A Method and System for Coordinated Control of Power Plant Side Intelligent Agents Based on Disturbance Sensitivity

By adopting a power plant-side intelligent agent coordination control method based on disturbance sensitivity, the multi-level coordination problem of distributed resources in new power systems is solved, enabling precise regulation and safety assurance in extreme environments, and improving economic efficiency and resource sustainability.

CN123121670APending Publication Date: 2026-11-17STATE GRID JIANGSU ELECTRIC POWER CO LTD MARKETING SERVICE CENT +2
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202611439612.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-16
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In new power systems, the spatial dispersion, temporal randomness, and physical characteristic deviation of distributed flexible resources make it difficult to achieve precise control through multi-level coordination between the master station, power plants, and resource side. Existing technologies are prone to power quality deterioration and safety risks under extreme environments, and are also economically inefficient.

Method used

A power plant-side intelligent agent coordinated control method based on disturbance sensitivity is adopted. By injecting small signal disturbances into controllable nodes, the voltage-active and reactive power sensitivity matrices are calculated, and a least squares optimization model is constructed. Combined with voltage safety constraints and energy storage SOC management, reactive power priority adjustment and active power compensation hierarchical control are realized, zero marginal cost resources are prioritized, and a daily SOC regression mechanism is introduced.

Benefits of technology

It enables precise response to the control targets of the main station in extreme environments, ensuring grid security, reducing curtailment losses, improving economic efficiency, ensuring the continuous regulation capability of energy storage resources, reducing communication delays, and supporting efficient collaborative operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123121670A_ABST
    Figure CN123121670A_ABST
Patent Text Reader

Abstract

A method and system for coordinated control of power plant-side intelligent agents based on disturbance sensitivity includes: injecting disturbances at each controllable node, collecting voltage steady-state response data of target nodes across the entire network, and calculating a sensitivity matrix; when the master station issues a reactive power regulation command, the power plant-side intelligent agent constructs a reactive power optimization target based on the sensitivity matrix, solves for the optimal reactive power regulation increment, and issues it to each controllable node for execution; calculating the predicted voltage of each target node, and performing active power regulation if there is still a risk of voltage exceeding the limit; using the predicted voltage of each target node as a reference voltage, and combining it with the active power regulation command, constructing an active power optimization target, solving for the active power regulation amount of each controllable node, and issuing it to each controllable node; when the power grid is in a non-emergency state and the deviation between the active power regulation command and the power of the power plant tie line is less than a preset tolerance threshold, the SOC of the energy storage node is returned to the preset SOC threshold. This invention can support the efficient coordination and stable operation of new power systems under complex operating conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of power system operation and control technology, and particularly relates to a method and system for coordinated control of power plant-side intelligent agents based on disturbance sensitivity. Background Technology

[0002] Against the backdrop of the accelerated construction of new power systems, extreme weather scenarios, such as "extreme heat and no wind" and "no sunlight during evening peak hours," are occurring frequently, posing unprecedented challenges to the safe and stable operation of power systems. In this context, large-scale distributed and flexible resources such as distributed generation, electrochemical energy storage, and temperature-controlled loads have become key supports for coping with extreme conditions and ensuring power supply. However, these resources are highly dispersed spatially, exhibit strong randomness and volatility in time, and their physical characteristics deviate significantly under extreme environments. This uncertainty on both the "source-load" and "source" sides makes multi-level coordination between the main station, power plants, and resource sides extremely difficult. Global control targets issued by the main station are often difficult to execute accurately at the power plant level, and blindly responding to main station targets while ignoring the power plant status can easily deteriorate power quality at the power plant level and even affect grid security.

[0003] Currently, there are three main technical approaches for coordinated control of distributed flexible resources, but all have limitations in addressing the aforementioned extreme conditions and multi-level coordination requirements. The first is a centralized optimization method based on a global physical model, such as optimal power flow. While theoretically achieving network-wide optimization, this method heavily relies on precise grid topology and model parameters. Under extreme weather conditions, the grid topology is prone to temporary changes, and physical parameters drift drastically with the environment, leading to model parameter distortion. Control strategies based on precise sensitivity calculations are easily rendered ineffective, even triggering regulation oscillations. The second is a distributed control method based on local rules, such as inverter droop control. This method offers fast response and does not rely on communication, but lacks coordination. Each resource responds only to the overall objective, leading to mutually canceling or over-adjusting regulation actions. This is not only economically inefficient but also often fails to guarantee the effectiveness of the global control objectives issued by the master station. The third is the traditional aggregation control method. Existing plant-side aggregators often use capacity superposition to allocate master station commands, ignoring the complex network constraints and resource coupling relationships within the plant. This easily triggers local safety protection when executing high-power commands from the master station, causing control failure.

[0004] Chinese patent application CN120377289A discloses a distribution network voltage control method based on data-driven power flow and sensitivity analysis. This method constructs a distribution network voltage control model based on sensitivity analysis values ​​to achieve effective control of the distribution network voltage. However, it still has the following shortcomings: The core of this method is a neural network regression model trained based on historical power flow data. In extreme scenarios such as "extreme heat and no wind," or when a grid fault causes temporary topology changes and drastic parameter drift, the historical data becomes invalid, leading to severe distortion of the analyzed sensitivity, which in turn causes control failure or even grid oscillation. This method only focuses on the voltage control model of the distribution network itself and does not consider how to coordinate with the global dispatch instructions issued by the master station under the new power system, failing to solve the problem of local safety limit violations easily triggered by the substation when executing master station instructions. This method does not perform long-term management of the dynamic physical boundaries of the control resources during voltage control, which can easily lead to the loss of regulation capacity of energy storage due to overcharging / over-discharging during continuous control. Furthermore, it does not distinguish the order of call for zero-cost reactive power and high-cost active power (such as curtailed solar power), resulting in poor overall operational economy.

[0005] In summary, existing technologies urgently require a plant-side intelligent agent coordination control method based on the master station's control objectives, which no longer relies on physical model parameters, can perceive local security boundaries in real time, and can achieve accurate response to the master station's control objectives, in order to support the efficient coordination and stable operation of new power systems under extremely complex operating conditions. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a plant-side intelligent agent coordination control method and system based on disturbance sensitivity.

[0007] The present invention adopts the following technical solution.

[0008] In a first aspect, the present invention discloses a plant-side intelligent agent coordinated control method based on disturbance sensitivity, the method comprising the following steps: Step 1: Periodically inject active and reactive power disturbances not exceeding the preset disturbance thresholds at each controllable node managed by the intelligent agent on the power plant side, and collect voltage steady-state response data of the target nodes of the entire network. Calculate the voltage-active power sensitivity matrix and voltage-reactive power sensitivity matrix for the current period based on the disturbance and voltage steady-state response data. Step 2: When the master station issues a reactive power regulation command to the plant-side intelligent agent, the plant-side intelligent agent constructs a reactive power optimization target by combining the voltage-reactive power sensitivity matrix of the current cycle, solves the optimal reactive power regulation increment, and issues it to each controllable node for execution; the predicted voltage of each target node is calculated based on the optimal reactive power regulation increment. If there is still a risk of node voltage exceeding the limit after reactive power regulation, then proceed to step 3 for active power regulation. Step 3: Using the predicted voltage of each target node as the reference voltage, construct a voltage safety hard constraint based on the voltage-active power sensitivity matrix, and combine the active power adjustment command issued by the master station to construct the active power optimization target and solve for the active power adjustment amount of each controllable node. Then, send the active power adjustment amount to each controllable node for active power adjustment. Step 4: When the intelligent agent detects that the power grid is in a non-emergency state and the deviation between the active power regulation command issued by the master station and the power of the plant tie line is less than the preset tolerance threshold, the SOC of the energy storage node is smoothly returned to the preset SOC threshold.

[0009] More preferably, In step 2, the reactive power regulation instructions include the grid connection point target voltage regulation instructions and the plant-wide reactive power regulation instructions; for different reactive power regulation instructions, corresponding reactive power optimization targets and constraints are constructed respectively.

[0010] More preferably, When the master station issues a target voltage regulation command for the grid connection point to the power plant-side intelligent agent, its reactive power optimization target is:

[0011] in, This is a subset of the row vectors corresponding to the nodes that exceed the limit in the voltage-reactive power sensitivity matrix. To find the reactive power adjustment increment vector for each controllable node, we need to determine the solution. This is the column vector of the desired adjustment amount of the target node voltage; When the master station issues a reactive power adjustment command for the entire plant to the plant-side intelligent agent, its reactive power optimization objective is:

[0012] in, The sum of reactive power commands issued by the master station; Ω represents the set of all controllable nodes; For the first j The reactive power adjustment increment of each controllable node.

[0013] More preferably, In step 3, the voltage safety hard constraint specifically refers to:

[0014] in, Predict voltage vectors for each target node; This is the voltage-active power sensitivity matrix; , These are the active power regulation vectors of each energy storage node in the controllable node and the active power reduction vectors of each distributed power generation node, respectively. , These are the lower and upper limits of the safe operating voltage for each target node, respectively.

[0015] More preferably, In step 3, the active power optimization objective is specifically:

[0016] in, This represents the active power regulation of energy storage at each energy storage node. This represents the active power reduction of each distributed power node. The current reference switching power; Active power adjustment instructions issued by the master station; The penalty weights applied to the reduction of distributed power sources.

[0017] More preferably, In step 4, the step of smoothly regressing the SOC of the energy storage node to a preset SOC threshold specifically includes: calculating the average charging and discharging power of the energy storage node based on the current SOC and the preset SOC threshold, combined with the rated capacity of the energy storage and the remaining time of the day; constructing a voltage-constrained SOC regression model and solving it to obtain the active power adjustment of the energy storage node at each time.

[0018] More preferably, The objective function of the SOC regression model is as follows:

[0019] in, This represents the actual active power output vector of each energy storage node at the current moment; This is the adjustment increment vector for the active power of each energy storage node; This represents the average charge and discharge power vector for each energy storage node. The average charge and discharge power of each energy storage node is determined as follows:

[0020] in, The current moment; The preset SOC threshold; The current state of charge (SOC); This refers to the rated capacity of the energy storage. This represents the average charge and discharge power of the energy storage node.

[0021] Secondly, the present invention discloses a power plant-side intelligent agent coordination control system based on the aforementioned method, including a sensitivity matrix construction module, a reactive power regulation module, an active power regulation module, and an energy storage node SOC regression module. The sensitivity matrix construction module periodically injects active and reactive power disturbances not exceeding a preset disturbance threshold at each controllable node managed by the intelligent agent on the power plant side, and collects voltage steady-state response data of target nodes across the entire network. Based on the disturbance and voltage steady-state response data, it calculates the voltage-active power sensitivity matrix and voltage-reactive power sensitivity matrix for the current period. The reactive power regulation module, when the master station sends a reactive power regulation command to the plant-side intelligent agent, the plant-side intelligent agent constructs a reactive power optimization target by combining the voltage-reactive power sensitivity matrix of the current cycle, solves the optimal reactive power regulation increment, and sends it to each controllable node for execution; based on the optimal reactive power regulation increment, it calculates the predicted voltage of each target node; if there is still a risk of node voltage exceeding the limit after reactive power regulation, it proceeds to step 3 for active power regulation. The active power regulation module uses the predicted voltage of each target node as the reference voltage, constructs a voltage safety hard constraint based on the voltage-active power sensitivity matrix, and combines the active power regulation command issued by the master station to construct the active power optimization target and solve for the active power regulation amount of each controllable node. The active power regulation amount is then sent to each controllable node for active power regulation. The SOC regression module for energy storage nodes smoothly regresses the SOC of the energy storage node to the preset SOC threshold when the intelligent agent detects that the power grid is in a non-emergency state and the deviation between the active power regulation command issued by the master station and the power of the plant tie line is less than the preset tolerance threshold.

[0022] Thirdly, the present invention provides a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of the first aspects of the present invention.

[0023] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the first aspects of the present invention.

[0024] The beneficial effects of this invention are compared with those of the prior art: 1. To address the issue of power grid topology changes or drastic parameter drifts that may be caused by extreme weather, a small-signal perturbation method is adopted to obtain the real-time sensitivity of flexible resources in a wide area. This method breaks through the dependence of traditional methods on topology and physical parameters, and can accurately perceive local safety boundaries, making it suitable for the control of new power systems under extreme conditions.

[0025] 2. A least-squares optimization model with safety constraints was constructed, overcoming the limitation of traditional methods that struggle to balance responding to superior commands and ensuring local safety. By taking the master station's control target as the tracking object and considering local voltage safety constraints, the model can find the control solution that best approximates the master station command while strictly ensuring the safety of the plant, thus solving the pain point of the plant-side intelligent agent easily triggering over-limit protection when executing high-power commands. An innovative hierarchical control logic of prioritizing reactive power voltage regulation and providing active power compensation as a safety net is implemented. Reactive power resources are prioritized to smooth voltage deviations, and when reactive power regulation is insufficient, energy storage and photovoltaic active power are used for coordinated voltage regulation, significantly reducing the loss of renewable energy curtailment.

[0026] 3. A tiered, weighted resource allocation strategy is adopted. When responding to master station commands, priority is given to allocating zero-marginal-cost distributed generation reactive power resources and low-cost energy storage time-shift resources, with distributed generation active power reduction only implemented when absolutely necessary. Compared to traditional simple threshold-based or proportional allocation methods, this invention minimizes curtailment and improves the overall economic efficiency of power plant operations.

[0027] 4. Addressing the characteristic of energy storage as a confined energy resource, a daily SOC (State of Charge) regression mechanism is introduced. Utilizing non-emergency grid conditions or periods of relaxed master station commands, the energy storage's state of charge is automatically restored to a stable threshold through trajectory tracking optimization. This mechanism prevents the energy storage from losing its regulatory capacity due to depletion or full charging under continuous extreme operating conditions, ensuring the system's continuous emergency response capability in the future.

[0028] 5. By adopting local computing and real-time control based on plant-side intelligent agents, the latency and congestion caused by direct communication between a large number of underlying devices and the master station are avoided, ensuring millisecond-level response and accurate execution of master station control commands, and supporting the efficient operation of the "master station-plant-resource" three-layer architecture. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating the plant-side intelligent agent coordinated control method based on disturbance sensitivity according to the present invention. Figure 2 This is a diagram illustrating the effect of local node voltage constraint control on the power plant-side intelligent agent in response to the control target of the master station in Embodiment 2 of the present invention. Figure 3 This is a diagram showing the coordinated charging and discharging power of the energy storage system in response to the master station's commands within one day in Embodiment 2 of the present invention; Figure 4 This is a graph showing the change in the state of charge of the energy storage system within a day and its return trajectory at the end of the day in Embodiment 2 of the present invention. Figure 5 This is a comparison chart of typical adjustable point voltage-active power sensitivity online measurement values ​​and theoretical values ​​in Embodiment 2 of the present invention; Figure 6This is a scatter plot of the overall sensitivity prediction deviation in Embodiment 2 of the present invention. Detailed Implementation

[0030] 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 of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0031] like Figure 1 As shown, this invention discloses a plant-side intelligent agent coordinated control method based on disturbance sensitivity, specifically including the following steps: Step 1: Periodically inject active and reactive power disturbances not exceeding the preset disturbance thresholds at each controllable node managed by the intelligent agent on the power plant side, and collect voltage steady-state response data of the target nodes of the entire network. Calculate the voltage-active power sensitivity matrix and voltage-reactive power sensitivity matrix for the current period based on the disturbance and voltage steady-state response data. The preset disturbance threshold is selected based on a comprehensive consideration of the power grid's normal operation safety, voltage response sampling signal-to-noise ratio, and sensitivity matrix linear identification accuracy. This avoids both excessive disturbances affecting the power grid's steady-state operation and insufficient disturbances causing voltage response signal distortion. In this invention, a preferred threshold is... pu.

[0032] Step 2: When the master station issues a reactive power regulation command to the plant-side intelligent agent, the plant-side intelligent agent constructs a reactive power optimization target by combining the voltage-reactive power sensitivity matrix of the current cycle, solves the optimal reactive power regulation increment, and issues it to each controllable node for execution; the predicted voltage of each target node is calculated based on the optimal reactive power regulation increment. If there is still a risk of node voltage exceeding the limit after reactive power regulation, then proceed to step 3 for active power regulation. The reactive power regulation instructions include the grid connection point target voltage regulation instructions and the plant-wide reactive power regulation instructions; for different reactive power regulation instructions, corresponding reactive power optimization targets and constraints are constructed respectively.

[0033] When the master station issues a target voltage regulation command for the grid connection point to the power plant-side intelligent agent, its reactive power optimization target is:

[0034] in, This is a subset of the row vectors corresponding to the nodes that exceed the limit in the voltage-reactive power sensitivity matrix. To find the reactive power adjustment increment vector for each controllable node, we need to determine the solution. This is the column vector of the desired adjustment amount of the target node voltage; When the master station issues a reactive power adjustment command for the entire plant to the plant-side intelligent agent, its reactive power optimization objective is:

[0035] in, The sum of reactive power commands issued by the master station; Ω represents the set of all controllable nodes; For the first j The reactive power adjustment increment of each controllable node.

[0036] Step 3: Using the predicted voltage of each target node as the reference voltage, construct a voltage safety hard constraint based on the voltage-active power sensitivity matrix, and combine the active power adjustment command issued by the master station to construct the active power optimization target and solve for the active power adjustment amount of each controllable node. Then, send the active power adjustment amount to each controllable node for active power adjustment. The voltage safety hard constraint is specifically as follows:

[0037] in, Predict voltage vectors for each target node; This is the voltage-active power sensitivity matrix; , These are the active power regulation vectors of each energy storage node in the controllable node and the active power reduction vectors of each distributed power generation node, respectively. , These are the lower and upper limits of the safe operating voltage for each target node, respectively.

[0038] The active power optimization objective is specifically as follows:

[0039] in, This represents the active power regulation of energy storage at each energy storage node. This represents the active power reduction of each distributed power node. The current reference switching power; Active power adjustment instructions issued by the master station; The penalty weights applied to the reduction of distributed power sources.

[0040] The penalty weight satisfy That is, the penalty weight is a normal number much greater than 1. By constraining the active power reduction amplitude of distributed power sources through the large-weight penalty term, the active power resources of energy storage are prioritized to complete the voltage safety bottom-line regulation, thereby minimizing the curtailment loss of new energy.

[0041] Step 4: When the intelligent agent detects that the power grid is in a non-emergency state and the deviation between the active power regulation command issued by the master station and the power of the plant tie line is less than the preset tolerance threshold, the SOC of the energy storage node is smoothly returned to the preset SOC threshold.

[0042] The step of smoothly regressing the SOC of the energy storage node to a preset SOC threshold specifically includes: calculating the average charging and discharging power of the energy storage node based on the current SOC and the preset SOC threshold, combined with the rated capacity of the energy storage and the remaining time of the day; constructing a voltage-constrained SOC regression model and solving it to obtain the active power adjustment of the energy storage node at each time.

[0043] The objective function of the SOC regression model is as follows:

[0044] in, This represents the actual active power output vector of each energy storage node at the current moment; This is the adjustment increment vector for the active power of each energy storage node; This represents the average charge and discharge power vector for each energy storage node. The average charge and discharge power of each energy storage node is determined as follows:

[0045] in, The current moment; The preset SOC threshold is the target SOC. The current state of charge (SOC); This refers to the rated capacity of the energy storage. This represents the average charge and discharge power of the energy storage node.

[0046] Example 1: like Figure 1 As shown, this invention discloses a plant-side intelligent agent coordinated control method based on disturbance sensitivity, specifically including the following steps: Step 1: Periodically inject active and reactive power disturbances not exceeding the preset disturbance thresholds at each controllable node managed by the intelligent agent on the power plant side, and collect voltage steady-state response data of the target nodes of the entire network. Calculate the voltage-active power sensitivity matrix and voltage-reactive power sensitivity matrix for the current period based on the disturbance and voltage steady-state response data. The preset disturbance threshold is selected based on a comprehensive consideration of the power grid's normal operation safety, voltage response sampling signal-to-noise ratio, and sensitivity matrix linear identification accuracy. This avoids both excessive disturbances affecting the power grid's steady-state operation and insufficient disturbances causing voltage response signal distortion. In this invention, a preferred threshold is... pu.

[0047] Specifically, according to the preset update cycle (e.g., taking 15 minutes as the time step), the perturbation not exceeding the preset perturbation threshold is periodically injected into the controllable variables managed by the intelligent agent on the plant side, the voltage steady-state response data of the target nodes of the entire network is collected, the control variable sensitivity matrix under the current time step is updated using the finite difference method, and the implicit voltage safety boundary under the current operating condition is established.

[0048] The online measurement of the voltage-active power sensitivity matrix and the voltage-reactive power sensitivity matrix specifically includes: Step 1.1: Establish baseline operating status. The intelligent volume measurement device collects the current baseline operating data of the plant's internal systems and records it. The reference voltage magnitude vector of each node .

[0049] Step 1.2: Measure the voltage-active power sensitivity matrix and select each controllable node sequentially. j (e.g., photovoltaic or energy storage nodes), keeping the power of other nodes constant, only at the node j Injected active power disturbance Subsequently, the power plant-side intelligent agent continuously monitors the voltage amplitude of the target node at a set sampling frequency. When the relative fluctuation rate of the voltage amplitude is continuously less than the preset steady-state threshold (e.g., 0.1%) within a set time window (e.g., 1 second), it is determined that the power grid transient process has ended and the system has reached a new steady state, and the voltage steady-state response data after the system stabilizes is collected. Calculate the sensitivity matrix based on the finite difference principle. S P The j The formula for calculating column elements is:

[0050] in, The first in the voltage-active power sensitivity matrix i line, number j Column elements; For the first i The reference voltage of each voltage monitoring node; To give the first j After adding active power disturbance to the first controllable node, the first... i Voltage steady-state response data of each voltage monitoring node; This represents the active power disturbance.

[0051] After traversing all controllable nodes, the active power disturbance is removed, and the system is restored to the baseline state.

[0052] Step 1.3: Measure the voltage-reactive power sensitivity matrix, sequentially for each controllable node. j Injected reactive power disturbance The network voltage response vector after the intelligent agent acquisition system stabilizes. Calculate the sensitivity matrix S Q The j The formula for calculating column elements is:

[0053] in, The first in the voltage-reactive power sensitivity matrix i line, number j Column elements; To give the first j After adding reactive power disturbance to the first controllable node, the second... i Voltage steady-state response data of each voltage monitoring node; This represents the reactive power disturbance.

[0054] Finally, a full-dimensional sensitivity matrix reflecting the actual physical characteristics of the current power grid was constructed.

[0055] Step 2: When the master station issues a reactive power regulation command to the plant-side intelligent agent, the plant-side intelligent agent constructs a reactive power optimization target by combining the voltage-reactive power sensitivity matrix of the current cycle, solves the optimal reactive power regulation increment, and issues it to each controllable node for execution; the predicted voltage of each target node is calculated based on the optimal reactive power regulation increment. If there is still a risk of node voltage exceeding the limit after reactive power regulation, then proceed to step 3 for active power regulation. The plant-side intelligent agent receives reactive power regulation commands from the master station, including two modes: target voltage regulation commands at the grid connection point and reactive power regulation commands for the entire plant. Following the principle of "reactive power priority, active power backup," it prioritizes utilizing the inverter reactive power capacity of distributed generation resources in controllable nodes. Based on the voltage-reactive power sensitivity matrix, it calculates the reactive power adjustable boundary, constructs and solves a least-squares optimization model, and maximizes the response to the master station's reactive power regulation commands while ensuring voltage safety.

[0056] Specifically, reactive power coordinated optimization includes the following steps: Step 2.1: Dynamically calculate the reactive power constraint boundary of the distributed generation resource inverter; The intelligent agent reads the active power output of each distributed generation resource at the current moment. and inverter apparent power rated capacity Based on capacity constraints, computing nodes j Adjustable upper limit of reactive power and lower limit :

[0057] Step 2.2: Construct and solve the least squares model for reactive power tracking. The power plant-side agent constructs the optimization objective based on the type of reactive power adjustment command issued by the master station. If the master station issues a target voltage regulation command for the grid connection point, then the following least squares problem with linear constraints is constructed, using the reactive power regulation of distributed generation resources as the optimization variable:

[0058] in, This is a subset of the row vectors corresponding to the nodes that exceed the limit in the voltage-reactive power sensitivity matrix. For the reactive power regulation vector of the controllable node; For the first j The reactive power regulation of each controllable node; Controllable node before adjustment j Unproductive efforts; This is the column vector of the desired adjustment amount of the target node voltage.

[0059] The desired adjustment column vector of the target node voltage Specifically, the calculation follows a fusion approach that targets the main station's instructions and imposes local security as a hard constraint. If the main station issues a target voltage adjustment command for the grid connection point At the local security threshold [ , Within the range, the voltage tracking reference equal ;like If the limit is exceeded or no specific target is issued, corrections will be made for the overvoltage node to ensure the voltage tracking reference is maintained. Equal to the local security threshold upper limit Correction for undervoltage nodes, enabling voltage tracking reference. Equal to the lower limit of the local security threshold Finally, calculate the desired regulation of the target node voltage. ,in, This represents the current actual voltage vector of the over-limit node.

[0060] The optimal reactive power increment is obtained by using an optimization algorithm. The policy was issued and implemented to improve voltage distribution by utilizing low-cost reactive power resources in advance, thus reserving a safety margin for subsequent active power regulation.

[0061] When the master station issues a plant-wide reactive power adjustment command to the plant-side intelligent agent, it tracks the power deviation. To optimize the variables, the following least squares model is constructed:

[0062] in, The sum of reactive power commands issued by the master station; Ω is the set of adjustable reactive power resource nodes; For the first j The reactive power adjustment increment of each controllable node; For the reactive power adjustment increment column vector of the controllable node; This is the voltage-reactive power sensitivity matrix; This represents the voltage vector of each target node before reactive power regulation; , These are the lower and upper limits of safe operation for each voltage node, respectively. For the first j The reactive power regulation of each controllable node; Controllable node before adjustment j Unproductive efforts; , They are nodes j The reactive power has adjustable upper and lower limits.

[0063] By solving the corresponding least squares model, the optimal reactive power increment can be obtained using an optimization algorithm. And it was issued and implemented.

[0064] Step 2.3: Utilizing the voltage-reactive power sensitivity matrix S Q With optimal reactive power increment Perform prediction calculations to determine the predicted voltage vector for each target node. Is there a voltage higher than the set safety limit? or below the lower limit If a node exists, it is determined that the local node still has a risk of exceeding the voltage limit. At this point, the agent updates the predicted voltage vectors for each target node. Proceed to step 3. When residual voltage exceeds the limit, active power coordination optimization will play a backup support role; when the voltage is within the safe range, active power coordination optimization will respond to the active power dispatching instructions issued by the master station based on this safe benchmark.

[0065] The predicted voltage of each target node is calculated as follows:

[0066] Step 3: Using the predicted voltage of each target node as the reference voltage, construct a voltage safety hard constraint by combining the voltage-active power sensitivity matrix, construct an active power optimization target by combining the active power adjustment command issued by the master station, and solve for the active power adjustment amount of each controllable node. Then, send the active power adjustment amount to each controllable node for active power adjustment. The power plant-side agent receives active power regulation commands from the master station. The agent calculates the dynamic power boundary based on the real-time state of energy storage (SOC) and constructs a joint optimization model that includes energy storage regulation and distributed generation resource reduction. With the goal of minimizing the deviation from the master station's power command and with node voltage not exceeding limits as a strong constraint, the optimal active power regulation is solved and executed.

[0067] Specifically, active power coordination optimization includes the following steps: Step 3.1: Calculate the dynamic power boundary of energy storage based on the SOC state. The agent calculates the dynamic power boundary of the energy storage battery based on the current state of charge of the battery. Rated capacity Control step size and the rated power of the converter Calculate the maximum allowable discharge power of the energy storage. and maximum charging power The discharge boundary is limited by the smaller of the remaining charge and the rated power of the equipment:

[0068] The charging limit is limited by the smaller of the remaining charging capacity and the device's rated power:

[0069] in, , These represent the lower and upper limits of the state of charge (SOC) of the energy storage battery, respectively.

[0070] Step 3.2: Construct a joint optimization model in response to master station commands. This stage of active power coordinated optimization, as a step in responding to master station active power control commands, receives the predicted voltage vectors of each target node transmitted in Step 2. As the initial state reference Energy storage active power regulation and the reduction of active power generation resources ( ) are the joint optimization variables. The objective function is as follows:

[0071] in, This represents the active power regulation of energy storage at each energy storage node. This represents the active power reduction of each distributed power node. The current reference switching power; Active power adjustment instructions issued by the master station; The penalty weights applied to the reduction of distributed power sources, and satisfying That is, the penalty weight is a normal number much greater than 1. By constraining the active power reduction amplitude of distributed power sources through the large-weight penalty term, the active power resources of energy storage are prioritized to complete the voltage safety bottom-line regulation, thereby minimizing the curtailment loss of new energy.

[0072] In the joint optimization model, a voltage-active power sensitivity matrix is ​​introduced. S P A hard voltage safety constraint ensures that the regulated voltage remains within a safe range.

[0073] in, Predicted voltage at each target node after reactive power regulation; This is the voltage-active power sensitivity matrix; , These are the active power regulation of each energy storage node in the controllable node and the active power reduction vector of each distributed power node, respectively. , These are the lower and upper limits of the safe operating voltage for each target node, respectively. , These represent the lower limit of the maximum allowable charging power and the upper limit of the maximum allowable discharging power for each energy storage node in the controllable node; This is a vector of the actual active power output of each distributed power node at the current moment.

[0074] By solving this model, the agent can prioritize the use of energy storage resources and, when necessary, assist in reducing the output of distributed generation, in order to achieve optimal tracking of the master station's commands, while strictly adhering to the local voltage safety baseline.

[0075] Step 4: When the intelligent agent detects that the power grid is in a non-emergency state and the deviation between the active power regulation command issued by the master station and the power of the plant tie line is less than the preset tolerance threshold, the SOC of the energy storage node is smoothly returned to the preset SOC threshold.

[0076] The power grid being in a non-emergency state means that the voltage amplitude of all monitored target nodes within the plant is within the set safe operating threshold range, and no emergency alarm signals such as power grid fault ride-through or frequency anomaly are received; the preset tolerance threshold, in this embodiment, is preferably 5% of the plant's rated capacity, that is, the plant does not need to call up energy storage on a large scale for high-power charging and discharging in order to track the master station's instructions. Furthermore, when the intelligent agent detects that the power grid is in a non-emergency state and the deviation between the active power regulation command issued by the master station and the power of the plant tie line is less than a preset tolerance threshold, based on the current SOC and the preset SOC threshold, and combined with the remaining time until the end of the day to calculate the target recovery power, the energy storage SOC is smoothly restored to the preset SOC threshold through trajectory tracking optimization, ensuring that the system has continuous regulation capability. Specifically, this includes the following steps: Step 4.1: Calculate the target power of regression based on the remaining time. Based on the current SOC and the preset SOC threshold, and combining the rated capacity of the energy storage and the remaining time until the end of the day, calculate the target recovery power, which is the average charging and discharging power of the energy storage node. The calculation method is as follows:

[0077] in, The current moment; The preset SOC threshold is the target SOC. The current state of charge (SOC); This refers to the rated capacity of the energy storage. This represents the average charge and discharge power of the energy storage node.

[0078] Step 4.2: Construct a voltage-constrained SOC regression model and solve for the active power regulation of the energy storage node at each time step; Solving the trajectory tracking optimization problem with voltage safety constraints: To prevent new voltage exceedances caused by SOC regression, a constrained trajectory tracking optimization problem is constructed. The goal is to make the actual energy storage output as close as possible to the voltage limit. Meanwhile, the voltage constraint is treated as an insurmountable hard condition, and the SOC regression model is specifically as follows:

[0079] in, This represents the actual active power output vector of each energy storage node at the current moment; This is the adjustment increment vector for the active power of each energy storage node; This represents the average charge and discharge power vector for each energy storage node. This represents the voltage vector of each target node at present; , These are the lower and upper limits of the safe operating voltage for each target node, respectively. This is the voltage-active power sensitivity matrix.

[0080] By solving this model, the energy storage charging and discharging strategy can be intelligently adjusted during the window period of grid voltage safety to ensure that the energy storage capacity returns to the ideal level at the end of the day and maintain the long-term sustainable regulation capability of the system.

[0081] During the callback process, if an active or reactive power adjustment command is received from the master station, or if the grid voltage exceeds the limit and enters an emergency operation state, the SOC regression control is suspended, and priority is given to responding to grid regulation and safety constraints. Once the grid returns to a non-emergency state and the deviation between the active power adjustment command issued by the master station and the power of the plant tie line is less than the preset tolerance threshold, the SOC smooth callback will continue to be executed.

[0082] This invention presents a plant-side intelligent agent coordinated control method based on the sensitivity of distributed flexible resource disturbances. Its key feature is that it breaks through the dependence of traditional control methods on grid topology and model parameters, making it suitable for the regulation of wide-area aggregated flexible resources. It uses an online sensitivity matrix of the plant-side intelligent agent to perceive local safety boundaries; it constructs a least-squares-based collaborative optimization model, using the control targets issued by the master station as the optimization tracking target and local node voltage safety as a hard constraint; and it achieves hierarchical collaborative control by setting differentiated cost weights for reactive power regulation of distributed generation resources inverters, active power regulation of energy storage systems, and active power reduction of distributed generation resources.

[0083] Example 2: To more clearly illustrate the outstanding substantive features of this invention and the significant progress it brings to the prior art, an application example of implementing this invention is described below.

[0084] This embodiment aims to verify, through simulation examples, the effectiveness of the plant-side intelligent agent coordinated control method based on distributed flexible resource disturbance sensitivity proposed in this invention in responding to the master station's control objectives, ensuring local voltage safety, and managing resource sustainability under extreme operating conditions.

[0085] This embodiment is validated based on a typical 30-node power plant model containing multiple distributed power sources and energy storage systems. The system includes three photovoltaic-energy storage hybrid nodes (connected to nodes 23, 28, and 30 respectively), representing the distributed flexible resources managed by the power plant's intelligent agent. The scenario is set as an extreme high-load day, simulating typical extreme conditions such as "extreme heat and no wind." Under these conditions, local voltage fluctuations are severe, and the master station issues high-frequency, high-power-range control commands to ensure overall grid balance, requiring peak shaving and valley filling in the power plant area.

[0086] like Figure 2 As shown, after adopting the strategy of this invention, the voltage curves (blue curves) of each node in the power plant are strictly limited to the safe operating range. Compared with not adopting this strategy (red curves), this method, while ensuring the local voltage safety baseline, achieves maximum tracking of the master station's power commands through intelligent allocation of adjustment margin, effectively avoiding the risk of exceeding limits.

[0087] like Figure 3As shown, after receiving instructions from the power plant's intelligent agent, the energy storage system charges and discharges according to the calculated optimal power, prioritizing its bidirectional regulation capability to absorb deviations. Because the optimization model sets a high penalty term for photovoltaic curtailment, this strategy effectively reduces unnecessary curtailment, demonstrating that this invention can achieve optimal economic efficiency in resource allocation order when responding to master station instructions.

[0088] like Figure 4 As shown, after the energy storage SOC fluctuates significantly due to the response to the master station's commands during the day, the plant-side intelligent agent activates the regression module during the nighttime period when the load is stable. The energy storage output is intelligently scheduled, causing its state of charge trajectory to smoothly converge to the preset target value. This ensures that the energy storage has sufficient adjustment margin before the next day or the next extreme operating condition, guaranteeing the long-term execution capability of the plant-side intelligent agent for the master station's control tasks.

[0089] like Figure 5 and Figure 6 As shown, the voltage-active / reactive sensitivity value measured by the small-signal perturbation method in this invention ( Figure 5 The red crosses in the diagram are different from the theoretical values ​​calculated based on the accurate topological model. Figure 5 The blue circles in the image are highly consistent, and the overall prediction bias ( Figure 6 The parameters are within a very small range. This strongly demonstrates that the present invention can perceive power grid characteristics with high fidelity and high accuracy in environments with unknown or drifting parameters, providing a solid foundation for the accurate execution of control strategies.

[0090] In summary, the method of this invention can effectively solve the problem of regulating distributed flexible resources in extreme scenarios, achieve high-precision tracking of the control target on the master station side and reliable protection of local voltage safety, while taking into account both economy and resource sustainability.

[0091] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0092] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0093] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0094] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A plant-side intelligent agent coordinated control method based on disturbance sensitivity, characterized in that, The method includes the following steps: Step 1: Periodically inject active and reactive power disturbances not exceeding the preset disturbance thresholds at each controllable node managed by the intelligent agent on the power plant side, and collect voltage steady-state response data of the target nodes of the entire network. Calculate the voltage-active power sensitivity matrix and voltage-reactive power sensitivity matrix for the current period based on the disturbance and voltage steady-state response data. Step 2: When the master station issues a reactive power regulation command to the plant-side intelligent agent, the plant-side intelligent agent constructs a reactive power optimization target by combining the voltage-reactive power sensitivity matrix of the current cycle, solves the optimal reactive power regulation increment, and issues it to each controllable node for execution; the predicted voltage of each target node is calculated based on the optimal reactive power regulation increment. If there is still a risk of node voltage exceeding the limit after reactive power regulation, then proceed to step 3 for active power regulation. Step 3: Using the predicted voltage of each target node as the reference voltage, construct a voltage safety hard constraint based on the voltage-active power sensitivity matrix, and combine the active power adjustment command issued by the master station to construct the active power optimization target and solve for the active power adjustment amount of each controllable node. Then, send the active power adjustment amount to each controllable node for active power adjustment. Step 4: When the intelligent agent detects that the power grid is in a non-emergency state and the deviation between the active power regulation command issued by the master station and the power of the plant tie line is less than the preset tolerance threshold, the SOC of the energy storage node is smoothly returned to the preset SOC threshold.

2. The plant-side intelligent agent coordination control method according to claim 1, characterized in that: In step 2, the reactive power regulation instructions include the grid connection point target voltage regulation instructions and the plant-wide reactive power regulation instructions; for different reactive power regulation instructions, corresponding reactive power optimization targets and constraints are constructed respectively.

3. The plant-side intelligent agent coordination control method according to claim 2, characterized in that: When the master station issues a target voltage regulation command for the grid connection point to the power plant-side intelligent agent, its reactive power optimization target is: in, This is a subset of the row vectors corresponding to the nodes that exceed the limit in the voltage-reactive power sensitivity matrix. To find the reactive power adjustment increment vector for each controllable node, we need to determine the solution. This is the column vector of the desired adjustment amount of the target node voltage; When the master station issues a reactive power adjustment command for the entire plant to the plant-side intelligent agent, its reactive power optimization objective is: in, The sum of reactive power commands issued by the master station; Ω represents the set of all controllable nodes; For the first j The reactive power adjustment increment of each controllable node.

4. The plant-side intelligent agent coordination control method according to claim 1, characterized in that: In step 3, the voltage safety hard constraint specifically refers to: in, Predict voltage vectors for each target node; This is the voltage-active power sensitivity matrix; , These are the active power regulation vectors of each energy storage node in the controllable node and the active power reduction vectors of each distributed power generation node, respectively. , These are the lower and upper limits of the safe operating voltage for each target node, respectively.

5. The plant-side intelligent agent coordination control method according to claim 4, characterized in that: In step 3, the active power optimization objective is specifically: in, This represents the active power regulation of energy storage at each energy storage node. This represents the active power reduction of each distributed power node. The current reference switching power; Active power adjustment instructions issued by the master station; The penalty weights applied to the reduction of distributed power sources.

6. The plant-side intelligent agent coordination control method according to claim 1, characterized in that: In step 4, the step of smoothly regressing the SOC of the energy storage node to a preset SOC threshold specifically includes: calculating the average charging and discharging power of the energy storage node based on the current SOC and the preset SOC threshold, combined with the rated capacity of the energy storage and the remaining time of the day; constructing a voltage-constrained SOC regression model and solving it to obtain the active power adjustment of the energy storage node at each time.

7. The plant-side intelligent agent coordination control method according to claim 6, characterized in that: The objective function of the SOC regression model is as follows: in, This represents the actual active power output vector of each energy storage node at the current moment; This is the adjustment increment vector for the active power of each energy storage node; This represents the average charge and discharge power vector for each energy storage node. The average charge and discharge power of each energy storage node is determined as follows: in, The current moment; The preset SOC threshold; The current state of charge (SOC); This refers to the rated capacity of the energy storage. This represents the average charging and discharging power of the energy storage node.

8. A power plant-side intelligent agent coordinated control system based on disturbance sensitivity according to any one of claims 1-7, comprising a sensitivity matrix construction module, a reactive power regulation module, an active power regulation module, and an energy storage node SOC regression module, characterized in that: The sensitivity matrix construction module periodically injects active and reactive power disturbances not exceeding a preset disturbance threshold at each controllable node managed by the intelligent agent on the power plant side, and collects voltage steady-state response data of target nodes across the entire network. Based on the disturbance and voltage steady-state response data, it calculates the voltage-active power sensitivity matrix and voltage-reactive power sensitivity matrix for the current period. The reactive power regulation module, when the master station sends a reactive power regulation command to the plant-side intelligent agent, the plant-side intelligent agent constructs a reactive power optimization target by combining the voltage-reactive power sensitivity matrix of the current cycle, solves the optimal reactive power regulation increment, and sends it to each controllable node for execution; based on the optimal reactive power regulation increment, it calculates the predicted voltage of each target node; if there is still a risk of node voltage exceeding the limit after reactive power regulation, it proceeds to step 3 for active power regulation. The active power regulation module uses the predicted voltage of each target node as the reference voltage, constructs a voltage safety hard constraint based on the voltage-active power sensitivity matrix, and combines the active power regulation command issued by the master station to construct the active power optimization target and solve for the active power regulation amount of each controllable node. The active power regulation amount is then sent to each controllable node for active power regulation. The SOC regression module for energy storage nodes smoothly regresses the SOC of the energy storage node to the preset SOC threshold when the intelligent agent detects that the power grid is in a non-emergency state and the deviation between the active power regulation command issued by the master station and the power of the plant tie line is less than the preset tolerance threshold.

9. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-7.

Citation Information

Patent Citations

  • Power distribution network voltage control method based on data-driven power flow and sensitivity analysis

    CN120377289A