Virtual power plant regulation capacity dynamic quantitative evaluation and multi-level linkage closed-loop control method

Abstract

The application discloses a virtual power plant regulation capacity dynamic quantitative evaluation and multi-stage linkage closed-loop control method, and belongs to the technical field of power system operation and control. The steps comprise the following: constructing a dynamic capacity stereographic map based on heterogeneous resource physical parameters; decomposing a total power instruction into control primitives matched with a coordinated resource cluster and predicting risks; performing parallel safety deduction through a virtual power grid deduction environment and generating an instruction strategy package; issuing an instruction and monitoring execution deviation, triggering a traceability-collaborative compensation mechanism to suppress deviation and updating resource states. The application adopts the construction of a dynamic capacity stereographic map for accurate evaluation, generates an instruction strategy package in combination with virtual deduction for safe pre-control, and uses a traceability-collaborative compensation mechanism for stable closed-loop collaborative control strategy, which can change power grid safety risks from passive response to proactive prevention in advance, improve the accuracy, safety and stability of virtual power plant regulation, and effectively suppress control oscillation caused by disordered response of a large number of heterogeneous resources.

Description

Technical Field

[0001] This invention belongs to the field of power system operation and control technology, and in particular relates to a method for dynamic quantitative evaluation of the regulation capacity of virtual power plants and multi-level linkage closed-loop control. Background Technology

[0002] A virtual power plant is an advanced energy management system that aggregates geographically dispersed distributed energy sources, controllable loads, and energy storage systems through information and communication technologies and intelligent control algorithms. It functions as a unified power plant, participating in the operation of the electricity market and providing ancillary services to the power grid. Virtual power plants aim to improve the utilization efficiency of distributed resources, provide necessary flexibility and regulation capabilities to the power system, and are an important component in building new power systems.

[0003] Existing virtual power plant control methods typically employ a hierarchical control architecture. The upper-level dispatch center performs economic optimization calculations based on grid demand to arrive at a total power regulation target, which is then decomposed and distributed to individual aggregators or directly to resource terminals for execution. During execution, a closed-loop tracking method is used to monitor total output using traditional PID controllers or simple feedback mechanisms to correct for deviations. For the quantitative assessment of regulation capacity, a simple arithmetic summation of the current adjustable capacities of various resources is often employed.

[0004] However, the aforementioned existing technical solutions have significant drawbacks. First, the simple arithmetic summation-based capability assessment ignores the vast differences in the physical characteristics of different resources, such as response speed, duration, energy constraints, and location within the power grid. This leads to assessment results that deviate significantly from reality, often resulting in unexecutable dispatch commands. Second, the direct issuance and execution of control commands lacks prior assessment of potential safety issues such as local voltage exceedances and line overloads that may arise in a real power grid environment, posing significant safety risks. Finally, traditional closed-loop feedback control mechanisms struggle to distinguish between deviations caused by communication delays and equipment failures, easily leading to malfunctions. Furthermore, the lack of coordination during compensation can cause control oscillations, affecting system stability. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a method for dynamic quantitative evaluation and multi-level linkage closed-loop control of virtual power plant regulation capacity. The method adopts a collaborative control strategy that uses a dynamic capacity three-dimensional map for accurate evaluation, combines virtual simulation to generate instruction strategy packages for safety pre-control, and utilizes a source-collaborative compensation mechanism for stable closed-loop control. This method can improve the accuracy, safety and stability of virtual power plant regulation.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] The method for dynamic quantitative evaluation of virtual power plant regulation capacity and multi-level linkage closed-loop control includes the following steps:

[0008] Based on the real-time physical state parameters of various heterogeneous resources within the virtual power plant, a dynamic three-dimensional map characterizing the synergistic adjustment potential among heterogeneous resources is constructed and continuously updated. Heterogeneous resources are distributed energy resources with different response times, output characteristics, and physical constraints, including distributed photovoltaics, energy storage devices, and controllable loads.

[0009] The system receives total power regulation instructions from the power grid dispatch center and, based on a dynamic capability 3D map, decomposes the total power regulation instructions into several control elements that match the collaborative resource clusters. At the same time, it predicts the execution risk of each control element. The collaborative resource cluster is a set of distributed resources with collaborative regulation capabilities selected from the virtual power plant based on the electrical distance, response sensitivity, and regulation direction between heterogeneous resources in the dynamic capability 3D map.

[0010] Several control primitives are injected into a virtual power grid simulation environment to simulate the real power grid topology and state for parallel security simulation, generating an instruction strategy package that includes main power control instructions and supporting security fine-tuning instructions.

[0011] The instruction policy package is sent to the corresponding heterogeneous resource terminal for execution;

[0012] Monitor the deviation between the actual execution effect and the expected effect of the instruction strategy package, trigger the source tracing and collaborative compensation mechanism based on the dynamic capability three-dimensional map, suppress the deviation and update the resource status.

[0013] Preferably, constructing and continuously updating a dynamic three-dimensional map characterizing the synergistic adjustment potential among heterogeneous resources includes:

[0014] Obtain real-time physical state parameters and operating condition parameters of the virtual power plant. The real-time physical state parameters include the real-time power and electricity of distributed photovoltaic, energy storage devices and controllable loads.

[0015] Based on the equivalent physical model and coupling constraints of each resource type, and combined with real-time power, energy and operating condition parameters, a resource response feature association network that characterizes the response influence relationship between resources is calculated and generated.

[0016] By integrating real-time physical state parameters with resource response characteristics into a network, a dynamic capability 3D map is formed and updated in seconds. This map can query the optimal collaborative compensation resource objects and paths when any resource state changes.

[0017] Preferably, the total power adjustment command is decomposed into several control primitives that match the cooperative resource cluster, and the execution risks of each control primitive are predicted, including:

[0018] Receive total power adjustment command and extract the target power value and adjustment direction from the command;

[0019] Query the dynamic capability 3D map at the current moment to identify several candidate resource clusters that meet the requirements of adjustment direction and power margin;

[0020] The target power value is allocated to several candidate resource clusters according to optimization rules to form several control elements. The physical association information in the dynamic capability 3D map is used to predict the risk of local power grid parameter overruns that may be caused by the execution of each control element.

[0021] Preferably, the instruction strategy package that generates the main power control instruction and the corresponding safety fine-tuning instruction includes:

[0022] Several control primitives and anticipated potential risk points are injected into the virtual power grid simulation environment;

[0023] Transient power flow calculations are performed in a virtual power grid simulation environment to simulate power flow and equipment status changes after the execution of control primitives, and to generate safety risk simulation results.

[0024] If the safety risk simulation results indicate the existence of a safety risk, the system will automatically retrieve and call safety adjustment resources that can eliminate the risk from the dynamic capability 3D map, and generate a matching safety fine-tuning instruction bound to the control primitive with the risk.

[0025] If the safety risk simulation results indicate that there is no safety risk, then several control primitives will be directly used as the main power control command.

[0026] All control primitives and their corresponding safety fine-tuning instructions are encapsulated to form an instruction policy package.

[0027] Preferably, the accompanying security fine-tuning instructions that generate control primitives that are associated with potential risks include:

[0028] Analysis of the safety risk simulation results determined that the type of safety risk was either voltage over-limit or line overload.

[0029] To address the risk of voltage exceeding limits, resources with reactive power regulation capabilities near the risk point are retrieved from the dynamic capability 3D map, and reactive power regulation fine-tuning instructions are generated.

[0030] To address the risk of line overload, backup resource paths for power flow transfer are retrieved from the dynamic capacity 3D map, and power transfer fine-tuning instructions are generated.

[0031] Bind the reactive power regulation fine-tuning command or power transfer fine-tuning command to the control element that caused the risk in terms of time and logic to ensure that the two are issued and executed synchronously.

[0032] Preferably, the source tracing and collaborative compensation mechanism triggered by the dynamic capability 3D map, which suppresses deviations and updates resource status, includes:

[0033] Real-time acquisition of operating parameters at the virtual power plant grid connection point, and calculation of the actual aggregate output curve;

[0034] The actual aggregated output curve is compared with the expected output curve in the instruction strategy package to generate an output deviation signal and deviation characteristic pattern.

[0035] By matching and analyzing the deviation feature patterns with the dynamic capability 3D map, the source of the deviation can be determined and the core resource nodes causing the deviation and the related resource sets affected by them can be located.

[0036] If the source of the deviation is determined to be a local device failure, a precise compensation instruction is sent to the associated resource set, and the status marker of the failed resource node in the dynamic capability 3D map is updated.

[0037] If the source of the deviation is determined to be a transient disturbance that is not due to equipment failure, the current control command remains unchanged, and the deviation characteristic pattern is recorded in the deviation event log.

[0038] Preferably, sending precise compensation instructions to the associated resource set and updating the status markers of failed resource nodes in the dynamic capability 3D map includes:

[0039] Based on the dynamic capability 3D map, obtain the electrical distance and adjustment capability weight between core resource nodes and related resource sets;

[0040] Based on the weights of electrical distance and regulation capability, the power gap to be compensated is dynamically allocated to some or all of the resources in the associated resource set, generating differentiated compensation instructions;

[0041] Issue differentiated compensation instructions to enable related resource sets to coordinate actions to smooth out output deviation signals.

[0042] Preferably, after issuing the instruction policy package, the following operations are performed:

[0043] Within a time window, collect execution confirmation signals from each resource terminal;

[0044] If no execution confirmation signal for the critical resource is received within the time window, the execution of the resource instruction is deemed abnormal, and an instruction execution abnormality event is generated.

[0045] Based on abnormal command execution events, alternative resources are selected from the reserve resource pool using a dynamic capability 3D map, and an emergency switchover command is generated and issued.

[0046] Preferably, after suppressing the deviation and updating the resource status, the dynamic capability 3D map is updated, including:

[0047] Record each deviation event triggered by the source tracing and collaborative compensation mechanism, the compensation strategy, and the final effect to form a deviation event log;

[0048] Based on the deviation event log, the response characteristic parameters of the corresponding resource nodes in the dynamic capability 3D map are corrected, and the corrected response characteristic parameters are generated.

[0049] Update the dynamic capability stereo map for the next cycle using the corrected response characteristic parameters.

[0050] A virtual power plant's dynamic quantitative assessment of regulation capacity and a multi-level linkage closed-loop control system, used to implement the above methods, include:

[0051] Map building and updating module: Based on the real-time physical state parameters of various heterogeneous resources in the virtual power plant, it builds and continuously updates a dynamic three-dimensional map that characterizes the potential for coordinated regulation among heterogeneous resources. Heterogeneous resources are distributed energy resources with different response times, output characteristics and physical constraints, including distributed photovoltaics, energy storage devices and controllable loads.

[0052] Command decomposition and risk prediction module: It is used to receive the total power regulation command issued by the power grid dispatch center, and decompose the total power regulation command into several control elements that match the collaborative resource cluster based on the dynamic capability 3D map. At the same time, it predicts the execution risk of each control element. The collaborative resource cluster is a set of distributed resources with collaborative regulation capabilities selected from the virtual power plant based on the electrical distance, response sensitivity and regulation direction between heterogeneous resources in the dynamic capability 3D map.

[0053] Safety simulation and strategy package generation module: This module is used to inject several control primitives into a virtual power grid simulation environment that simulates the real power grid topology and state for parallel safety simulation, and generate an instruction strategy package that includes main power control instructions and supporting safety fine-tuning instructions.

[0054] Command issuance and execution module: used to issue command policy packages to the corresponding heterogeneous resource terminals for execution;

[0055] Deviation Monitoring and Collaborative Compensation Module: Used to monitor the deviation between the actual execution effect and the expected effect of the instruction strategy package. Based on the dynamic capability 3D map, it triggers a source tracing and collaborative compensation mechanism to suppress the deviation and update the resource status.

[0056] The present invention has the following advantages:

[0057] This invention constructs a dynamic capability 3D map, deeply coupling the physical characteristics of heterogeneous resources with the power grid topology, thereby achieving accurate and dynamic quantification of the regulation potential of virtual power plants. It overcomes the "overestimation" defect of traditional linear summation evaluation methods, making scheduling decisions based on more reliable physical facts, and improving the feasibility of control commands and the overall economic efficiency of system operation.

[0058] This invention introduces a feedforward-based safety pre-simulation mechanism based on a virtual power grid simulation environment, which can proactively identify and avoid potential power grid safety risks before control commands are issued. By generating an instruction strategy package that embeds corresponding safety fine-tuning instructions, safety assurance is integrated into the control process, realizing a shift from passive response to proactive prevention, and improving the safety and reliability of virtual power plants participating in power grid regulation.

[0059] This invention designs a source-tracing and collaborative compensation closed-loop mechanism based on a dynamic capability 3D map, which can intelligently identify the root cause of execution deviations and perform precise local collaborative compensation for faults or disturbances. This not only improves the response speed and accuracy of closed-loop control, but also effectively suppresses the risk of systemic oscillations that may be caused by the disordered response of massive heterogeneous resources, ensuring the stability and regulation quality of the overall control of the virtual power plant. Attached Figure Description

[0060] Figure 1 A flowchart illustrating the dynamic quantitative assessment of the regulation capacity of a virtual power plant and the multi-level linkage closed-loop control method.

[0061] Figure 2 Structural architecture diagram of a virtual power plant's dynamic quantitative evaluation of regulation capacity and a multi-level linkage closed-loop control system;

[0062] Figure 3 This is a schematic diagram of the collaborative weight distribution in the resource response feature association network in this embodiment of the invention;

[0063] Figure 4 This is a schematic diagram comparing the execution effect of the source tracing-collaborative compensation mechanism in an embodiment of the present invention. Detailed Implementation

[0064] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0065] like Figure 1 As shown, the method for dynamic quantitative evaluation of the regulation capacity of a virtual power plant and multi-level linkage closed-loop control includes the following steps:

[0066] Based on the real-time physical state parameters of various heterogeneous resources within the virtual power plant, a dynamic three-dimensional map characterizing the synergistic adjustment potential among heterogeneous resources is constructed and continuously updated. Heterogeneous resources are distributed energy resources with different response times, output characteristics, and physical constraints, including distributed photovoltaics, energy storage devices, and controllable loads.

[0067] The system receives total power regulation instructions from the power grid dispatch center and, based on a dynamic capability 3D map, decomposes the total power regulation instructions into several control elements that match the collaborative resource clusters. At the same time, it predicts the execution risk of each control element. The collaborative resource cluster is a set of distributed resources with collaborative regulation capabilities selected from the virtual power plant based on the electrical distance, response sensitivity, and regulation direction between heterogeneous resources in the dynamic capability 3D map.

[0068] Several control primitives are injected into a virtual power grid simulation environment to simulate the real power grid topology and state for parallel security simulation, generating an instruction strategy package that includes main power control instructions and supporting security fine-tuning instructions.

[0069] The instruction policy package is sent to the corresponding heterogeneous resource terminal for execution;

[0070] Monitor the deviation between the actual execution effect and the expected effect of the instruction strategy package, trigger the source tracing and collaborative compensation mechanism based on the dynamic capability three-dimensional map, suppress the deviation and update the resource status.

[0071] The method described in this embodiment can be applied to, for example... Figure 2 The virtual power plant regulation capacity dynamic quantitative evaluation and multi-level linkage closed-loop control system shown includes:

[0072] Map building and updating module: Based on the real-time physical state parameters of various heterogeneous resources in the virtual power plant, it builds and continuously updates a dynamic three-dimensional map that characterizes the potential for coordinated regulation among heterogeneous resources. Heterogeneous resources are distributed energy resources with different response times, output characteristics and physical constraints, including distributed photovoltaics, energy storage devices and controllable loads.

[0073] Command decomposition and risk prediction module: It is used to receive the total power regulation command issued by the power grid dispatch center, and decompose the total power regulation command into several control elements that match the collaborative resource cluster based on the dynamic capability 3D map. At the same time, it predicts the execution risk of each control element. The collaborative resource cluster is a set of distributed resources with collaborative regulation capabilities selected from the virtual power plant based on the electrical distance, response sensitivity and regulation direction between heterogeneous resources in the dynamic capability 3D map.

[0074] Safety simulation and strategy package generation module: This module is used to inject several control primitives into a virtual power grid simulation environment that simulates the real power grid topology and state for parallel safety simulation, and generate an instruction strategy package that includes main power control instructions and supporting safety fine-tuning instructions.

[0075] Command issuance and execution module: used to issue command policy packages to the corresponding heterogeneous resource terminals for execution;

[0076] Deviation Monitoring and Collaborative Compensation Module: Used to monitor the deviation between the actual execution effect and the expected effect of the instruction strategy package. Based on the dynamic capability 3D map, it triggers a source tracing and collaborative compensation mechanism to suppress the deviation and update the resource status.

[0077] This embodiment constructs a collaborative control architecture encompassing dynamic evaluation, feedforward pre-simulation, and closed-loop self-healing, breaking down the fragmented nature of traditional virtual power plant control. First, by constructing a dynamic capability 3D map that reflects the physical coupling and collaborative potential between heterogeneous resources in real time, the complex, nonlinear resource aggregation problem is transformed into a computable and predictable dynamic network model, providing a unified physical factual basis for accurate decision-making. Second, it abandons the sequential "decision first, verification later" approach, introducing risk pre-identification based on this map during the instruction decomposition stage. Furthermore, it deeply integrates the safety verification process with the generation of preventative control strategies within the virtual power grid simulation environment, ultimately outputting an instruction strategy package with an inherent safety assurance mechanism. Finally, in the execution stage, this method utilizes the physical correlation information of the dynamic capability 3D map to intelligently trace execution deviations, achieving a shift from global blind compensation to local precise collaborative compensation, forming a stable closed loop capable of self-diagnosis and self-repair.

[0078] Based on the real-time physical state parameters of various heterogeneous resources within the virtual power plant, a dynamic three-dimensional map characterizing the synergistic adjustment potential among heterogeneous resources is constructed and continuously updated, including:

[0079] Obtain real-time physical state parameters and operating condition parameters of the virtual power plant. The real-time physical state parameters include the real-time power and electricity of distributed photovoltaic, energy storage devices and controllable loads.

[0080] Based on the equivalent physical model and coupling constraints of each resource type, and combined with real-time power, energy and operating condition parameters, a resource response feature association network that characterizes the response influence relationship between resources is calculated and generated.

[0081] By integrating real-time physical state parameters with resource response characteristics into a network, a dynamic capability 3D map is formed and updated in seconds. This map can query the optimal collaborative compensation resource objects and paths when any resource state changes.

[0082] First, data acquisition and parameter initialization are performed. Remote terminal units (RTUs) deployed at distributed photovoltaic (PV) power plants, energy storage power stations, and controllable load clients acquire real-time physical and operational status parameters of the virtual power plant at a high frequency with an acquisition cycle of 100 milliseconds to 1 second, using industrial communication protocols such as IEC 61850 or Modbus TCP / IP. Real-time physical status parameters mainly include the active power P, reactive power Q, and state of charge (SOC) of each resource node; operational status parameters include equipment switching status, PV inverter operating mode, and controllable load response priority. This raw data forms the foundational data layer for constructing a dynamic capability 3D map. Second, a resource response characteristic correlation network is constructed. This network is a weighted directed graph used to quantify the mutual influence of various resources in the power grid topology. The construction of this network relies on the equivalent physical model of the distribution network, particularly the node impedance matrix Z. The node impedance matrix describes the voltage response caused by injecting a unit current into any node in the network, intuitively reflecting the electrical distance between nodes. The weights of the response influence relationships between resources are also considered. The influence weight from resource j to resource i can be quantified using the following functional relationship:

[0083] ;

[0084] in, The voltage sensitivity factor is extracted from the node impedance matrix. It represents the degree to which the power change of resource j affects the voltage at resource i. This value directly reflects the tightness of electrical coupling. It is the maximum adjustable power margin that resource j can provide at the current moment, which is determined by its real-time physical state parameters, such as the SOC range of energy storage or the curtailment margin of photovoltaics. The power response rate of resource j is typically measured in megawatts per second and reflects its regulation sensitivity.

[0085] The function F is a normalized weighting function that integrates sensitivity, power margin and response rate of different dimensions into a dimensionless weight value. The higher the weight value, the stronger the collaborative compensation ability of resource j to resource i. , , The preset weighting coefficients are used, and they satisfy the following conditions: These coefficients reflect the system's emphasis on electrical distance, energy adequacy, and response speed. For example, in scenarios requiring emergency voltage regulation, these coefficients can be increased. The value of .

[0086] The voltage sensitivity reference value is usually taken as the maximum absolute value of the voltage sensitivity between resource i and all resource nodes on its feeder. The rated power of resource j is used to allocate the current power margin. Converted into a dimensionless percentage form, it reflects the depth of resource regulation. The maximum physical response rate that resource j can achieve is used to determine the real-time response rate. Normalization.

[0087] Finally, the dynamic capability 3D map is fused, generated, and continuously updated. This map is a multi-dimensional data structure that uses real-time physical state parameters as dynamic attributes of nodes and resource response feature association networks as the connection relationships and weights between nodes. Upon receiving a new set of real-time physical state parameters, the map is immediately updated. For example, if the SOC of an energy storage device decreases, its available power margin for upward adjustment... This decreases accordingly, leading to a reduction in the associated weights of all energy storage devices. Dynamic reduction. The refresh cycle of the entire map is controlled on the order of seconds, typically 1 to 5 seconds, to ensure that it accurately reflects the collaborative adjustment potential of all resources within the virtual power plant at any given time.

[0088] The core function of the dynamic capability 3D map is that when a change in the state of a resource is needed, such as a sudden drop in the output of distributed photovoltaic power, an optimized graph search algorithm can be executed on the map to quickly locate one or more energy storage devices or controllable loads with the highest associated weight with the photovoltaic node and sufficient self-adjustment margin. These are then used as the optimal collaborative compensation resource object and path, providing millisecond-level decision support for subsequent closed-loop control. For example... Figure 3 As shown, a resource response feature association network generated based on a dynamic capability 3D map is presented, in which the depth quantization of matrix weights characterizes the collaborative adjustment potential among heterogeneous resource nodes under the power system topology.

[0089] One application example is that, in a virtual power plant, the output of a distributed photovoltaic node i drops sharply due to cloud shading. The optimal compensation resource j needs to be retrieved. First, the voltage sensitivity factor of resource j to node i is calculated using the node impedance matrix. Set the voltage sensitivity reference value to 0.05V / A. The voltage is 0.1V / A, and the state of charge (SOC) of energy storage device j is 75% in real time, corresponding to the maximum adjustable power margin currently available. Its rated power is 80kW. The power is 100kW, and its current power response rate is measured. Its physical maximum response rate is 40kW / s. The power output is 50 kW / s. The weighting factors for sensitivity, power margin, and response rate are set as follows: , , Substitute into the specific normalized weighted function expression Calculations are performed to obtain the collaborative response weight of resource j to resource i at that moment. The calculation result is updated in real time to the dynamic capability 3D map as a dynamic attribute, so that when the graph search algorithm is executed, the energy storage device j can be identified as the optimal collaborative compensation path to smooth the output fluctuation of the photovoltaic node i due to the high weight value.

[0090] The system receives total power regulation commands from the power grid dispatch center and, based on a dynamic capability 3D map, decomposes these commands into several control primitives that match the coordinated resource clusters. Simultaneously, it predicts the execution risks of each control primitive, including:

[0091] Receive total power adjustment command and extract the target power value and adjustment direction from the command;

[0092] Query the dynamic capability 3D map at the current moment to identify several candidate resource clusters that meet the requirements of adjustment direction and power margin;

[0093] The target power value is allocated to several candidate resource clusters according to optimization rules to form several control elements. The physical association information in the dynamic capability 3D map is used to predict the risk of local power grid parameter overruns that may be caused by the execution of each control element.

[0094] First, the command reception and parsing process is executed. When the virtual power grid control system receives a total power regulation command through a remote signaling and control channel conforming to the IEC 60870-5-104 standard, the command decomposition and risk prediction module is immediately triggered. This module first parses the command message and extracts two key parameters: the target power value. With adjustment direction For example, the instruction "increase by 10 megawatts" would... For 10 megawatts, The first step is positive. Secondly, candidate resource clusters are filtered based on a dynamic capability 3D map. Using the current moment as an index, the dynamically updated capability 3D map is queried. This involves traversing pre-divided or dynamically aggregated collaborative resource clusters within the map. A collaborative resource cluster is a set of resources combined based on principles such as proximity by electrical distance and similar response characteristics.

[0095] The selection process follows two hard constraints: first, the overall adjustment direction of the resource cluster must be consistent with the instruction adjustment direction. Matching, for example, requires filtering resource clusters with generation potential or interruptible loads when issuing upward adjustment orders; secondly, the aggregated available power margin of resource clusters. The value must be greater than zero. All resource clusters meeting the criteria are identified as candidate resource clusters. Next, target power allocation and control primitive generation are performed. The total target power value is... Assigning resources to several shortlisted candidate resource clusters is a multi-objective optimization problem. A weighted allocation strategy is adopted, allocating resources based on the comprehensive performance indicators of each candidate cluster. (Assignment weights are then defined.) The calculation formula is:

[0096] ;

[0097] in, The weights are assigned to the k-th candidate resource cluster. , , These are the normalized available power margin, average response rate, and unit adjustment cost for the resource cluster, all of which are obtained directly from the dynamic capability 3D map. , , The weighting coefficient is determined by the scheduling strategy, for example, by increasing it during emergency adjustments. The value increases during economic allocation. Value, and The power allocated to the k-th resource cluster. Then equals Multiply by its weighting percentage. Each Resource cluster k, target power The combination of these elements constitutes a control unit.

[0098] Finally, a dynamic capability 3D map is used to predict execution risks. For each generated control primitive, the resource response features embedded in the map are correlated with network information, particularly the voltage sensitivity matrix, to perform a rapid static safety assessment. The execution of this control primitive is then estimated. Power injection or absorption may cause voltage shifts at several critical grid nodes that are electrically closest to it. If the estimated voltage deviation If the voltage deviation exceeds a preset safety threshold, such as 80% of the ±5% voltage deviation range required by national standards (i.e., 4%), a high-risk warning is marked for that control element. This step does not perform a complete power flow calculation, but rather serves as a rapid screening mechanism to identify the control elements most likely to cause local power grid instability problems, providing a focus for subsequent, more detailed safety simulations.

[0099] One application example is that when a virtual power plant receives a "10MW increase" total power adjustment command from the power grid dispatch center, it first parses and extracts the target power value. 10MW and adjustable direction For the positive direction, candidate resource clusters 1 and 2 that meet the adjustment direction and have a power margin greater than zero are then selected from the dynamic capability 3D map; the weighting coefficient ratio is set to the power margin. Response rate Adjusting costs If the normalized indices of resource cluster 1 are respectively , , The normalized indices for resource cluster 2 are as follows: , , Substitute into the weighted allocation formula The weight of resource cluster 1 is calculated. The weight of resource cluster 2 The power allocated to the control primitive of resource cluster 1 is then... The power allocated to the control primitives of resource cluster 2 ;

[0100] During the risk prediction phase, the execution risk is estimated using a sensitivity matrix. Potential node voltage shift The risk level was 4.2%, which exceeded the preset safety threshold of 4%. Therefore, a high-risk warning label was set for the control element corresponding to resource cluster 2, thus completing the precise decomposition from macro scheduling instructions to micro control elements with risk assessment.

[0101] Several control primitives are injected into a virtual power grid simulation environment to simulate the real power grid topology and state for parallel security simulation, generating an instruction strategy package that includes main power control instructions and supporting security fine-tuning instructions, including:

[0102] Several control primitives and anticipated potential risk points are injected into the virtual power grid simulation environment;

[0103] Transient power flow calculations are performed in a virtual power grid simulation environment to simulate power flow and equipment status changes after the execution of control primitives, and to generate safety risk simulation results.

[0104] If the safety risk simulation results indicate the existence of a safety risk, the system will automatically retrieve and call up safety adjustment resources that can eliminate the risk from the dynamic capability 3D map, and generate a matching safety fine-tuning instruction bound to the control primitive with the risk.

[0105] If the safety risk simulation results indicate that there is no safety risk, then several control primitives will be directly used as the main power control command.

[0106] All control primitives and their corresponding safety fine-tuning instructions are encapsulated to form an instruction policy package.

[0107] First, the simulation environment is initialized and control primitives are injected. The complete topology of the current distribution network, line impedance parameters, transformer tap positions, and real-time operating status of all distributed resources are retrieved from the central database or real-time data platform to construct an accurate virtual power grid simulation environment, which can be regarded as a real-time digital twin system.

[0108] Several control primitives, especially those marked with potential risk points, are injected into the simulation environment as disturbance events to be executed. Next, parallel transient power flow calculations and security risk assessments are conducted. For each injected control primitive, an independent simulation instance is launched, performing transient power flow calculations in its dedicated virtual environment. This calculation is not a simple steady-state power flow analysis, but rather simulates a short time window, such as 0 to 5 seconds, from the issuance of a command to the completion of the adjustment action, analyzing the dynamic changes in voltage, frequency, and line power flow at key nodes of the power grid during this period.

[0109] After the simulation is completed, the simulation results are automatically compared with preset safety constraint boundaries, such as whether the node voltage is maintained between 0.95 and 1.05 per unit and whether the line load rate is below 100%. If all indicators are within the safe range, the safety risk simulation result of the control element is determined to be risk-free; if any indicator exceeds the limit, it is determined to be a safety risk, and the type, location, and severity of the exceedance are recorded. Then, for the identified safety risks, an adaptive fine-tuning instruction generation mechanism is initiated.

[0110] If the safety risk simulation indicates a risk, such as a control element causing the 35 kV bus voltage to drop to 0.94 per unit, the system will immediately use this risk information as an index to query the dynamic capability 3D map. It will then search for the nearest resource with rapid reactive power regulation capabilities, such as a static var compensator or energy storage inverter. Based on the available reactive power capacity and response sensitivity of this resource on the map, the system will automatically calculate the minimum reactive power compensation required to raise the voltage to a safe range and generate a corresponding safety fine-tuning command, such as "STATCOM-03, issue 0.8 Mvar reactive power". This fine-tuning command will be strongly bound to the original control element that caused the risk, ensuring that the two are logically related and executed in tandem.

[0111] Finally, the instructions are encapsulated to form the final instruction policy package. All control primitives that have undergone safety simulation are integrated. For control primitives whose simulation results indicate no risk, they are directly encapsulated as main power control instructions. For control primitives whose simulation results indicate risk, they are encapsulated as instruction pairs along with their corresponding safety fine-tuning instructions. All these main power control instructions and instruction pairs are uniformly organized into a structured data packet, namely the instruction policy package. This policy package clearly defines the target object, adjustment amount, execution priority, and coordination relationship of each instruction, constituting a complete scheduling and execution scheme that has undergone closed-loop safety verification.

[0112] One application example is that a virtual power plant issues commands to a certain collaborative resource cluster. The active power increase control element is first injected into a virtual power grid simulation environment synchronized with the real power grid topology for transient power flow calculation. The simulation results show that after executing the command, the predicted per-unit value of the 35kV bus voltage will drop to The voltage was found to be below the safety threshold of 0.95, indicating a potential risk of voltage exceeding limits. This risk information was then used as an index to query the dynamic capability 3D map. The energy storage inverter with the closest electrical distance was located, and its voltage reactive power sensitivity coefficient was obtained. Substitute into the reverse control formula Calculate the required reactive power compensation, i.e. This generates a matching safety fine-tuning command that "issues 0.8MVar reactive power" and strongly binds it to the original control element. Finally, the main power control command and the matching safety fine-tuning command are encapsulated into a structured command strategy package, realizing the transformation from an unverified control element to an endogenous safety scheduling scheme.

[0113] If the security risk simulation results indicate the existence of a security risk, the system will automatically retrieve and call upon security adjustment resources that can eliminate the risk from the dynamic capability 3D map, and generate corresponding security fine-tuning instructions bound to the control primitives of the existing risk, including:

[0114] Analysis of the safety risk simulation results determined that the type of safety risk was either voltage over-limit or line overload.

[0115] To address the risk of voltage exceeding limits, resources with reactive power regulation capabilities near the risk point are retrieved from the dynamic capability 3D map, and reactive power regulation fine-tuning instructions are generated.

[0116] To address the risk of line overload, backup resource paths for power flow transfer are retrieved from the dynamic capacity 3D map, and power transfer fine-tuning instructions are generated.

[0117] Bind the reactive power regulation fine-tuning command or power transfer fine-tuning command to the control element that caused the risk in terms of time and logic to ensure that the two are issued and executed synchronously.

[0118] First, precise classification and location of safety risks are performed. After the safety risk projection results are generated from transient power flow calculations, these results are analyzed in depth. All grid parameters exceeding safety thresholds are traversed and classified according to the nature of the over-limit indicators. If the per-unit voltage value of a node is below 0.95 or above 1.05, the risk type is determined to be voltage over-limit. If the calculated power flow value of a line exceeds 100% of its rated transmission capacity, it is determined to be line overload.

[0119] Simultaneously, the specific location of the risk occurrence will be accurately recorded, such as the node number of the voltage overrun or the line identifier of the line overload. Secondly, differentiated resource retrieval and fine-tuning instructions will be generated for different risk types. If the risk type is voltage overrun, a neighborhood search will be performed on the dynamic capability 3D map centered on the node where the overrun occurred. The search target is the resource with the closest electrical distance and independent reactive power regulation capability, such as energy storage devices equipped with four-quadrant inverters, STATCOM static var compensators, or photovoltaic power plants with leading-phase operation capability.

[0120] Based on voltage sensitivity analysis, the minimum reactive power regulation required to pull the voltage back to the safe range is calculated. The system generates a reactive power adjustment command, such as "Energy storage station ES-02, generate 1.2 Mvar reactive power." If the risk type is line overload, the system's search strategy is completely different. It searches the grid topology for alternative resource paths that can share the power flow. This means the system searches upstream of the overloaded line for alternative resources that can increase power generation or downstream for alternative resources that can absorb the load, thus achieving power flow transfer between different branches without changing the total regulation power.

[0121] Once a feasible power flow transfer path is determined, a power transfer fine-tuning command will be generated. This is usually a command pair, such as "PV-07 photovoltaic power station, reduce active power output by 0.5 MW" and "CL-03 controllable load cluster, increase load by 0.5 MW".

[0122] Finally, the fine-tuning instructions are logically bound and synchronously encapsulated with the main instructions. The generated reactive power regulation fine-tuning instructions or power transfer fine-tuning instructions are not simply simple additional instructions. At the data structure level, this fine-tuning instruction is strongly associated with the control primitive that initially caused the safety risk. This binding ensures that in the final instruction policy package, the two are treated as an inseparable execution unit. In terms of timing logic, they are set to execute synchronously or with a slight time delay. For example, the fine-tuning instruction must be executed 500 milliseconds before or simultaneously with the main instruction to ensure that safety measures are in place before the main power regulation action has an effect. This binding and timing setting is key to ensuring that the accompanying safety fine-tuning instructions can effectively eliminate risks.

[0123] One application example is when a virtual power plant performs voltage over-limit risk handling for grid node N-105, it deeply analyzes the transient power flow calculation results to identify the per-unit voltage value of that node. The voltage exceeded the safety threshold of 1.05, accurately classified as a high voltage over-limit risk. Subsequently, a neighborhood search was performed on the dynamic capability 3D map centered on node N-105, locating an energy storage inverter ES-03 with reactive power regulation capabilities. The required reactive power regulation was calculated through voltage sensitivity analysis, and a sensitivity factor was set. Substitute into the formula Calculation Based on this, a matching safety fine-tuning instruction of "inductive absorption of 0.2 Mvar reactive power" is generated. Subsequently, at the data structure level, this fine-tuning instruction is strongly correlated and bound with the main power regulation instruction that causes the risk, and a synchronous execution sequence is set to ensure that before the main instruction increases active power output and causes the voltage to rise further, ES-03 has performed reactive power compensation in advance, thereby forcibly pulling the node voltage back to the safe range of 1.05 per unit value.

[0124] After issuing the instruction policy package, perform the following operations:

[0125] After issuing the instruction policy package, collect the execution confirmation signals from each resource terminal within a time window;

[0126] If no execution confirmation signal for the critical resource is received within the time window, the execution of the resource instruction is deemed abnormal, and an instruction execution abnormality event is generated.

[0127] Based on abnormal command execution events, alternative resources are selected from the reserve resource pool using a dynamic capability 3D map, and an emergency switchover command is generated and issued.

[0128] First, execution confirmation monitoring is initiated after the command is issued. Upon successful distribution of the command policy package to all target heterogeneous resource terminals, an independent timer is started for each command, especially those critical resource commands whose contribution to the total power adjustment target exceeds a certain threshold, such as 5%. This timer has a set time window. This time window is estimated based on the physical response characteristics of the resource and the maximum latency of the communication network, and is typically set between 2 and 10 seconds. Within this time window, the system continuously listens for execution confirmation signals from the corresponding resource terminal via the reverse communication link. This signal is a standardized message indicating that the terminal has correctly received, parsed, and begun executing the instruction.

[0129] Perform timeout checks and generate exception events. If within the preset time window... If a confirmation signal for execution of a critical resource is not received and the timer times out, this situation will be immediately classified as an instruction execution exception event. A structured event record containing the exception resource ID, the unconfirmed instruction content, and the event timestamp will be generated. This judgment mechanism does not wait for the actual output deviation to occur, but proactively identifies potential execution failure risks in the initial stage of instruction execution.

[0130] Finally, an emergency resource switchover is performed based on the dynamic capability 3D map. Once an abnormal command execution event occurs, the emergency switchover logic is triggered immediately. This is done using the ID of the abnormal resource and its unexecuted power adjustment. As input, a rapid backup resource search is performed on the dynamic capability 3D map at the current moment. The goal of the search is to find alternatives from a pre-defined pool of backup resources with high reserve coefficients and rapid response capabilities, such as online battery energy storage systems or industrial loads that can be quickly interrupted. The optimal resource is selected. The selection criteria are similar to those for coordinated compensation, prioritizing resources with short electrical distances, fast response times, and low adjustment costs. Once a replacement resource is identified, an emergency switching command is immediately generated. The command's target power remains the same as the original command, but the target object is changed to the newly selected replacement resource. This command is then sent through the highest-priority communication channel. Simultaneously, in the dynamic capability 3D map, the original abnormal resource is temporarily marked as "communication interrupted" or "response abnormal" to avoid issuing commands to it again within a short period.

[0131] One application example is when a virtual power plant executes a frequency regulation command, it issues a command to a controlled industrial electric boiler with a rated regulating power of 100kW. The resource's contribution to the overall target is assessed as 8%, exceeding the preset trigger threshold of 5%, and is then activated. The timer. If the reverse communication link does not send a standardized execution confirmation message within 5 seconds, an execution anomaly is immediately determined and an event log containing the device ID is generated. At this time, the unexecuted power adjustment amount of the abnormal resource is extracted. Assuming the current command requires it to output 80% of its capacity, a dynamic capability 3D map is then retrieved, and a group of distributed battery energy storage systems with the closest electrical distance and a response time of more than 3 seconds is located in the reserve resource pool. The current effective adjustment margin of these systems is then determined. It is 150kW, which meets the requirements. The switching conditions were then determined. The target power for the emergency switching command was then calculated as follows: The command is then sent to the energy storage system via the highest priority channel, and the electric boiler is marked as "abnormal response" on the map.

[0132] Monitor the deviation between the actual execution effect and the expected effect of the instruction strategy package, trigger a source tracing and collaborative compensation mechanism based on a dynamic capability 3D map, suppress the deviation and update the resource status, including:

[0133] Real-time acquisition of operating parameters at the virtual power plant grid connection point, and calculation of the actual aggregate output curve;

[0134] The actual aggregated output curve is compared with the expected output curve in the instruction strategy package to generate an output deviation signal and deviation characteristic pattern.

[0135] By matching and analyzing the deviation feature patterns with the dynamic capability 3D map, the source of the deviation can be determined and the core resource nodes causing the deviation and the related resource sets affected by them can be located.

[0136] If the source of the deviation is determined to be a local device failure, a precise compensation instruction is sent to the associated resource set, and the status marker of the failed resource node in the dynamic capability 3D map is updated.

[0137] If the source of the deviation is determined to be a transient disturbance that is not due to equipment failure, the current control command remains unchanged, and the deviation characteristic pattern is recorded in the deviation event log.

[0138] First, high-frequency deviation monitoring and feature extraction are performed. Using phasor measurement units (PMUs) or high-precision energy meters deployed at the common grid connection point (PCC) connecting the virtual power plant and the main grid, the operating parameters of the grid connection point, primarily total active power, are collected in real time at millisecond to second rates. These collected values ​​are serialized to form the actual aggregated power output curve. Simultaneously, the expected power output curve on the same time axis is extracted from the issued instruction strategy package. By subtracting the two curves point by point, a continuous power output deviation signal is generated. :

[0139] ;

[0140] in, It is the actual aggregate output at time t. This is the expected output of the command. It not only focuses on the magnitude of the deviation, but also utilizes signal processing techniques such as short-time Fourier transform or wavelet analysis to analyze the deviation signal. The analysis extracts the deviation characteristics and patterns, such as the time of occurrence, duration, rate of change, and frequency of fluctuation. These patterns are the key basis for subsequent source tracing analysis.

[0141] Deviation tracing and localization are performed based on feature matching. When the absolute value of the output deviation signal remains greater than a preset dead zone threshold (e.g., 2% of the total regulation capacity) for a certain duration, such as more than 10 seconds, the tracing-and-coordinated compensation mechanism is automatically triggered. The extracted deviation feature patterns are used as query input for pattern matching on a dynamic capacity 3D map. Since the map stores typical response characteristics and fault features of various heterogeneous resources, for example, distributed photovoltaic power generation due to cloud cover will exhibit a gradual decrease over minutes, while inverter disconnection will exhibit a step-like drop over milliseconds.

[0142] By matching the most similar feature patterns, the source of the deviation can be identified with a very high probability, and the core resource node causing the deviation can be directly located on the map's topology. Simultaneously, based on the resource response characteristics and network associations, the set of associated resources most closely electrically coupled to that node can be identified. Finally, differentiated collaborative compensation and state updates are performed.

[0143] Based on the source tracing results, two different processing strategies are adopted. If the source of the deviation is determined to be a local device failure, such as a BMS failure and offline status of an energy storage unit, collaborative compensation will be initiated immediately. A precise compensation command will be sent to the previously identified associated resource set, such as the downstream controllable air conditioning cluster. The power of the command will be exactly equal to the power deficit caused by the failed device.

[0144] Simultaneously, the status marker of the failed resource node in the dynamic capability 3D map is updated to "faulty" or "unavailable" to ensure its removal in the next round of scheduling optimization. If the source of the deviation is a transient disturbance not related to equipment failure, such as a small-scale fluctuation in illumination, its deviation characteristic pattern is usually characterized by small amplitude, short duration, and self-recovery. In this case, no active intervention is required, and the currently issued control commands remain unchanged to avoid frequent disturbances to the system. However, the deviation characteristic pattern of this event will be fully recorded in the deviation event log for subsequent offline optimization and self-learning of the model.

[0145] Through a closed-loop process of deviation monitoring, source tracing, and collaborative compensation, the virtual power plant possesses real-time self-diagnosis and repair capabilities. This mechanism can accurately distinguish between hard equipment faults and external environmental disturbances, and adopt optimal response strategies. It can maintain the power response commitment to the outside world through rapid collaborative compensation, and ensure the continuous accuracy of the internal control model by dynamically updating resource status, thereby improving the overall robustness of the virtual power plant's operation and the execution accuracy of dispatch commands. Figure 4 This demonstrates the dynamic response process by which a virtual power plant, after detecting a deviation of actual output from the expected curve, automatically invokes related resources through a source tracing and collaborative compensation mechanism to smooth out the deviation, thereby enabling the grid-connected point to aggregate output and re-lock onto the expected target.

[0146] One application example is real-time monitoring at the grid connection point (PCC) of a virtual power plant. At time t, the actual aggregated output is collected by the phasor measurement unit. The output is 220kW, while the expected output in the instruction strategy package is... When the power is 250kW, the deviation calculation formula is used. The output deviation signal is calculated. The deviation was -30kW. Since the magnitude of this deviation exceeded the preset dead zone threshold and exhibited a step-down characteristic at the millisecond level, the deviation characteristic pattern was matched and analyzed with the dynamic capability 3D map. It was determined that the source of the deviation was a local equipment failure event of the photovoltaic inverter disconnecting from the grid in the resource cluster. The core resource node that failed and the associated energy storage resource set with the closest electrical coupling with it were located. Subsequently, a precise compensation strategy was executed. The failed photovoltaic node was marked as "unavailable" in the dynamic capability 3D map, and a precise compensation command of +30kW was issued to the associated energy storage resource set according to the power gap. This quickly smoothed out the output deviation in the local area and ensured the stability and robustness of the overall response of the virtual power plant.

[0147] If the deviation is determined to originate from a local device failure, a precise compensation command is sent to the associated resource set, and the status markers of the failed resource nodes in the dynamic capability 3D map are updated, including:

[0148] Based on the dynamic capability 3D map, obtain the electrical distance and adjustment capability weight between core resource nodes and related resource sets;

[0149] Based on the weights of electrical distance and regulation capability, the power gap to be compensated is dynamically allocated to some or all of the resources in the associated resource set, generating differentiated compensation instructions;

[0150] Issue differentiated compensation instructions to enable related resource sets to coordinate actions to smooth out output deviation signals.

[0151] First, precise invocation of compensation resource parameters is performed. After locating the core resource node (i.e., the failed device) and the set of associated resources affected by it, a deep data query is performed on the dynamic capability 3D map using these two objects as indexes. Key parameters extracted include the electrical distance between the core resource node and each available resource in the associated resource set, quantified by the voltage sensitivity factor S; simultaneously, the current real-time adjustable power margin of each resource in the associated resource set is also extracted. Information such as response rate R and preset adjustment cost C, which are weighted by adjustment capabilities, forms the basis for differentiated allocation decisions.

[0152] Perform dynamic allocation calculations to compensate for power gaps. Assume the power gap to be compensated due to the failure of a core resource node is... This gap needs to be allocated to n resources in the associated resource set. The compensation power allocated to the j-th resource... It is through a comprehensive performance index It is derived through weighted calculation. The formula for calculating this indicator is:

[0153] ;

[0154] in, Let be the comprehensive performance index of the j-th resource. , , These are the normalized values ​​of the voltage sensitivity of the j-th resource relative to the failed node, the current available power margin, and the response rate. These values ​​are obtained by mapping the original physical quantities to the range of 0 to 1 to eliminate the influence of dimensions. , , These are preset weighting coefficients, and the sum of the three is 1. They can be dynamically adjusted according to the control strategy. For example, in scenarios requiring rapid response, they can be temporarily adjusted. The weight is increased from 0.3 to 0.6. After calculating the comprehensive performance index of all available resources, the specific compensation power allocated to the j-th resource is:

[0155] ;

[0156] in, This is the sum of the comprehensive performance indicators of all available resources K in the associated resource set. Using this formula, resources with shorter electrical distances, larger adjustment margins, and faster response times will automatically undertake more compensation tasks, forming differentiated compensation instructions. Finally, differentiated compensation instructions are generated and issued. The calculated compensation indicators for each resource undertaking a compensation task will be... Values ​​are encapsulated into control messages conforming to their terminal communication protocols, such as "Energy storage unit ES-05, linearly increase active power output by 2.1 MW within the next 5 seconds." These instructions are marked as high priority and concurrently distributed to various resource terminals in the associated resource set via a dedicated scheduling channel. The timestamps or start signals included in the instructions ensure that all resources can act in concert, and their aggregation effect will precisely smooth out the output deviation signal caused by equipment failure at the grid connection point, restoring the overall output of the virtual power plant to the expected level.

[0157] One application example is a virtual power plant identifying a power shortfall that needs to be compensated due to the failure of a core resource node. For a power rating of 30kW, the parameters of resource 1 and resource 2 within the associated resource set are first obtained from the dynamic capability 3D map. The normalized voltage sensitivity, available power margin, and response rate indices are then set as follows: , , and resource 2 , , The weighting coefficients are set as follows: , , Substitute into the formula The comprehensive performance index of resource 1 was calculated. Comprehensive performance indicators of resource 2 Then, according to the formula Dynamic power allocation is performed, and the compensation power allocated to resource 1 is calculated. The compensation power allocated to resource 2 Finally, the calculated precise values ​​are encapsulated into differentiated compensation commands and issued concurrently, enabling Resource 1 and Resource 2 to work together to smooth out the output deviation signal and achieve precise self-healing with minimal grid disturbance.

[0158] After suppressing deviations and updating resource status, the dynamic capability 3D map is updated, including:

[0159] Record each deviation event triggered by the source tracing and collaborative compensation mechanism, the compensation strategy, and the final effect to form a deviation event log;

[0160] Based on the deviation event log, the response characteristic parameters of the corresponding resource nodes in the dynamic capability 3D map are corrected, and the corrected response characteristic parameters are generated.

[0161] Update the dynamic capability stereo map for the next cycle using the corrected response characteristic parameters to improve the map's prediction accuracy.

[0162] First, the structured recording of deviation events is executed. Each time the source tracing and collaborative compensation mechanism is triggered and successfully handles a deviation event, the key information of this event is automatically structured and stored, forming a new deviation event log entry. This entry contains at least the following fields: unique event ID, trigger timestamp, initial deviation characteristic pattern, core resource node located by the system, the final compensation strategy adopted (including the associated resource sets invoked and their respective compensation amounts), the actual response time of the compensation action, and the final evaluation of the output deviation suppression effect. This log library constitutes a valuable data asset containing rich operational experience.

[0163] Offline calibration of response characteristic parameters is performed based on log data. An offline calibration program is initiated at a fixed period, such as weekly. This program iterates through all events recorded in the deviation event log within the past period. For each event, the program focuses on the difference between the model parameters of the core resource nodes before the event and their actual performance at the time of the event. For example, if the log shows that a certain energy storage unit's actual power ramp-up rate after receiving a compensation command... Far below its rated response characteristic parameters registered in the dynamic capability stereo map. Based on data from multiple similar events, parameter identification algorithms such as recursive least squares or gradient descent are used to re-estimate the response characteristic parameters of the resource node, generating corrected response characteristic parameters. The correction process is as follows:

[0164] ;

[0165] in, The learning rate, or Kalman gain, determines the degree of influence of a single actual observation on the model parameter update. Its value typically ranges from 0.01 to 0.1 to ensure the stability of the learning process. This process applies not only to the response rate but also to the correction of other key parameters such as available capacity and adjustment cost.

[0166] Finally, the dynamic capability 3D map is updated using the corrected parameters. After the offline calibration program completes the parameter reassessment of all relevant resource nodes, a parameter update package is generated. During a low-business period, such as early morning, this update package, containing all corrected response characteristic parameters, is applied in batches to the underlying database of the dynamic capability 3D map, overwriting old parameters that may have been biased. This update marks the end of a complete learning cycle. In the next scheduling cycle, when the system builds or updates the dynamic capability 3D map, it will directly use these parameters, which have been corrected based on actual operational data and are closer to the real physical characteristics.

[0167] One application example is when the system records a deviation event log for energy storage unit ES-01 participating in collaborative compensation, showing the actual average power ramp-up rate of that unit after receiving the instruction. for The response characteristic parameters of this node are currently stored in the dynamic capability 3D map. for Set the learning rate during the offline correction phase. The value is 0.08, which is then substituted into the parameter correction formula. Calculations are performed to obtain the corrected response characteristic parameters. This value is then overwritten into the underlying map database in subsequent parameter update packages, ensuring that the quantitative evaluation in the next scheduling cycle can be calculated based on parameters that better reflect the actual physical performance. This achieves a closed-loop self-evolution of the evaluation model from historical bias to improved future prediction capabilities.

[0168] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can make equivalent substitutions or modifications based on the technical solution and concept of the present invention within the scope of the technology disclosed in the present invention, and such modifications should also be considered to fall within the scope of protection of the present invention.

Claims

1. A method for dynamic quantitative evaluation and multi-level linkage closed-loop control of virtual power plant regulation capacity, characterized in that, Includes the following steps: Based on the real-time physical state parameters of various heterogeneous resources within the virtual power plant, a dynamic three-dimensional map characterizing the synergistic adjustment potential among heterogeneous resources is constructed and continuously updated. Heterogeneous resources are distributed energy resources with different response times, output characteristics, and physical constraints, including distributed photovoltaics, energy storage devices, and controllable loads. The system receives total power regulation instructions from the power grid dispatch center and, based on a dynamic capability 3D map, decomposes the total power regulation instructions into several control elements that match the collaborative resource clusters. At the same time, it predicts the execution risk of each control element. The collaborative resource cluster is a set of distributed resources with collaborative regulation capabilities selected from the virtual power plant based on the electrical distance, response sensitivity, and regulation direction between heterogeneous resources in the dynamic capability 3D map. Several control primitives are injected into a virtual power grid simulation environment to model the real power grid topology and state for parallel security simulation, generating an instruction strategy package containing main power control instructions and supporting security fine-tuning instructions, including: Several control primitives and anticipated potential risk points are injected into the virtual power grid simulation environment; Transient power flow calculations are performed in a virtual power grid simulation environment to simulate power flow and equipment status changes after the execution of control primitives, and to generate safety risk simulation results. If the safety risk simulation results indicate the existence of a safety risk, the system will automatically retrieve and call up safety adjustment resources that can eliminate the risk from the dynamic capability 3D map, and generate a matching safety fine-tuning instruction bound to the control primitive with the risk. If the safety risk simulation results indicate that there is no safety risk, then several control primitives will be directly used as the main power control command. All control primitives and their corresponding safety fine-tuning instructions are encapsulated to form an instruction policy package; The instruction policy package is sent to the corresponding heterogeneous resource terminal for execution; Monitor the deviation between the actual execution effect and the expected effect of the instruction strategy package, trigger a source tracing and collaborative compensation mechanism based on a dynamic capability 3D map, suppress the deviation and update the resource status, including: Real-time acquisition of operating parameters at the virtual power plant grid connection point, and calculation of the actual aggregate output curve; The actual aggregated output curve is compared with the expected output curve in the instruction strategy package to generate an output deviation signal and deviation characteristic pattern. By matching and analyzing the deviation feature patterns with the dynamic capability 3D map, the source of the deviation can be determined and the core resource nodes causing the deviation and the related resource sets affected by them can be located. If the source of the deviation is determined to be a local device failure, a precise compensation instruction is sent to the associated resource set, and the status marker of the failed resource node in the dynamic capability 3D map is updated. If the source of the deviation is determined to be a transient disturbance that is not due to equipment failure, the current control command remains unchanged, and the deviation characteristic pattern is recorded in the deviation event log.

2. The method for dynamic quantitative evaluation and multi-level linkage closed-loop control of virtual power plant regulation capacity according to claim 1, characterized in that, Constructing and continuously updating a dynamic capability 3D map characterizing the synergistic adjustment potential among heterogeneous resources includes: Obtain real-time physical state parameters and operating condition parameters of the virtual power plant. The real-time physical state parameters include the real-time power and electricity of distributed photovoltaic, energy storage devices and controllable loads. Based on the equivalent physical model and coupling constraints of each resource type, and combined with real-time power, energy and operating condition parameters, a resource response feature association network that characterizes the response influence relationship between resources is calculated and generated. By integrating real-time physical state parameters with resource response characteristics into a network, a dynamic capability 3D map is formed and updated in seconds. This map can query the optimal collaborative compensation resource objects and paths when any resource state changes.

3. The method for dynamic quantitative evaluation and multi-level linkage closed-loop control of virtual power plant regulation capacity according to claim 1, characterized in that, The total power adjustment command is decomposed into several control primitives that match the cooperative resource cluster, and the execution risks of each control primitive are predicted, including: Receive total power adjustment command and extract the target power value and adjustment direction from the command; Query the dynamic capability 3D map at the current moment to identify several candidate resource clusters that meet the requirements of adjustment direction and power margin; The target power value is allocated to several candidate resource clusters according to optimization rules to form several control elements. The physical association information in the dynamic capability 3D map is used to predict the risk of local power grid parameter overruns that may be caused by the execution of each control element.

4. The method for dynamic quantitative evaluation and multi-level linkage closed-loop control of virtual power plant regulation capacity according to claim 1, characterized in that, The accompanying safety fine-tuning instructions generated and bound to the control primitives that pose risks include: Analysis of the safety risk simulation results determined that the type of safety risk was either voltage over-limit or line overload. To address the risk of voltage exceeding limits, resources with reactive power regulation capabilities near the risk point are retrieved from the dynamic capability 3D map, and reactive power regulation fine-tuning instructions are generated. To address the risk of line overload, backup resource paths for power flow transfer are retrieved from the dynamic capacity 3D map, and power transfer fine-tuning instructions are generated. Bind the reactive power regulation fine-tuning command or power transfer fine-tuning command to the control element that caused the risk in terms of time and logic to ensure that the two are issued and executed synchronously.

5. The method for dynamic quantitative evaluation and multi-level linkage closed-loop control of virtual power plant regulation capacity according to claim 1, characterized in that, After issuing the instruction policy package, perform the following operations: Within a time window, collect execution confirmation signals from each resource terminal; If no execution confirmation signal for the critical resource is received within the time window, the execution of the resource instruction is deemed abnormal, and an instruction execution abnormality event is generated. Based on abnormal command execution events, alternative resources are selected from the reserve resource pool using a dynamic capability 3D map, and an emergency switchover command is generated and issued.

6. The method for dynamic quantitative evaluation and multi-level linkage closed-loop control of virtual power plant regulation capacity according to claim 1, characterized in that, Sending precise compensation instructions to the associated resource set and updating the status markers of failed resource nodes in the dynamic capability 3D map includes: Based on the dynamic capability 3D map, obtain the electrical distance and adjustment capability weight between core resource nodes and related resource sets; Based on the weights of electrical distance and regulation capability, the power gap to be compensated is dynamically allocated to some or all of the resources in the associated resource set, generating differentiated compensation instructions; Issue differentiated compensation instructions to enable related resource sets to coordinate actions to smooth out output deviation signals.

7. The method for dynamic quantitative evaluation and multi-level linkage closed-loop control of virtual power plant regulation capacity according to claim 1, characterized in that, After suppressing deviations and updating resource status, the dynamic capability 3D map is updated, including: Record each deviation event triggered by the source tracing and collaborative compensation mechanism, the compensation strategy, and the final effect to form a deviation event log; Based on the deviation event log, the response characteristic parameters of the corresponding resource nodes in the dynamic capability 3D map are corrected, and the corrected response characteristic parameters are generated. Update the dynamic capability stereo map for the next cycle using the corrected response characteristic parameters.

8. A virtual power plant regulation capacity dynamic quantitative evaluation and multi-level linkage closed-loop control system, used to implement the method of any one of claims 1-7, characterized in that, include: Map building and updating module: Based on the real-time physical state parameters of various heterogeneous resources in the virtual power plant, it builds and continuously updates a dynamic three-dimensional map that characterizes the potential for coordinated regulation among heterogeneous resources. Heterogeneous resources are distributed energy resources with different response times, output characteristics and physical constraints, including distributed photovoltaics, energy storage devices and controllable loads. Command decomposition and risk prediction module: It is used to receive the total power regulation command issued by the power grid dispatch center, and decompose the total power regulation command into several control elements that match the collaborative resource cluster based on the dynamic capability 3D map. At the same time, it predicts the execution risk of each control element. The collaborative resource cluster is a set of distributed resources with collaborative regulation capabilities selected from the virtual power plant based on the electrical distance, response sensitivity and regulation direction between heterogeneous resources in the dynamic capability 3D map. Safety simulation and strategy package generation module: This module is used to inject several control primitives into a virtual power grid simulation environment that simulates the real power grid topology and state for parallel safety simulation, and generate an instruction strategy package that includes main power control instructions and supporting safety fine-tuning instructions. Command issuance and execution module: used to issue command policy packages to the corresponding heterogeneous resource terminals for execution; Deviation Monitoring and Collaborative Compensation Module: Used to monitor the deviation between the actual execution effect and the expected effect of the instruction strategy package. Based on the dynamic capability 3D map, it triggers a source tracing and collaborative compensation mechanism to suppress the deviation and update the resource status.

Images