Power adjustment method and device, computer equipment and computer readable storage medium

Abstract

The invention relates to a power adjusting method and device, computer equipment and a computer readable storage medium. The method comprises the following steps: acquiring real-time measurement data of each region; determining a local state measurement value according to the real-time measurement data, and updating a local state initial value of each region in the lower-layer model prediction controller based on the local state measurement value and a corresponding local state prediction value; updating a global state initial value of an upper-layer model prediction controller by using the updated local state initial value of each region in the current period, and correcting a global dynamic behavior prediction model based on the updated global state initial values to obtain a global state prediction value of the next period; and according to the global state prediction value, the reserve capacity cost of each region and the reference power adjustment amount, a region reference power instruction of the next period is generated in combination with a preset global constraint condition. By adopting the method, the frequency adjustment precision and the dynamic response performance in a renewable energy source scene can be improved.

Description

Technical Field

[0001] This application relates to the field of new energy power technology, and in particular to a power regulation method, device, computer equipment, computer-readable storage medium, and computer program product. Background Technology

[0002] The rapid increase in the penetration rate of new energy generation in the power system has led to the power system gradually exhibiting characteristics of low inertia and weak damping, resulting in poor frequency stability. Traditional Automatic Generation Control (AGC) mainly relies on the regulation capabilities of synchronous generators. However, in environments with a high proportion of new energy grid integration, the random volatility and low inertia characteristics of wind and solar power significantly reduce the frequency response capability of the power system and exacerbate the dynamic fluctuations of Area Control Error (ACE). At the same time, the integration of new regulation resources such as energy storage and demand response further increases the complexity of multi-timescale dynamic coupling in the system. Traditional automatic generation control strategies based on proportional-integral (PI) control face severe challenges in terms of dynamic response speed and multi-resource collaborative optimization.

[0003] Currently, research on automatic generation control for new energy power systems still has many shortcomings: for example, although the fuzzy adaptive PI control strategy can adapt to the fluctuations in new energy output by dynamically adjusting the PI parameters, its adjustment range is limited by the physical inertia of the synchronous generator unit, making it difficult to cope with rapid frequency fluctuations on the order of seconds; the method of adding virtual inertia response function to new energy power plants by combining virtual synchronous generator (VSG) technology does not consider the coordination problem of resources on multiple time scales, and the virtual inertia parameters are fixed, making it unable to adapt to a wide range of operating scenarios. Summary of the Invention

[0004] Based on this, it is necessary to provide a power regulation method, device, computer equipment, computer-readable storage medium, and computer program product that can effectively solve the synergistic problem of "global optimization accuracy - local computational efficiency - dynamic scenario adaptability" in the automatic generation control of high-proportion new energy power systems, so as to improve the frequency regulation accuracy and dynamic response performance in renewable energy scenarios.

[0005] In a first aspect, this application provides a power regulation method, comprising:

[0006] The real-time measurement data of each region at the current moment is obtained by the lower-level model prediction controller tracking the power adjustment of each region in response to the regional reference power command issued by the upper-level model prediction controller at the current moment.

[0007] The local state measurement value of each region at the current time is determined based on the real-time measurement data, and the initial local state value of each region in the lower-level model prediction controller is updated based on the local state measurement value and the local state prediction value corresponding to the current time.

[0008] The global initial state of the upper-layer model prediction controller is updated using the updated local initial state values ​​of each region in the current period, and the global dynamic behavior prediction model of the upper-layer model prediction controller is corrected based on the global initial state values ​​before and after the update.

[0009] The updated global state initial value is input into the corrected global dynamic behavior prediction model to obtain the global state prediction value for the next cycle generated by the upper-level model prediction controller.

[0010] Based on the global state prediction value, the standby capacity cost of each region, and the reference power adjustment amount of each region, the upper-level model predictive controller generates the region reference power command for the next cycle in combination with the preset global constraints.

[0011] The power parameters of each region are adjusted by executing the regional reference power command for the next cycle.

[0012] Secondly, this application also provides a power regulation device, comprising:

[0013] The data acquisition module is used to acquire real-time measurement data of each region at the current moment. The real-time measurement data is obtained by the lower-level model prediction controller tracking the power adjustment of each region in response to the regional reference power command issued by the upper-level model prediction controller at the current moment.

[0014] The state determination module is used to determine the local state measurement value of each region at the current time based on the real-time measurement data, and update the initial local state value of each region in the lower-level model prediction controller based on the local state measurement value and the local state prediction value corresponding to the current time.

[0015] The rolling time-domain feedback module is used to update the global initial value of the upper-layer model prediction controller using the updated local initial value of each region in the current period, and to correct the global dynamic behavior prediction model of the upper-layer model prediction controller based on the global initial value before the update and the updated global initial value.

[0016] The state prediction module is used to input the updated global state initial value into the corrected global dynamic behavior prediction model to obtain the global state prediction value for the next cycle generated by the upper-level model prediction controller.

[0017] The instruction generation module is used to generate the regional reference power instruction of the upper-level model predictive controller in the next cycle based on the global state prediction value, the standby capacity cost of each region, and the reference power adjustment amount of each region, combined with preset global constraints.

[0018] The power adjustment module is used to execute the regional reference power command of the next cycle to adjust the power parameters of each region.

[0019] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the power regulation method described in any of the embodiments of the first aspect.

[0020] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the power regulation method described in any of the embodiments of the first aspect.

[0021] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the power regulation method described in any of the embodiments of the first aspect.

[0022] The aforementioned power regulation method, apparatus, computer equipment, computer-readable storage medium, and computer program product utilize a hierarchical centralized-distributed hybrid model predictive controller architecture. In the lower-level model predictive controller, the execution status of regional reference power commands issued by each region to the upper-level model predictive controller is tracked, and local initial state values ​​are updated based on real-time measurement data from each region. In the upper-level model predictive controller, the global initial state value is updated every preset period based on the local initial state value of the lower-level model predictive controller, and the global dynamic behavior prediction model is corrected. The corrected global dynamic behavior prediction model and the updated initial state value are then used to predict the next period. The global state prediction value is obtained, and then the regional reference power command for the next period is obtained by combining the global state prediction value of the next period with the reserve capacity cost, reference power adjustment amount and preset constraints. The command is then issued to each region for execution. This not only reduces the dimensionality of decision variables when the upper-level model predictive controller performs global optimization and reduces the amount of data computation of the upper-level model predictive controller, but also enables dynamic adaptive adjustment by updating the initial state based on the actual measurement value tracked and detected by the lower-level model predictive controller. This improves the frequency regulation accuracy and dynamic response performance in renewable energy scenarios and has the characteristics of high global optimization accuracy, fast local calculation efficiency and strong adaptability to dynamic scenarios. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating a power regulation method in one embodiment;

[0025] Figure 2 This is a flowchart illustrating the local power regulation steps in one embodiment;

[0026] Figure 3 This is a flowchart illustrating the communication topology weight configuration steps in one embodiment;

[0027] Figure 4 This is a flowchart illustrating the steps for constructing local state measurement values ​​in one embodiment;

[0028] Figure 5 This is a flowchart illustrating the steps involved in constructing the system's dynamic equations in one embodiment.

[0029] Figure 6 This is a schematic diagram of an automatic power generation control dynamic model in one embodiment;

[0030] Figure 7 This is a schematic diagram of a two-layer model predictive control architecture in one embodiment;

[0031] Figure 8 This is a schematic diagram of a region simulation in one embodiment;

[0032] Figure 9 This is a schematic diagram of the region simulation results in one embodiment;

[0033] Figure 10 This is a schematic diagram of the simulation results of the tie line power in one embodiment;

[0034] Figure 11 This is a schematic diagram of the dynamic adjustment process in region 2 in one embodiment;

[0035] Figure 12 This is a schematic diagram of simulation results at different permeability rates in one embodiment;

[0036] Figure 13 This is a structural block diagram of the power regulation device 1300 in one embodiment;

[0037] Figure 14This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0039] It should be noted that the terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" as used in this application refers to two or more. The term "and / or" as used in this application refers to one of the solutions, or any combination of multiple solutions. The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with relevant regulations.

[0040] In one embodiment, such as Figure 1 As shown, a power regulation method is provided. This embodiment illustrates the application of this method to a server. It is understood that this method can also be applied to a terminal, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps S102 to S112. Wherein:

[0041] Step S102: Obtain real-time measurement data for each region at the current moment.

[0042] The region can be a region defined based on geographical location or power grid architecture.

[0043] For example, the server can use the lower-level model predictive controller to track the power adjustment of each region in response to the regional reference power command issued by the upper-level model predictive controller at the current moment, and obtain the real-time measurement data of each region at the current moment. The real-time measurement data may include, but is not limited to, the output power of each region, the grid operating frequency, the tie-line power, the unit output, etc.

[0044] Model Predictive Control (MPC) is an online method for solving open-loop optimal control problems in the finite-time domain, achieving dynamic optimization of the controlled system. At each sampling time, it predicts future behavior based on the current state, continuously optimizes the future control sequence, and implements the control input only at the current time. This embodiment employs a two-layer model predictive controller to perform the optimization problem in the power regulation process. The upper-layer model predictive controller is a centralized model predictive controller used to handle the global economic optimization of reserve allocation and power exchange plans for all regions at a slow time scale (i.e., with a preset period as the decision time), continuously optimizing the area control error (ACE) and reserve capacity allocation of each region to reduce the dimensionality of decision variables. The lower-layer model predictive controller is a distributed model predictive controller used to handle the local dynamic adjustment of frequency tracking and energy storage charging and discharging between regions at a fast time scale (i.e., with each moment as the decision time), dynamically coordinating the output of adjustable units and new energy power plants in each region.

[0045] Step S104: Determine the local state measurement value of each region at the current time based on the real-time measurement data, and update the initial local state value of each region in the lower-level model prediction controller based on the local state measurement value and the local state prediction value corresponding to the current time.

[0046] In the scenario of multi-regional coordinated frequency regulation in a power system, the dynamic behavior of each region (the change in local state value between the previous and current moments) is jointly determined by the interaction of its internal adjustable generating units, renewable energy power plants, and energy storage systems. Local state measurements can be used to quantify the real-time state of each region and its contribution to frequency regulation.

[0047] Local state predictions can be used to characterize the prediction results of the lower-level model predictive controller for the local state value at the current time based on the local state value at the previous time step.

[0048] The initial local state value can be used as the local state value at the current time step to participate in the calculation of the predicted local state value at the next time step.

[0049] For example, the server can use the lower-level model predictive controller to extract variable features of adjustable generating units, new energy power stations, and energy storage systems within each region from real-time measurement data, and use these variable features to construct the local state measurement value of each region at the current moment. The server then processes the local state measurement value and the predicted local state value of each region at the current moment, and uses the processed data result as the updated initial local state value to update the initial state of each region within the lower-level model predictive controller.

[0050] Step S106: Update the global initial state of the upper-layer model predictive controller using the updated local initial state values ​​of each region in the current period, and correct the global dynamic behavior prediction model of the upper-layer model predictive controller based on the global initial state values ​​before and after the update.

[0051] The initial global state value can be used as the global state value for the current period and participate in the calculation of the predicted global state value for the next period. The global state value can be used to represent the integrated result of the dynamic characteristics and coupling relationships of all regions.

[0052] The global dynamic behavior prediction model can be used to predict the global state value in the (k+1)th period based on the global state value in the kth period.

[0053] For example, the upper-level model predictive controller and the lower-level model predictive controller achieve target consistency (controlling the frequency deviation of each region within a preset fluctuation range) through rolling time-domain feedback (the lower-level model predictive controller transmits its local initial state value to the upper-level model predictive controller in each cycle) and instruction decomposition (the upper-level model predictive controller is responsible for global adjustment of all regions, and the lower-level model predictive controller is responsible for local adjustment of each region). That is, within each cycle, the upper-level model predictive controller updates the global initial state value of all regions within its own region based on the local initial state values ​​of each region stored by the lower-level model predictive controller. For instance, when a cycle update is reached, the upper-level model predictive controller can use the local initial state value stored by the lower-level model predictive controller at the current moment for computation, and use the result as the global initial state value for the current cycle. Alternatively, the upper-level model predictive controller can calculate the global initial state value for the current cycle using the average of the local initial state values ​​of the lower-level model predictive controller at various moments within the current cycle.

[0054] Optionally, in some implementations, the calculation method for the global initial state value can refer to the calculation method for the local initial state value described above. The local initial state values ​​of each region are processed to obtain the global state measurement value. The global state prediction value and the global state measurement value of the upper-level model predictive controller in the current period are processed, and the processed data result is used as the global initial state value of the upper-level model predictive controller.

[0055] The server can use the deviation between the initial global state value before and after the update to correct the global dynamic behavior prediction model of the upper-level model predictive controller. Alternatively, the server can correct the global dynamic behavior prediction model based on the relationship between the initial global state values ​​before and after the update.

[0056] Step S108: Input the updated global state initial value into the corrected global dynamic behavior prediction model to obtain the global state prediction value for the next cycle generated by the upper-level model prediction controller.

[0057] For example, the server can input the updated global state initial value into the corrected global dynamic behavior prediction model for processing, and use the processed data result as the global state prediction value for the next cycle generated by the upper-level model prediction controller.

[0058] Step S110: Based on the global state prediction value, the standby capacity cost of each region, and the reference power adjustment amount of each region, and combined with the preset global constraints, generate the regional reference power command of the upper-level model predictive controller in the next cycle.

[0059] For example, the server can utilize the upper-level model predictive controller to construct an objective function based on the global state prediction value, the reserve capacity cost of each region, and the reference power adjustment amount of each region. By solving the objective function using preset global constraints, the optimal solution that minimizes the Area Control Error (ACE) and reserve capacity cost within the prediction time domain is obtained. Based on the optimal solution, the output parameters of adjustable generating units, renewable energy plants, and energy storage systems in each region are determined, and the regional reference power command for the upper-level model predictive controller in the next cycle is generated. The global constraints may include, but are not limited to, power balance constraints, reserve capacity limit constraints, and control input limiting constraints.

[0060] Step S112: Execute the regional reference power command for the next cycle to adjust the power parameters of each region.

[0061] For example, the server can distribute the regional reference power command for the next cycle generated by the upper-level model predictive controller to each region for execution, thereby adjusting the power parameters of each region. Subsequently, the server can also use the lower-level model predictive controller to continue tracking the execution of the regional reference power command for the next cycle in each region, collect real-time measurement data of each region at the next moment, and repeat the operations of steps S102 to S112 for rolling optimization adjustment.

[0062] The aforementioned power regulation method utilizes a hierarchical centralized-distributed hybrid model predictive controller architecture. In the lower-level model predictive controller, the execution status of regional reference power commands issued by the upper-level model predictive controller is tracked in each region, and local initial state values ​​are updated based on real-time measurement data from each region. In the upper-level model predictive controller, the global initial state value is updated every preset period based on the local initial state value of the lower-level model predictive controller, and the global dynamic behavior prediction model is corrected. The corrected global dynamic behavior prediction model and the updated initial state value are used to predict the global state prediction value for the next period. Then, based on the global state prediction value for the next period, combined with reserve capacity cost, reference power adjustment amount, and preset global constraints, the regional reference power command for the next period is obtained and issued to each region for execution. This not only reduces the dimensionality of decision variables during global optimization by the upper-level model predictive controller and reduces the data computation load of the upper-level model predictive controller, but also enables dynamic adaptive adjustment by updating the initial state based on the actual measurement values ​​tracked and detected by the lower-level model predictive controller. This improves the frequency regulation accuracy and dynamic response performance in renewable energy scenarios, and features high global optimization accuracy, fast local computation efficiency, and strong adaptability to dynamic scenarios.

[0063] In one exemplary embodiment, such as Figure 2 As shown, after step S104, the system may further include steps S202 to S206. Wherein:

[0064] Step S202: Input the updated initial local state values ​​of each region into the local dynamic behavior prediction model to obtain the local state prediction value for the next time step generated by the lower-level model prediction controller.

[0065] Among them, the local dynamic behavior prediction model can be used to predict the local state value at time k+1 based on the local state value at time k.

[0066] Local state predictions can be used to characterize the local state values ​​of each region predicted by the lower-level model predictive controller at the next time step.

[0067] For example, the server can refer to the above-described method for generating global state prediction values, and in the lower-level model prediction controller, input the updated local state initial values ​​of each region into the local dynamic behavior prediction model for calculation and processing, and use the processing results as the local state prediction values ​​of each region generated by the lower-level model prediction controller at the next moment.

[0068] Step S204: Based on the power deviation parameter between the local state prediction value at the next moment and the reference power parameter corresponding to the regional reference power command, as well as the local operating cost of each region, and combined with the preset local constraints, generate the regional reference power command for each region at the next moment for the lower-level model predictive controller.

[0069] For example, the server can extract the predicted unit power value for each region at the next time step from the predicted local state value for each region at the next time step. The difference between the predicted unit power value and the reference power parameter corresponding to the regional reference power command is used as the power deviation parameter. The following operations are performed for each region: In the lower-level module predictive controller, an objective function is constructed using the power deviation parameter and the local operating layer. Based on preset local constraints, the objective function is solved to obtain the optimal solution that tracks the upper-level command and minimizes the local cost. The lower-level model predictive controller generates the power adjustment scheme for each region at the next time step based on the optimal solution for each region. The local constraints may include, but are not limited to, power balance constraints and unit ramp rate constraints.

[0070] Step S206: Execute the power adjustment scheme for the next moment to adjust the power parameters of each region.

[0071] For example, the server can use the lower-level model predictive controller to send the power regulation scheme for the next time step to each region for execution in order to adjust the power parameters of each region on a fast time scale.

[0072] In this embodiment, by employing a lower-level model predictive controller to achieve local dynamic adjustment for each region on a fast time scale, the response speed of power regulation can be improved.

[0073] In one exemplary embodiment, such as Figure 3 As shown, the power regulation method provided in this application may further include the following steps S302 to S308. Wherein:

[0074] Step S302: Construct the node feature vector corresponding to each region by using the regional control deviation parameters, state of charge parameters and new energy fluctuation intensity of each region.

[0075] For example, in a power system containing Automatic Generation Control (AGC), there are typically nodes corresponding to different regions. By encoding these complex and diverse nodes, information closely related to the current node state is selected for overall system evaluation. For instance, factors such as the grid's regional control deviation parameters, state of charge (SOC), and the intensity of renewable energy fluctuations define node characteristics. Therefore, the server can construct node feature vectors for each region in the lower-level model predictive controller using the regional control deviation parameters, SOC parameters, and renewable energy fluctuation intensity for each region.

[0076] Alternatively, in some implementations, the node feature vectors can be constructed using the following feature vector equation:

[0077] ,

[0078] In the formula: This is for regional control error; It is in a charged state; The intensity of new energy fluctuations.

[0079] Step S304: Using the node feature vectors of each region and the preset weight matrix, calculate the attention weights between regions.

[0080] For example, the server can measure the importance of inter-node communication by calculating the inter-node attention coefficients corresponding to each region, and determine the inter-node communication that is critical for frequency regulation in the power system. The server can calculate the attention weights between regions by performing calculations on the node feature vectors of each pair of regions and the preset weight matrix in the lower-level model predictive controller.

[0081] Alternatively, in some implementations, attention weights can be calculated using the following formula:

[0082] ,

[0083] In the formula: This is the weight matrix. This is the attention parameter vector.

[0084] Step S306: Normalize the attention weights to obtain normalized weights between regions.

[0085] For example, the server can perform calculations using the attention weights between regions and the maximum value of those attention weights, and use the processed result as the normalized weights between regions. This eliminates the quantization differences between different weights, ensuring that the weights of each region fall within a specific range, facilitating comparison and calculation on the same scale, and improving system stability.

[0086] Alternatively, in some implementations, the following formula can be used for normalization:

[0087] .

[0088] Step S308: Configure the communication topology weights between regions in the lower-level model prediction controller according to the normalized weights.

[0089] For example, the server can directly use the normalized weights between regions as communication topology weights to configure the lower-level model prediction controller. Alternatively, it can multiply the original communication topology weights by the normalized weights and use the product as the new communication topology weights to update the lower-level model prediction controller.

[0090] In this embodiment, by introducing a graph attention mechanism to dynamically identify the coupling strength between regions and adaptively optimize the distributed communication topology weights of the lower-level model prediction controller, the complexity of collaborative computation can be reduced and the processing efficiency of the lower-level model prediction controller can be improved.

[0091] In an exemplary embodiment, the power adjustment method provided in this application may further include: setting the communication topology weight between regions corresponding to the normalized weight to zero when the normalized weight is less than a preset sparsity threshold.

[0092] In this embodiment, by setting the communication topology weight between regions with normalized weights less than a preset sparsity threshold to zero, weak connections can be eliminated, reducing the amount of data computation for the lower-level model predictive controller.

[0093] In one exemplary embodiment, such as Figure 4 As shown, step S104 may include steps S402 to S406. Wherein:

[0094] Step S402: Extract new energy output data, power exchange parameters, load disturbance parameters, unit output parameters and state of charge parameters from real-time measurement data.

[0095] For example, the server can extract the new energy output data, power exchange parameters, load disturbance parameters, unit output parameters and state of charge parameters of each region from the real-time measurement data of each region at the current moment, based on the characteristics of new energy output fluctuation and load disturbance.

[0096] Step S404: Input the new energy output data, power exchange parameters and load disturbance parameters of each region into the preset system dynamic equation to obtain the frequency deviation parameters of each region at the current time.

[0097] The system dynamic equations can be constructed using a multi-resource joint state-space model with the frequency deviation parameter as the control objective.

[0098] For example, the server can construct a multi-resource joint state-space model for the characteristics of new energy output fluctuations and load disturbances, and set the frequency deviation parameter as the control objective, thereby forming the system dynamic equations. The new energy output data, power exchange parameters, and load disturbance parameters of each region are input into the preset system dynamic equations for calculation and processing, thereby obtaining the frequency deviation parameters of each region at the current time.

[0099] Step S406: Using the frequency deviation parameters, unit output parameters and state of charge parameters of each region, construct the local state measurement values ​​of each region.

[0100] For example, the server can construct local state measurement values ​​in vector form in the lower-level model predictive controller using the frequency deviation parameters, unit output parameters and state of charge parameters of each region, that is, obtain the local state vector.

[0101] In this embodiment, by using the system dynamic equation to solve for the frequency deviation parameter, the unit output parameter and the state of charge parameter are extracted from the real-time measurement data. The local state measurement value is constructed using the frequency deviation parameter, the unit output parameter, and the state of charge parameter. This can improve the quantification of the relationship between the output of new energy and the frequency deviation, and improve the accuracy and reliability of the local state measurement value.

[0102] In one exemplary embodiment, such as Figure 5 As shown, a method for constructing the dynamic equations of a system is provided, including the following steps S502 to S510. Wherein:

[0103] Step S502: Based on the additional power compensation amount generated by the wind turbine under rotor kinetic energy control and the power change amount generated by the wind turbine under pitch control, construct the wind power transfer function related to the frequency deviation of the wind turbine.

[0104] For example, the server can use inertial simulation to simulate the output power adjustment of a wind turbine under varying wind speeds, thereby enhancing the grid's inertial response support. The rotor kinetic energy characteristics of the wind turbine are modeled as an equivalent first-order inertial element, whose dynamic characteristics are consistent with those of traditional turbines. Therefore, the additional power compensation generated by the wind turbine under rotor kinetic energy control can be used to construct the following first wind transfer function related to frequency deviation:

[0105] ,

[0106] In the formula: This indicates the additional power compensation amount generated by rotor kinetic energy control. This indicates frequency deviation. This represents the system's dynamic adjustment gain coefficient. s represents the complex frequency variable in the Laplace transform. The time constant representing mechanical inertia.

[0107] Pitch control of wind turbines dynamically adjusts the blade angle of attack to change the aerodynamic capture efficiency of the rotor, achieving precise power regulation of the unit. However, although this control method has wide-range power regulation capability, its dynamic response exhibits typical inertial characteristics due to the inherent hysteresis of the mechanical transmission coefficient. Therefore, a second wind transfer function related to the frequency deviation of the wind turbine can be constructed using the power change generated by the wind turbine under pitch control, as shown below:

[0108] ,

[0109] In the formula: Provides power variation for pitch control. This is the primary frequency modulation coefficient. is the variable pitch response time constant. And s are the same as in the above formula.

[0110] Step S504: Using the output ratio of the photovoltaic power station, the command issuance time, and the inverter execution time, construct the photovoltaic transfer function related to the frequency deviation of the photovoltaic power station.

[0111] For example, the frequency response of a photovoltaic (PV) power plant needs to consider dynamic time delay characteristics. Therefore, the output ratio of the PV power plant, the command issuance time, and the inverter execution time can be used to characterize the frequency modulation command transmission delay and the dynamic response process of power electronic devices through a dual-inertial system, constructing the PV transfer function related to frequency deviation of the PV power plant as shown below:

[0112] ,

[0113] In the formula: The proportion of power output from photovoltaics; The time when the instruction was issued; This represents the inverter's execution time. The value 's' is the same as in the formula above.

[0114] Step S506: Construct a dynamic power response function of distributed energy storage related to frequency deviation using the droop coefficient and differential gain coefficient; construct a state-of-charge response function using the dynamic power response function and the rated capacity of distributed energy storage; and obtain the energy storage power function by combining the dynamic power response function and the state-of-charge response function.

[0115] For example, when facing power demands from various dynamic changes in the external power grid, distributed energy storage units can rapidly switch from one power level to another, accurately outputting the corresponding power to participate in grid frequency regulation, stabilizing the grid frequency within a certain range to cope with sudden power fluctuations or peak loads. The dynamic power response of energy storage can dynamically adjust charging and discharging power proportionally and differentially according to the grid frequency deviation. Therefore, the following distributed energy storage dynamic power response function related to frequency deviation can be constructed using the droop coefficient and differential gain coefficient:

[0116] ,

[0117] In the formula: The droop factor (MW / Hz) determines the static frequency modulation capability; The differential gain (MW·s / Hz) suppresses the rate of change of frequency. Same as the formula above.

[0118] The state-of-charge (POC) parameter represents the ratio of the remaining charge of a distributed energy storage unit to its rated battery capacity at a given moment. By considering the POC parameter of the distributed energy storage unit at the previous moment, along with its capacity and power, the POC parameter can be predicted and dynamically adjusted for the next moment to maintain the POC balance of the distributed energy storage units and preserve power system frequency stability. Therefore, the POC response function can be constructed using the energy storage dynamic power response function and the rated capacity of the distributed energy storage, as shown below:

[0119] ,

[0120] In the formula: SOC represents the state of charge parameter. η represents the charge / discharge efficiency coefficient. Rated energy storage capacity (MWh); during discharge >0, the State of Charge (SOC) parameter decreases during charging. <0, the state of charge (SOC) parameter increases.

[0121] The server can combine the above-mentioned energy storage dynamic power response function and state of charge response function to obtain the energy storage power function.

[0122] Step S508: Fit the deterministic component of the load disturbance using the daily load curve, construct the random component of the load disturbance using a Gaussian distribution, and combine the deterministic and random components to obtain the load disturbance function.

[0123] For example, power system load power exhibits multi-timescale random disturbance characteristics. Its fluctuations are influenced by user behavior patterns, climate sensitivity, and the time-varying coupling effects of the grid topology, requiring quantitative characterization using a probability distribution model. For instance, the deterministic component of the load disturbance can be fitted using the daily load curve. The random component of the load disturbance can be constructed using a Gaussian distribution with a mean of 0 and a variance of σ². Combining the deterministic and random components, the following load disturbance function can be obtained:

[0124] ,

[0125] in, The deterministic component and the random component are obtained by fitting the daily load curve. It follows a mean of 0 and a variance of σ. 2 Gaussian process.

[0126] The load regulation capability of a power system, characterized by its frequency-power static characteristic coefficient, can be quantitatively modeled using the equivalent damping coefficient, resulting in the following expression for load frequency regulation:

[0127] ,

[0128] In the formula: This is the load frequency regulation coefficient, typically ranging from 1% to 3%.

[0129] Step S510: Using the wind power transfer function, photovoltaic transfer function, energy storage power function, and load disturbance function, and with the frequency deviation parameter as the control objective, construct the system dynamic equation.

[0130] For example, the server can calculate wind power generation using the wind transfer function, photovoltaic power generation using the photovoltaic transfer function, energy storage charging and discharging power using the energy storage power function, and load disturbance using the load disturbance function. It can also calculate the load frequency regulation using the load frequency regulation formula. Furthermore, it can calculate the net power imbalance using the wind power generation, photovoltaic power generation, energy storage charging and discharging power, and grid interaction power. Finally, it can calculate the frequency deviation using the net power imbalance, load disturbance, load frequency regulation, and the system's equivalent inertial time constant. This leads to the system dynamic equations shown below:

[0131] ,

[0132] In the formula: Similar to the formula above, this represents the frequency deviation. The system's equivalent inertial time constant. This represents the net power imbalance. △Pload represents the load disturbance. This indicates the load frequency regulation. Pwind represents wind power generation. P PV Indicates photovoltaic power generation capacity, P ess This indicates the energy storage charging and discharging power. P grid This indicates the power exchange between power grids.

[0133] Alternatively, in some implementations, the sensitivity of wind-solar-storage power output to frequency can also be quantified using a Jacobian matrix:

[0134] .

[0135] Alternatively, in other implementations, a formula can be constructed based on all of the above formulas, such as... Figure 6 The figure shows an automatic power generation control dynamic model that includes multiple types of resources such as wind, solar and energy storage.

[0136] In this embodiment, by using the above-mentioned method to target the characteristics of power output fluctuation and load disturbance of new energy sources, an automatic power generation control dynamic model containing multiple types of resources such as wind, solar and energy storage is constructed. This model can accurately characterize the characteristics of wind power inertial simulation and pitch control, photovoltaic dual inertial time delay, energy storage droop control and SOC balancing and load probability distribution, making the model more consistent with the actual operation scenario of a high proportion of new energy power grid.

[0137] In one exemplary embodiment, such as Figure 7 As shown, a two-layer model predictive control architecture consisting of an upper-layer MPC (upper-layer model predictive controller) and a lower-layer MPC (lower-layer model predictive controller) is provided. In the upper-layer MPC, power data at the setpoint (i.e., each region) and the last optimization result are subjected to long-term rolling optimization, and the optimization result is input into the long-term predictive model to obtain the long-term optimization result. In the lower-layer MPC, the long-term optimization result of the upper-layer MPC is tracked to predict in the short-term predictive model, generating the power regulation mode for the controlled object. Short-term feedback correction is performed using real-time measurement data of the controlled object and the short-term optimization result of the short-term predictive model, achieving synchronization between the lower-layer and upper-layer MPCs through time-domain rolling.

[0138] Optionally, in some implementations, in multi-regional coordinated frequency regulation scenarios of the power system, the upper-level centralized MPC is responsible for optimizing the overall economy and stability. It needs to integrate the frequency deviation, power exchange, and reserve capacity information of each region to generate regional reference power commands. Its core objective is to suppress the impact of new energy output fluctuations and load disturbances on the grid frequency by continuously optimizing regional control deviations and reserve capacity allocation, while minimizing frequency regulation costs.

[0139] The global state vector integrates the dynamic characteristics and coupling relationships of all regions, and its specific expression is as follows:

[0140] ;

[0141] The control input is the reference power adjustment amount for each zone:

[0142] ;

[0143] The global dynamic behavior is described by the following discrete state-space equations, that is, the following formula is the global dynamic behavior prediction model:

[0144] ;

[0145] in: This is the state transition matrix, representing the regional inertia and frequency coupling; To control the input matrix; The vector of new energy output and load disturbance; This is the perturbation input matrix.

[0146] The upper-level MPC minimizes the ACE bias and reserve capacity cost in the prediction time domain, and its objective function equation is:

[0147] .

[0148] Constraints include the following types:

[0149] Power balance: ;

[0150] Spare capacity limit: ;

[0151] Control input limit: .

[0152] The initial state value is updated and the global dynamic behavior prediction model is corrected every cycle:

[0153] ,

[0154] In the formula: This is the Kalman filter gain matrix. These are actual measured values.

[0155] Alternatively, in other implementations, the lower-level MPC can define the local state vector in the following manner:

[0156] .

[0157] Considering a multi-regional collaborative scenario in a power system, the dynamic behavior of region i is described by dynamic equations:

[0158] ,

[0159] Where: control input , This is a local disturbance.

[0160] The optimization objective for region i is to track upper-layer instructions and minimize local cost, and its specific expression is: .

[0161] The constraint equations include the following types of constraints:

[0162] Power balance: ;

[0163] Unit ramp rate: .

[0164] Update the initial state based on the local state observer:

[0165] .

[0166] Alternatively, in some implementations, such as Figure 8 As shown, a regional simulation diagram in the IEEE 33-node model is provided, including Region 1, Region 2, and Region 3. A photovoltaic power station and a wind farm are configured in Region 1 and Region 2 respectively, and Region 3 is connected to an energy storage power station, achieving an overall renewable energy penetration rate of 30%. Power exchange between regions is achieved through tie lines, with specific transmission power as follows: Region 2 transmits 132MW to Region 1 and 341MW to Region 3, and Region 3 transmits 282MW to Region 1. Regarding frequency regulation system configuration, Regions 1, 2, and 3 have frequency regulation capacities of 556MW, 662MW, and 570MW respectively, corresponding to regional loads of 2770MW, 2162MW, and 3028MW. Frequency regulation units within each region exchange information through a ring communication network.

[0167] Refer to the simulation parameters in Table 1 and the model parameters in Table 2 below. Figure 8 Simulation experiments were conducted in the area shown:

[0168] Table 1 Simulation Parameters

[0169]

[0170] Table 2 Model Parameters

[0171]

[0172] (1) Load step change of 0.2 pu in region 2

[0173] In the simulation experiment, a load surge event occurred in Region 2 at t=10s, with an increment of 432MW (per unit value 0.2). Since this disturbance of 432MW did not exceed the preset frequency regulation capacity of 662MW in Region 2, power balance adjustment could be completed through its internal frequency regulation resources without the need for cross-regional power support from Regions 1 and 3. Under this condition, the power transmission of the interconnection lines between regions strictly maintained the initial planned value.

[0174] like Figure 9 The dynamic frequency response curves shown demonstrate that, compared to traditional centralized PI control strategies, the Automatic Generation Control (AGC) method provided in this application exhibits the following technical advantages during transient processes: By constructing a dynamic frequency regulation model for the generator set, the coordinated response mechanism between new energy units and traditional thermal power is optimized, raising the minimum frequency point by approximately 0.15 Hz; during the frequency recovery phase, a multi-timescale coordinated control strategy is employed, effectively suppressing approximately 23% overshoot. This verifies the effectiveness of the proposed control strategy in enhancing the absorption capacity of new energy sources and improving the dynamic quality of the system.

[0175] Figure 10 The dynamic response characteristics of the system tie-line power are demonstrated. Under initial steady-state conditions, region 2 transmits power to regions 1 and 3 via tie lines 12 and 23, respectively. When a 432MW load surge occurs at t=10s, to maintain power balance within region 2, the transmitted power of tie lines 12 and 23 instantaneously decreases by approximately 18.7% and 9.4%, respectively, and then gradually recovers to the initial setpoint through dynamic adjustment. Notably, the control strategy proposed in this application exhibits superior dynamic performance compared to traditional methods during the power recovery phase: power overshoot is reduced by 37%, and the settling time is shortened by 2.8 seconds, verifying its power smoothing control capability.

[0176] Given that Region 2 has the capacity margin to autonomously complete power regulation (disturbance value 432MW < frequency regulation capacity 662MW), this application focuses on analyzing the coordinated frequency regulation mechanism of its multiple types of units. Figure 11The dynamic adjustment process of AGC commands for synchronous generator units and wind farms within Region 2 was revealed. Data shows that the initial response difference among the units reached a maximum of 15.6MW, but through the coordination of the distributed consensus algorithm, the command deviation converged to within ±0.5MW within 23 seconds. This confirms the effectiveness of the constructed control architecture in the group control of heterogeneous units, achieving optimized allocation of active power between thermal power and renewable energy units.

[0177] (2) Penetration rates of different new energy sources

[0178] To assess the adaptability of the proposed control strategy under different renewable energy proportions, this application constructs comparative cases with renewable energy penetration rates of 45% and 60% by replacing thermal power units with equal capacity. Figure 12 The system frequency dynamic response curves under different penetration rates are presented. Simulation results show that as the renewable energy penetration rate increases from 30% to 60%, the maximum frequency deviation in each region exhibits a systematic decreasing trend. Notably, although a high proportion of renewable energy integration reduces the system's equivalent inertia, the optimized scheduling of the rapid frequency regulation capability of renewable energy units through the strategy proposed in this application shortens the system frequency regulation response time by approximately 38%, effectively offsetting the negative impact of reduced inertia.

[0179] Analysis shows that compared to the average response delay of 5-8 seconds for thermal power units, renewable energy units can complete more than 80% of power regulation within 0.5 seconds. The control algorithm in this application dynamically adjusts the frequency regulation weighting coefficient, enabling renewable energy units to handle 72% of the initial frequency regulation demand in a 60% penetration scenario. This collaborative mechanism based on multi-energy response characteristics reduces the maximum frequency deviation by 0.12 Hz compared to the 30% scenario at 60% penetration, and shortens the time to recover the frequency to steady state by 4.3 seconds. This phenomenon verifies the unique advantage of the proposed strategy in high-proportion renewable energy systems—achieving simultaneous improvement in system frequency regulation performance and renewable energy absorption capacity by deeply exploring the dynamic response potential of renewable energy.

[0180] In the embodiments described above, an Automatic Generation Control (AGC) dynamic model incorporating multiple resource types (wind, solar, and energy storage) is constructed to address the characteristics of new energy output fluctuations and load disturbances. This model accurately characterizes features such as wind power inertial simulation and pitch control, photovoltaic dual-inertial time delay, energy storage droop control and state of charge balancing, and load probability distribution, making the model more closely aligned with the actual operation scenarios of high-proportion new energy power grids. A "centralized-distributed" hybrid control architecture is designed. The upper-layer centralized MPC continuously optimizes regional control deviations and reserve capacity allocation, focusing on global economic optimization. The lower-layer distributed MPC dynamically coordinates adjustable units and new energy power plants within the region, achieving fast-timescale local adjustments. Through rolling time-domain feedback and command decomposition, the global optimization accuracy and local computational efficiency are balanced, improving system response speed. A graph attention mechanism is introduced to dynamically identify the coupling strength between regions, adaptively optimize the communication topology weights of the lower-layer distributed MPC, eliminate weak connections, reduce collaborative computation complexity, reduce redundant communication and latency, and enhance the ability of multi-regional collaborative frequency regulation to process spatiotemporal correlation information. This strategy significantly improves the accuracy of system frequency regulation and dynamic response performance, effectively enhances the frequency stability of high-proportion new energy power systems, and solves the shortcomings of traditional methods in terms of dynamic response, multi-resource coordination and computational efficiency.

[0181] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0182] Based on the same inventive concept, this application also provides a power regulation device for implementing the power regulation method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more power regulation device embodiments provided below can be found in the limitations of the power regulation method described above, and will not be repeated here.

[0183] In one exemplary embodiment, such as Figure 13As shown, a power regulation device 1300 is provided, including: a data acquisition module 1302, a state determination module 1304, a rolling time domain feedback module 1306, a state prediction module 1308, a command generation module 1310, and a power regulation module 1312, wherein:

[0184] The data acquisition module 1302 is used to acquire real-time measurement data of each region at the current moment. The real-time measurement data is obtained by the lower-level model predictive controller tracking the power adjustment of each region in response to the regional reference power command issued by the upper-level model predictive controller at the current moment.

[0185] The state determination module 1304 is used to determine the local state measurement value of each region at the current time based on real-time measurement data, and update the initial local state value of each region in the lower-level model prediction controller based on the local state measurement value and the local state prediction value corresponding to the current time.

[0186] The rolling time-domain feedback module 1306 is used to update the global initial state value of the upper-level model predictive controller by using the updated local initial state value of each region in the current period, and to correct the global dynamic behavior prediction model of the upper-level model predictive controller based on the global initial state value before the update and the updated global initial state value.

[0187] The state prediction module 1308 is used to input the updated global state initial value into the corrected global dynamic behavior prediction model to obtain the global state prediction value for the next cycle generated by the upper-level model prediction controller.

[0188] The instruction generation module 1310 is used to generate the regional reference power instruction of the upper-level model predictive controller in the next cycle based on the global state prediction value, the standby capacity cost of each region and the reference power adjustment amount of each region, combined with the preset global constraints.

[0189] The power adjustment module 1312 is used to execute the regional reference power command for the next cycle to adjust the power parameters of each region.

[0190] In an exemplary embodiment, the power regulation device 1300 further includes a local optimization module, which is used to input the updated local state initial values ​​of each region into the local dynamic behavior prediction model to obtain the local state prediction value for the next time moment generated by the lower-level model prediction controller; based on the power deviation parameter between the local state prediction value for the next time moment and the reference power parameter corresponding to the regional reference power command, and the local operating cost of each region, combined with preset local constraints, the lower-level model prediction controller generates the regional reference power command for each region for the next time moment; and executes the power regulation scheme for the next time moment to adjust the power parameters of each region.

[0191] In an exemplary embodiment, the local optimization module is further configured to construct node feature vectors corresponding to each region using the regional control deviation parameters, state of charge parameters, and new energy fluctuation intensity of each region; calculate the attention weights between regions using the node feature vectors of each region and a preset weight matrix; normalize the attention weights to obtain normalized weights between regions; and configure the communication topology weights between regions in the lower-level model predictive controller according to the normalized weights.

[0192] In an exemplary embodiment, the local optimization module is further configured to set the communication topology weight between regions corresponding to the normalized weight to zero when the normalized weight is less than a preset sparsity threshold.

[0193] In an exemplary embodiment, the state determination module 1304 is further configured to extract new energy output data, power exchange parameters, load disturbance parameters, unit output parameters, and state of charge parameters from real-time measurement data; input the new energy output data, power exchange parameters, and load disturbance parameters of each region into a preset system dynamic equation to obtain the frequency deviation parameters of each region at the current moment. The system dynamic equation is constructed using a multi-resource joint state-space model with the frequency deviation parameters as the control objective; and construct the local state measurement values ​​of each region using the frequency deviation parameters, unit output parameters, and state of charge parameters of each region.

[0194] In an exemplary embodiment, the state determination module 1304 is further configured to: construct a wind power transfer function related to frequency deviation of the wind turbine generator based on the additional power compensation generated by the wind turbine generator under rotor kinetic energy control and the power change generated by the wind turbine generator under pitch control; construct a photovoltaic transfer function related to frequency deviation of the photovoltaic power station using the output ratio of the photovoltaic power station, the command issuance time, and the execution time of the inverter; construct a dynamic power response function of distributed energy storage related to frequency deviation using the droop coefficient and the differential gain coefficient; construct a state-of-charge response function using the dynamic power response function of energy storage and the rated capacity of distributed energy storage; obtain an energy storage power function by combining the dynamic power response function of energy storage and the state-of-charge response function; fit the deterministic component of the load disturbance using the daily load curve; construct the random component of the load disturbance using the Gaussian distribution; obtain a load disturbance function by combining the deterministic component and the random component; and construct a system dynamic equation using the wind power transfer function, the photovoltaic transfer function, the energy storage power function, and the load disturbance function, with the frequency deviation parameter as the control objective.

[0195] Each module in the aforementioned power regulation device 1300 can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0196] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 14 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores real-time measurement data, local initial state values, global initial state values, and global state prediction values. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a power regulation method.

[0197] Those skilled in the art will understand that Figure 14 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0198] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0199] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method embodiments.

[0200] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0201] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0202] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0203] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A power regulation method, characterized in that, The method includes: The real-time measurement data of each region at the current moment is obtained by the lower-level model prediction controller tracking the power adjustment of each region in response to the regional reference power command issued by the upper-level model prediction controller at the current moment. The local state measurement value of each region at the current time is determined based on the real-time measurement data, and the initial local state value of each region in the lower-level model prediction controller is updated based on the local state measurement value and the local state prediction value corresponding to the current time. The global initial state of the upper-layer model prediction controller is updated using the updated local initial state values ​​of each region in the current period, and the global dynamic behavior prediction model of the upper-layer model prediction controller is corrected based on the global initial state values ​​before and after the update. The updated global state initial value is input into the corrected global dynamic behavior prediction model to obtain the global state prediction value for the next cycle generated by the upper-level model prediction controller. Based on the global state prediction value, the standby capacity cost of each region, and the reference power adjustment amount of each region, the upper-level model predictive controller generates the region reference power command for the next cycle in combination with the preset global constraints. The power parameters of each region are adjusted by executing the regional reference power command for the next cycle.

2. The method according to claim 1, characterized in that, After determining the local state measurement value of each region at the current time based on the real-time measurement data, and updating the initial local state value of each region in the lower-level model prediction controller based on the local state measurement value and the local state prediction value corresponding to the current time, the method further includes: The updated initial local state values ​​of each region are input into the local dynamic behavior prediction model to obtain the local state prediction value for the next time moment generated by the lower-level model prediction controller. Based on the power deviation parameter between the local state prediction value at the next moment and the reference power parameter corresponding to the regional reference power command, as well as the local operating cost of each region, and combined with the preset local constraints, the lower-level model prediction controller generates the regional reference power command for each region at the next moment. The power adjustment scheme for the next time step is executed to adjust the power parameters of each region.

3. The method according to claim 2, characterized in that, The method further includes: Each region is constructed using its regional control deviation parameters, state of charge parameters, and new energy fluctuation intensity. The attention weights between regions are calculated using the node feature vectors of each region and a preset weight matrix. The attention weights are normalized to obtain normalized weights between regions; Configure the communication topology weights between regions in the lower-level model prediction controller according to the normalized weights.

4. The method according to claim 3, characterized in that, The method further includes: If the normalized weight is less than a preset sparsity threshold, the communication topology weight between the regions corresponding to the normalized weight is set to zero.

5. The method according to claim 1, characterized in that, The step of determining the local state measurement value of each region at the current moment based on the real-time measurement data includes: The new energy output data, power exchange parameters, load disturbance parameters, unit output parameters and state of charge parameters are extracted from the real-time measurement data. The new energy output data, power exchange parameters, and load disturbance parameters of each region are respectively input into the preset system dynamic equation to obtain the frequency deviation parameters of each region at the current time. The system dynamic equation is constructed using a multi-resource joint state-space model with the frequency deviation parameters as the control target. Local state measurement values ​​for each region are constructed using frequency deviation parameters, unit output parameters, and state of charge parameters for each region.

6. The method according to claim 5, characterized in that, The methods for constructing the system's dynamic equations include: Based on the additional power compensation generated by the wind turbine under rotor kinetic energy control and the power change generated by the wind turbine under pitch control, a wind transfer function related to the frequency deviation of the wind turbine is constructed. By utilizing the output ratio of the photovoltaic power station, the command issuance time, and the inverter execution time, a photovoltaic transfer function related to the frequency deviation of the photovoltaic power station is constructed. A dynamic power response function for distributed energy storage related to frequency deviation is constructed using the droop coefficient and the differential gain coefficient. A state-of-charge response function is constructed using the dynamic power response function and the rated capacity of the distributed energy storage. The energy storage power function is obtained by combining the dynamic power response function and the state-of-charge response function. The deterministic component of the load disturbance is fitted using the daily load curve, and the random component of the load disturbance is constructed using a Gaussian distribution. The load disturbance function is obtained by combining the deterministic component and the random component. Using the wind power transfer function, the photovoltaic transfer function, the energy storage power function, and the load disturbance function, and with the frequency deviation parameter as the control target, the system dynamic equation is constructed.

7. A power regulation device, characterized in that, The device includes: The data acquisition module is used to acquire real-time measurement data of each region at the current moment. The real-time measurement data is obtained by the lower-level model prediction controller tracking the power adjustment of each region in response to the regional reference power command issued by the upper-level model prediction controller at the current moment. The state determination module is used to determine the local state measurement value of each region at the current time based on the real-time measurement data, and update the initial local state value of each region in the lower-level model prediction controller based on the local state measurement value and the local state prediction value corresponding to the current time. The rolling time-domain feedback module is used to update the global initial value of the upper-layer model prediction controller using the updated local initial value of each region in the current period, and to correct the global dynamic behavior prediction model of the upper-layer model prediction controller based on the global initial value before the update and the updated global initial value. The state prediction module is used to input the updated global state initial value into the corrected global dynamic behavior prediction model to obtain the global state prediction value for the next cycle generated by the upper-level model prediction controller. The instruction generation module is used to generate the regional reference power instruction of the upper-level model predictive controller in the next cycle based on the global state prediction value, the standby capacity cost of each region, and the reference power adjustment amount of each region, combined with preset global constraints. The power adjustment module is used to execute the regional reference power command of the next cycle to adjust the power parameters of each region.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images