is possible to cluster massive distributed resources. The K-means++ clustering algorithm based on improved index weights used in this invention is as follows: Figure 3 As shown.

[0063] After grouping distributed resources, it is necessary to further characterize the aggregation and control performance of the clusters. This will effectively reduce the dimensionality of decision variables and improve the feasibility of multiple control resources participating in distribution network regulation. Specifically, based on the regulation performance indicators of distributed resources within each cluster, a cluster characteristic index aggregation model is calculated to obtain the aggregated controllable capacity of the resource aggregate. Evaluation of the accuracy of regulation and average unit control cost parameters As shown in equations (3-6)-(3-8): (3-6) (3-7) (3-8) In the formula, and Let be the maximum up-adjustment capacity and the maximum down-adjustment capacity of the k-th resource aggregate at time t, respectively. and Let be the adjustable capacity and adjustable capacity of the distributed resource unit y within the k-th resource aggregate at time t, respectively. , , and These represent the number of the three types of resources within the cluster and the total number of resources.

[0064] Based on the aforementioned index definitions, algorithm clustering, and aggregation calculation processes, real-time aggregation and regulation capabilities for massive distributed resources can be obtained. The virtual power plant constructed in this invention further introduces feeder load power control technology on top of distributed resource aggregation control. Furthermore, due to the strong coupling between active and reactive power in the distribution network, the adjustment of distributed resource aggregation power at each node also affects node voltage, thus impacting the feeder load power control effect. Therefore, the aggregation regulation capacity of the virtual power plant considering both power regulation methods is not a simple summation of their theoretical values. Since the control cycle of automatic power control is relatively short, typically on a timescale of seconds to minutes, voltage sensitivity can be used to describe the impact of node injected power changes on node voltage, thereby allowing the derivation of the virtual power plant's aggregation regulation capacity. and The expression: (3-9) (3-10) In the formula: and This represents the total adjustable capacity of all adjustable resources on node i. Here, both upward and downward adjustments are from the source side perspective. and This is the vector of active power adjustment for distributed resources at each node. and This represents the adjustable capacity and adjustable capacity of the distributed resources on node i. When no distributed resources are connected to the node, this value is 0. and This represents the maximum adjustable value of the SVC compensation for each node at the current moment. and These are the sensitivity vectors of the i-node voltage to changes in the reactive and active power injected into each node.

[0065] After forming several distributed resource aggregates and obtaining automatic power control instructions for each aggregate, a second level of decomposition is needed to break down the power control requirements into specific distributed resource units. To address this, this invention constructs a cluster instruction decomposition model based on the principle of economy. Essentially, it is a process of prioritizing and sequentially adjusting the control of resources within the cluster according to control costs. Its objective function and constraints are as follows: (3-11) (3-12) In the formula, Let be the power control command for the k-th resource aggregate at time t. , and These represent the unit active power regulation of distributed photovoltaic, energy storage, and electric vehicles within the cluster, respectively.

[0066] Step 4: Construct a collaborative control optimization model. Introduce a voltage sensitivity matrix based on a linearized power flow model to construct a collaborative control optimization model that considers the coupling effect between feeder loads and distributed resources, thereby maximizing revenue.

[0067] Specifically, a voltage sensitivity matrix based on a linearized power flow model is introduced to characterize the relationship between node power injection and voltage response. A multi-objective optimization model is constructed with the goal of maximizing system benefits, including power response efficiency, cost expenditure, and load fluctuation. The model considers the coupling effect between feeder load regulation behavior and distributed resource injection power to achieve coordinated system control. Voltage sensitivity is updated in real time to provide feedback on regulation effects and optimize subsequent command issuance.

[0068] Feeder load power control is achieved through voltage regulation. This invention uses distributed resources and Static Var Compensators (SVCs) on some nodes as adjustable active and reactive power sources within the network, i.e., regulating node voltage through parallel resources. Based on the previous analysis of feeder load power control technology, it is necessary to further clarify the impact of node injected power changes on network voltage. Current research often uses the inversion of the Jacobian matrix of the system power flow to obtain the voltage sensitivity matrix, establishing a linear relationship between node voltage and node power changes. This significantly reduces computational complexity and is widely used in real-time voltage optimization problems. However, when the system operating state changes, the Jacobian matrix needs to be recalculated, and the larger the system topology, the longer the computation time. Therefore, this invention proposes a linearized voltage sensitivity calculation method based on a linearized power flow model to facilitate real-time voltage sensitivity calculation and meet the computational speed requirements of APC.

[0069] According to the concept of the linearized power flow model, the voltage of each node is simultaneously affected by the voltage of the balancing node and the power injected from the PQ node. These can be regarded as equivalent voltage sources and current sources, respectively. After the former is grounded and short-circuited, the latter can be regarded as the effect of the equivalent current of each node on the voltage of the PQ node. Therefore, its physical meaning is consistent with the voltage sensitivity of the PQ node type. The corresponding expression for the node voltage sensitivity can be obtained by taking the partial derivatives of the node voltage V with respect to the active power injected P and the reactive power injected Q. (4-1) (4-2) It should be noted that the voltage sensitivity matrix itself should be a scalar value. The above formula only gives the theoretical expression based on the linearized power flow model. In practical applications, it is necessary to further derive the analytical expression of the voltage sensitivity matrix, such as formulas (4-3) and (4-4).

[0070] (4-3) (4-4) In the formula, and These represent the operators for finding the real and imaginary parts of the target matrix, respectively.

[0071] Thus far, the parameters characterizing the system's topology have been obtained. Current running status Then, the active-voltage sensitivity and reactive-voltage sensitivity of each node can be calculated analytically using equations (4-3) and (4-4).

[0072] (1) Objective function Based on the preceding analysis, the participation of virtual power plants in automatic power control first requires the distributed resource aggregate and reactive power compensation equipment as direct decision-making objects to complete the first-level automatic power control command decomposition. Furthermore, the primary objective of virtual power plant operation is to ensure its own profitability. Therefore, this invention takes maximizing the net revenue from frequency regulation ancillary services provided by the virtual power plant as the goal of collaborative control of the two types of resources, establishing the following objective function: (4-5) In the formula: c k = + + ; The total control cost of the virtual power plant. The revenue from providing frequency modulation ancillary services at time t includes two parts: revenue from frequency modulation mileage and revenue from AGC capacity compensation. Let $t$ be the control cost of the k-th resource aggregate at time $t$. Let i be the active power control cost of the load at node i at time t; and These represent the active power of the resource aggregate and the node load, respectively. This refers to the automatic power control cycle of a virtual power plant; here... This represents the average comprehensive frequency regulation performance index of the virtual power plant over the most recent 8 winning bid periods up to time t; The expected frequency regulation mileage at time t can be determined by the difference in tie-line power before and after optimization for each time period. The tie-line power before optimization is the actual measured value, and the optimized tie-line power can be obtained by adding equation (4-5) as a constraint. The decision variable is the active power regulation of each resource aggregation. And the reactive power compensation increment of each static var compensator (SVC). .

[0073] (2) Constraints In the process of automatic power control with coordinated response of multiple types of resources, the main consideration is the operation control constraints of adjustable resources; in addition, while applying feeder load power control technology, direct-controlled virtual power plants also need to constrain network voltage to ensure the safe operation of the distribution network.

[0074] Regional tie-line power constraints: In addition to transmission power constraints, regional tie-line power must also meet the control objectives given by the upper-level power grid, maintaining the tie-line power at the target value. Nearby. In actual control processes, there is a high possibility that the active power response may not be adequate; therefore, a certain response error is permissible, measured by the maximum permissible deviation rate. If we represent it this way, then the constraint can be expressed as: (4-6) In the formula: and These represent the maximum and minimum power output at the drop point, respectively.

[0075] Node voltage constraints: After regulating distributed resources and reactive power compensation equipment, the voltage of each node should still be maintained within a safe range, as shown in the following formula: (4-7) In the formula: This represents the initial measured voltage of node i at the current time t. This represents the voltage increment at that node under the combined influence of the Static Var Compensator (SVC) and the resource aggregation control. and These represent the maximum and minimum node voltages, respectively. According to national standards, in medium-voltage distribution networks... and They are 1.07 pu and 0.93 pu respectively.

[0076] Distributed resource cluster operation constraints: The resource aggregation model calculates the controllable capacity of each distributed resource cluster, enabling cluster adjustment... Maintain within the corresponding clause boundaries, as shown in equation (4-8), where , .

[0077] (4-8) Static Var Compensator (SVC) operating constraints: At any given time, the reactive power compensation amount of the SVC after adjustment should still remain within its allowable range. (4-9) In the formula, This represents the initial reactive power compensation of the l-th static var compensator (SVC) at time t. and These are its maximum and minimum reactive power compensation, respectively.

[0078] Voltage regulation and feeder load power regulation constraints: Based on the calculated linearized voltage sensitivity, the active power regulation of each node can be quickly calculated using the CVR coefficient, which meets the requirements of power allocation optimization for speed. The expression of this process is shown in equations (4-10)-(4-11).

[0079] (4-10) (4-11) In equation (4-10), the first matrix on the right is the voltage sensitivity matrix, which contains the sensitivity of the voltage at feeder node i to the reactive power changes injected into the SVC access node and the active power changes of the distributed resource access node. The update cycle of the entire network voltage sensitivity matrix is ​​the same as the distributed resource clustering and aggregation cycle and the automatic power control cycle. The second matrix represents the reactive or active power increments of all SVCs, distributed photovoltaics, energy storage systems (ES), and electric vehicles (EVs) within the network. Equation (4-11) contains... This represents the increase in active power of the load at feeder node i during time period t.

[0080] In summary, in the method for coordinated direct control of feeder load and distributed flexible resources of this invention, step 1 clarifies the constraints, power characteristics, and control costs of various types of resources by constructing adjustment characteristic models and control cost models for multiple types of distributed resources, providing basic data support for subsequent steps. The subsequent applications of step 1 are as follows: the performance index calculation in step 2 needs to be based on these model parameters (adjustable capacity needs to refer to the maximum output and SOC constraints in step 1); the clustering in step 3 and the optimization model in step 4 also require resource constraints and cost data. Step 2, by calculating and assigning weights to three major indicators—adjustable capacity, control accuracy, and unit control cost—achieves a quantitative assessment of the technical and economic efficiency of resources. Its output weighted index system is the core basis for step 3. The improved K-means++ clustering algorithm in step 3 needs to combine these weighted indicators for resource clustering to ensure that the clustering results meet both technical and economic objectives. Step 3 obtains resource aggregates through clustering and aggregation, establishes a two-level control structure, and completes the initial instruction decomposition. The output resource aggregate parameters and two-level scheduling rules provide the premise for Step 4. The collaborative optimization model in Step 4 needs to use the aggregates as control units and combine the constraints of the two-level control to construct a global optimization scheme. Step 4, based on the resource model, index weights, and clustering results from the previous steps, introduces a voltage sensitivity matrix to characterize the coupling relationship, constructs a multi-objective optimization model, and realizes collaborative control throughout the entire process. Simultaneously, the feedback of its control effect can also feed back into the model parameter optimization of the previous steps, forming a closed loop. Ultimately, it achieves the goal of fast, accurate, and low-cost intelligent control, applicable to practical scenarios such as virtual power plants and Automatic Power Control (APC). This method realizes high-frequency, large-amplitude, and low-cost flexible load control in distribution networks, improving the economy and responsiveness of virtual power plants in orderly power consumption and frequency regulation ancillary services.

[0081] Based on the same inventive concept, another embodiment of the present invention provides a system for coordinated direct control of feeder loads and distributed flexible resources, used to implement the method for coordinated direct control of feeder loads and distributed flexible resources in the aforementioned embodiment. The system includes: The module is used to model the regulation characteristics of various types of distributed resources, including regulation characteristic models and cost models for feeder loads, distributed photovoltaics, energy storage, and electric vehicles. The calculation module is used to calculate the control performance indicators of various types of distributed resources, including controllable capacity, control accuracy and unit control cost, and uses the analytic hierarchy process to configure the indicator weights. The control module is used to group and aggregate distributed resources based on index weights and clustering algorithms, and to build a two-level control architecture of the virtual power plant main control layer and the resource cluster internal control layer. The coordination module is used to introduce a voltage sensitivity matrix based on a linearized power flow model to construct a coordinated control optimization model that considers the coupling effect between feeder load and distributed resources, thereby maximizing benefits.

[0082] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. It should be understood that the invention is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for coordinated direct control of feeder loads and distributed flexible resources, characterized in that, include: The regulation characteristics of various types of distributed resources are modeled, and regulation characteristic models and cost models are constructed for feeder loads, distributed photovoltaics, energy storage and electric vehicles. Calculate the regulation performance indicators for various types of distributed resources, including controllable capacity, regulation accuracy, and unit regulation cost, and use the analytic hierarchy process (AHP) to configure the indicator weights. Based on index weights and clustering algorithms, distributed resources are grouped and aggregated to construct a two-level control architecture of virtual power plant main control layer and resource cluster internal control layer. By introducing a voltage sensitivity matrix based on a linearized power flow model, a collaborative control optimization model considering the coupling effect of feeder load and distributed resources is constructed to maximize benefits.

2. The method for coordinated direct control of feeder load and distributed flexible resources according to claim 1, characterized in that, The regulation characteristic model of the feeder load is constructed based on the ZIP load model and CVR coefficient, and the cost model of the feeder load is constructed based on the voltage level and real-time load power.

3. The method for coordinated direct control of feeder load and distributed flexible resources according to claim 2, characterized in that, The regulation characteristic model of the distributed photovoltaic system is constructed based on the active power operation constraints of the distributed photovoltaic system. The regulation characteristic model of the energy storage system is constructed based on the operating power constraints, energy constraints, and dynamic characteristics of the energy storage system. The regulation characteristic model of the electric vehicle system is constructed based on the operating power constraints, energy constraints, dynamic characteristics of the electric vehicle system, and the electricity demand of users. The cost models of the distributed photovoltaic system, energy storage system, and electric vehicle system are all based on fixed compensation and are constants. , and .

4. The method for coordinated direct control of feeder load and distributed flexible resources according to claim 3, characterized in that, The calculation of the regulation performance index includes: The adjustable capacity of the distributed photovoltaic system is 10% of its current operating power. The adjustable capacity of the energy storage is calculated based on its state of charge, charge / discharge power, and charge / discharge efficiency. The adjustable capacity of the electric vehicle is calculated based on user electricity demand and battery state of charge constraints.

5. The method for coordinated direct control of feeder load and distributed flexible resources according to claim 4, characterized in that, The accuracy rate of the regulation of the distributed resources is: (2-11) in, and These represent the actual operating power and reference operating power of the distributed resource during a certain historical period, respectively. The allowable error for adjusting this distributed resource; This represents the total number of historical periods involved in the statistical calculations.

6. The method for coordinated direct control of feeder load and distributed flexible resources according to claim 5, characterized in that, The method of configuring index weights using the analytic hierarchy process includes: The target layer is the regulation performance of distributed resources in the distribution network, the criterion layer is the technical and economic efficiency, and the indicator layer is based on the controllable capacity, regulation accuracy and unit regulation cost. Construct an indicator importance judgment matrix: (2-12) Where A represents the indicator importance judgment matrix, The matrix A represents the comparison result of the importance between indicator i and indicator j, where n represents the number of indicators; the elements of the indicator importance judgment matrix A satisfy... ,like i = j ,but ; Calculate the nth root of the product of indicators in each row of the indicator importance judgment matrix, and then normalize the result to obtain the indicator weights. and the eigenvector W composed of the weights of each indicator; Calculate the largest eigenvalue of the importance judgment matrix. Then calculate the consistency index. And combine the order of the random judgment matrix and the random consistency index Calculate the consistency ratio The calculation formula is: (2-15) (2-16) (2-17) If the calculated consistency ratio If the value is less than or equal to the preset threshold, the importance judgment matrix A of the indicator is deemed reasonable, and the weight of the indicator is determined.

7. The method for coordinated direct control of feeder load and distributed flexible resources according to claim 6, characterized in that, Based on index weights and clustering algorithms, distributed resources are grouped and aggregated to construct a two-level control architecture: the main control layer of the virtual power plant and the internal control layer of the resource cluster. This includes: An improved K-means++ clustering algorithm is used for resource clustering, including index normalization, selection of the optimal number of clusters, initialization of cluster centers, Euclidean distance calculation considering index weights, sample clustering and iterative update of cluster centers, until the location of the cluster centers is determined to complete the clustering and aggregation.

8. The method for coordinated direct control of feeder load and distributed flexible resources according to claim 7, characterized in that, The two-level control architecture includes: The virtual power plant's main control layer is responsible for receiving automatic power control commands and distributing them to various resource aggregates. The resource cluster internal control layer completes the automatic power control command splitting and minimum cost allocation based on the cluster's internal resource constraints.

9. The method for coordinated direct control of feeder load and distributed flexible resources according to claim 8, characterized in that, The voltage sensitivity matrix is ​​derived based on a linearized power flow model and is used to characterize the relationship between node power injection and voltage response. It is also updated in real time to provide feedback on the control effect. The objective function of the collaborative control optimization model is to maximize the net revenue of the virtual power plant. The constraints include regional tie-line power constraints, node voltage constraints, distributed resource cluster operation constraints, static var compensator operation constraints, and voltage regulation and load power regulation constraints.

10. A system for coordinated direct control of feeder loads and distributed flexible resources, characterized in that, include: The module is used to model the regulation characteristics of various types of distributed resources, including regulation characteristic models and cost models for feeder loads, distributed photovoltaics, energy storage, and electric vehicles. The calculation module is used to calculate the control performance indicators of various types of distributed resources, including controllable capacity, control accuracy and unit control cost, and uses the analytic hierarchy process to configure the indicator weights. The control module is used to group and aggregate distributed resources based on index weights and clustering algorithms, and to build a two-level control architecture of the virtual power plant main control layer and the resource cluster internal control layer. The coordination module is used to introduce a voltage sensitivity matrix based on a linearized power flow model to construct a coordinated control optimization model that considers the coupling effect between feeder load and distributed resources, thereby maximizing benefits.

Images