An optimal power distribution method for large-scale photovoltaic power station responding to secondary frequency regulation of power grid

Abstract

The application discloses an optimal power distribution method of a large-scale photovoltaic power station responding to secondary frequency modulation regulation of a power grid, and belongs to the technical field of power grid dispatching. The method comprises the following steps: using an FCM algorithm to cluster inverters in the large-scale photovoltaic power station; decomposing the large-scale photovoltaic power station problem into top-level control and bottom-level control according to the clustering result; summarizing the regulation capacity of all inverters in a clustering group, and distributing power in the top-level control according to the regulation capacity of the whole group obtained by summarizing; using an FATA algorithm to optimize in the top-level control; transmitting the optimal solution obtained in the top-level control to the bottom-level control, distributing AGC signals to each inverter in the group in proportion; outputting optimal regulation power input of each inverter, monitoring the response of the inverter in real time, and adjusting and optimizing in the next control period according to feedback data. The application can reduce the optimization complexity of the bottom-level control, improve the calculation efficiency of AGC scheduling of the large-scale power station, provide greater reactive power reserve while minimizing power deviation, and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic generation control (AGC) and hierarchical power control of photovoltaic power stations, and in particular to an optimal power allocation method for large-scale photovoltaic power stations in response to secondary frequency regulation of a power grid. Background Art

[0002] With the global emphasis on sustainable energy and environmental protection, countries are vigorously developing renewable energy. Solar photovoltaic (PV) power generation, owing to its pollution-free, emission-free, and renewable nature, has gained widespread adoption. According to statistics, global installed PV capacity exceeded 1,000 GW by 2023, with an average annual growth rate exceeding 40% over the past decade. However, the randomness and volatility of PV generation pose significant challenges to real-time power balancing in power grids.

[0003] To maintain grid stability and reliability, an increasing number of photovoltaic (PV) plants are incorporating Automatic Generation Control (AGC) to dynamically adjust output power in response to grid demand. However, AGC in large-scale PV plants faces several major challenges. First, PV plants must not only rapidly respond to AGC signals to maintain grid frequency and voltage stability but also consider other operational objectives, such as maintaining and optimizing reactive power regulation capabilities. This multi-objective optimization requirement complicates AGC scheduling. Second, traditional AGC scheduling methods primarily focus on optimizing the scheduling of a single control area or a small number of generating units. However, in large-scale PV plants, hundreds of inverters must be regulated simultaneously. The response characteristics and capabilities of different inverters vary significantly, making efficient coordination of these inverters for optimal overall performance a challenge. Furthermore, AGC scheduling must be completed within extremely short timeframes (e.g., within seconds) to ensure rapid response to grid changes. Traditional centralized control methods may not be able to meet such stringent real-time requirements, especially when dealing with a large number of inverters.

[0004] Some research has attempted to improve the AGC performance of PV power plants through control algorithms. For example, some model-predictive control-based technologies have been used to improve the regulation flexibility and response speed of PV inverters. Some intelligent control methods (such as neural fuzzy control and deep reinforcement learning) attempt to optimize the operating performance of PV power plants through more intelligent algorithms. However, these methods generally focus on optimizing small-scale or single inverters, and lack consideration of the overall optimization of distributed inverters within large-scale PV power plants. Summary of the Invention

[0005] The present invention proposes an optimal power allocation method for large-scale photovoltaic power plants in response to secondary frequency regulation of the power grid, which decomposes the overall power control problem of large-scale photovoltaic power plants into two levels: top-level control and bottom-level control. The top-level control is responsible for allocating AGC signals among multiple inverter groups. This level uses the Fuzzy C-Means (FCM) clustering algorithm to group inverters, thereby reducing the dimension and complexity of optimization, and optimizes the AGC response performance and reactive power reserve through the Fata morgana algorithm (FATA). Once the top-level control determines the power allocation of each cluster group, the bottom-level control is responsible for the specific power allocation within each photovoltaic power plant group. The bottom-level control, in accordance with the instructions of the top-level control, further distributes the AGC signal allocated to each group to each inverter within the group.

[0006] The purpose of the present invention is to realize an optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency regulation of the power grid through the following technical solution, including the following steps:

[0007] 1) Use the FCM algorithm to cluster inverters in all large-scale photovoltaic power plants and divide inverters with similar regulation characteristics into multiple groups;

[0008] 2) Based on the clustering results of the inverters, the large-scale photovoltaic power station problem is decomposed into top-level control and bottom-level control. The top-level control is responsible for power distribution between multiple cluster groups, while the bottom-level control is responsible for power distribution within each group. Top-level and bottom-level control models are established.

[0009] 3) Obtaining the AGC signal of the current photovoltaic power station, which is provided by the power grid dispatching center and is used to indicate the required power adjustment amount;

[0010] 4) The photovoltaic power station calculates the regulation capacity of each inverter based on the current operating status and regulation capacity of the inverter;

[0011] 5) Summarize the regulation capabilities of all inverters in each cluster group, obtain the regulation capability of the entire group by summing the regulation capabilities of each inverter, and allocate power in the top-level control based on the summarized regulation capability of the entire group to ensure that the power allocation of each group is reasonable and can meet the overall regulation requirements of the system;

[0012] 6) Use the FATA algorithm to optimize in the top-level control and initialize the parameters of the FATA algorithm;

[0013] 7) The optimal solution obtained in the top-level control is transmitted to the bottom-level control, and the AGC signal is proportionally distributed to each inverter in the group;

[0014] 8) Output the optimal regulated power input for each inverter, monitor the inverter response in real time, and make adjustments and optimizations in the next control cycle based on the feedback data.

[0015] Furthermore, in step 1), the FCM algorithm collects characteristic data of all inverters in the photovoltaic power station:

[0016]

[0017] Among them, U in formula (1) ij Indicates the membership degree of the i-th inverter to the j-th cluster; ||x i -v j || represents the Euclidean distance between the i-th inverter and the j-th cluster center; m is the fuzzification coefficient, which is used to control the fuzziness of the membership degree. Usually m>1. When m increases, the fuzziness of the membership degree also increases. Formula (2) is used to calculate the distance between the inverter and the cluster center v j The Euclidean distance is used to measure the similarity between the inverter and the cluster center. The smaller the distance, the more similar they are. In formula (2), D is the dimension of the feature vector; c represents the number of clusters; x id represents the dth eigenvalue of the i-th inverter; v jd represents the dth eigenvalue of the jth inverter; the membership matrix of the inverter is calculated by formula (1) and formula (2), thereby determining the cluster group to which each inverter belongs.

[0018] Furthermore, in step 2), the task of the top-level control is to optimize the power distribution to meet the dual-objective optimization model, i.e., minimizing the power deviation and maximizing the reactive power reserve. The optimization problem of the top-level control is expressed as:

[0019]

[0020] f=w1·f1+w2·(1-f2)(5) (5)

[0021] In formula (3), f1 represents the sum of power deviations, and the purpose is to minimize this value so that the output power of the photovoltaic power station is as close as possible to the AGC signal requirement; ΔP in (k) represents the AGC signal input to the entire PV power station at time k, which also represents the total power adjustment required by the grid dispatch center; n represents the total number of cluster groups in the PV power station; k represents the current time step; h represents the time step index, ranging from k to k+τ; in formula (4), f2 represents the sum of the maximum reactive power reserves of all cluster groups, and the goal is to maximize this value to provide sufficient reactive power regulation capability; Q max,iis the maximum reactive power that the i-th cluster group can provide, which indicates the amount of reactive power that each group can contribute under the current circumstances; in formula (5), w1 and w2 are weight factors used to adjust the importance of the two objectives. These weight factors reflect the relative attention of the PV power station to minimizing power deviation and maximizing reactive power reserve under specific operating conditions.

[0022] Furthermore, in step 3), the AGC signal is a step signal and a continuous step signal.

[0023] Furthermore, in step 5), the regulation capabilities of all inverters in each cluster group are summarized, and the regulation capability of the entire group is obtained by summing the regulation capabilities of each inverter, as shown in formula (10) and formula (11):

[0024]

[0025] where ΔP out,i (k) represents the total power regulation capability of the i-th cluster group at time k, Q out,i (k) represents the total reactive power regulation capability of the i-th cluster group at time k, m i is the number of inverters in the i-th cluster group.

[0026] Furthermore, step 6) specifically includes: setting the population size and the maximum number of iterations, initializing the light propagation parameters Para1 and Para2, selecting the mirage light filtering strategy to control the balance between global search and local search; using the definite integral method to calculate the integral area S of the population quality to determine whether each individual enters the next search process; executing the light propagation strategy to calculate the individual quality factor P and the individual fitness factor q, as shown in the following formula:

[0027]

[0028] Among them, P measures the quality of individuals in the light propagation group. In the power allocation optimization of large-scale photovoltaic power plants, it represents the relative performance of a power allocation scheme in the optimization objective. S is the area of ​​the fitness curve calculated based on the definite integral principle, which represents the overall performance of the power allocation scheme. best

[0029] and S worst Represent the integrated areas of the best and worst solutions respectively; q represents the individual fitness factor, which measures the performance of the power allocation scheme of each cluster group in responding to the grid AGC signal;

[0030] fit i represents the fitness of the i-th cluster group, reflecting the effectiveness of the group in meeting the goals of power deviation and reactive power reserve; fit best and fitworst are the best and worst fitness values ​​in the population, respectively. The light propagation strategy adjusts individual positions through refraction and reflection to optimize power distribution. In the optimization of power distribution in large-scale photovoltaic power plants, the light propagation strategy represents an iterative adjustment process of power distribution. By exploring a better distribution scheme, it ensures that the photovoltaic power plant achieves a balance between different goals, as shown in the following formula:

[0031] x next =x best +x·Para1 (14)

[0032] x next =x rand +x f ·Para2 (15)

[0033] Para1=tan(θ) (16)

[0034]

[0035] The refraction strategy is used for global search, parameter Para1 is responsible for global search, and parameter Para2 is responsible for local search; best is the current best individual; x rand represents a random individual in the population, represents the reference power allocation scheme; tan(θ) represents the refraction angle.

[0036] Furthermore, in step 7), for the bottom-level control model, the bottom-level control further distributes the AGC signal allocated to each cluster group to each inverter in the group according to the output of the top-level control. The allocation strategy is based on the regulation capability of the inverter, as shown in formula (18):

[0037]

[0038] where ΔP in,iw (k) is the power input allocated to the w-th inverter in the i-th cluster group, ΔP max,iw (k) and ΔP min,iw (k) are the maximum and minimum power regulation ranges of the inverter respectively.

[0039] Furthermore, in step 8), within each control cycle, the power deviation is evaluated in real time based on the power output and feedback data of each inverter. If the system detects that the inverter output deviates too much from the set power, the top-level control optimizes the power allocation plan by updating the next cycle. The specific optimization goal is to minimize the power deviation and ensure reactive power reserve, as shown in the following formula:

[0040]

[0041] Where ΔP in (k) represents the deviation of the input power of the photovoltaic power station at time step k; ΔP out,iw (k) represents the output power deviation of the w-th inverter in the i-th cluster group at time step k, and n represents the number of cluster groups. This optimization objective ensures that each inverter achieves more accurate power regulation in the next cycle based on the feedback data.

[0042] By minimizing the power deviation of the photovoltaic power plant, the present invention enables the photovoltaic power plant to quickly and accurately respond to changes in the AGC signal, ensuring that the photovoltaic power plant output power can closely follow the AGC instructions, thereby maintaining the frequency stability of the power grid. Furthermore, reactive power reserves are crucial for regulating the grid voltage, especially when the grid voltage needs to be quickly adjusted to cope with transient load changes. By optimizing reactive power reserves through the rapid responsiveness of the FATA algorithm, the present invention can improve the system's voltage regulation capability and ensure the voltage stability of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of an optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency regulation of a power grid according to an embodiment of the present invention;

[0044] Figure 2 This is a model diagram of an optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency regulation of the power grid according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0046] In this example, a large-scale photovoltaic power station containing a large number of inverters is taken as an example. These inverters need to work together to quickly and accurately adjust the output power under different lighting conditions and grid requirements, respond to AGC signals, and thus maintain the frequency stability and voltage balance of the grid. The optimal power allocation method for a large-scale photovoltaic power station in response to the secondary frequency regulation of the grid proposed in the present invention can improve the AGC response performance of the photovoltaic power station, ensure that the power station can quickly track the AGC signal, optimize the reactive power reserve, and improve the voltage regulation capability of the system. Its flow chart is as follows Figure 1 The specific implementation steps are as follows:

[0047] Step 1: Use the Fuzzy C-Means (FCM) algorithm to cluster the inverters in large-scale photovoltaic power plants, and divide the inverters with similar regulation characteristics into multiple groups to simplify the overall optimization process. FCM is a clustering method based on fuzzy sets, which allows each data point to belong to multiple clusters to different degrees (fuzzy membership). FCM can handle the differences between inverter characteristics more flexibly, so that clustering results that are more suitable for actual application scenarios may be obtained. In this step, the FCM algorithm collects the characteristic data of all inverters in the photovoltaic power plant.

[0048]

[0049] Among them, U in formula (1) ij Indicates the membership degree of the i-th inverter to the j-th cluster; ||x i -v j || represents the Euclidean distance between the i-th inverter and the j-th cluster center; m is the fuzzification coefficient, which is used to control the fuzziness of the membership degree. Usually m>1. When m increases, the fuzziness of the membership degree also increases. Formula (2) is used to calculate the distance between the inverter and the cluster center v j The Euclidean distance is used to measure the similarity between the inverter and the cluster center. The smaller the distance, the more similar they are. In formula (2), D is the dimension of the feature vector; c is the number of clusters; x id represents the dth eigenvalue of the i-th inverter; v jd Denotes the dth eigenvalue of the jth inverter. The membership matrix of the inverter is calculated by formula (1) and formula (2), thereby determining the cluster group to which each inverter belongs.

[0050] Step 2: Based on the clustering results of the inverters, the large-scale photovoltaic power station problem is decomposed into top-level control and bottom-level control. The top-level control is responsible for the power distribution between multiple cluster groups, while the bottom-level control is responsible for the power distribution within each group. The top-level and bottom-level control models are established. The control model is as follows: Figure 2 As shown in Figure 2, the task of top-level control is to optimize power distribution to meet the dual-objective optimization model, i.e., minimizing power deviation and maximizing reactive power reserve. The optimization problem of top-level control can be expressed as:

[0051]

[0052] f=w1·f1+w2·(1-f2) (5)

[0053] In formula (3), f1 represents the sum of power deviations, and the purpose is to minimize this value so that the output power of the photovoltaic power station is as close as possible to the AGC signal requirement; ΔP in(k) represents the AGC signal input to the entire PV power station at time k, which also represents the total power adjustment required by the grid dispatching center; n represents the total number of cluster groups in the PV power station; k represents the current time step;

[0054] h represents the time step index, ranging from k to k+τ. In formula (4), f2 represents the sum of the maximum reactive power reserves of all cluster groups. The goal is to maximize this value to provide sufficient reactive power regulation capability; Q max,i is the maximum reactive power that the i-th cluster group can provide, which represents the amount of reactive power that each group can contribute under the current circumstances. In formula (5), w1 and w2 are weight factors used to adjust the importance of the two objectives. These weight factors reflect the relative attention of the PV power plant to minimizing power deviation and maximizing reactive power reserve under specific operating conditions.

[0055] According to the above optimization objectives, the top-level control converts the total AGC signal ΔP in Assigned to each cluster group i, the allocation result is limited by the power regulation capability of each group, as shown in formula (6):

[0056] ΔP min,i (k)≤ΔP in,i (k)≤ΔP max,i (k), i=1,2,...,n (6)

[0057] where ΔP min,i (k) represents the minimum adjustment power range of the i-th cluster group at time k;

[0058] ΔP max,i (k) represents the maximum adjustment power range of the i-th cluster group at time k; ΔP in,i (k)

[0059] is the AGC signal assigned to the i-th cluster group, indicating the power adjustment task assigned to the group by the top-level control.

[0060] Step 3: Obtain the current PV plant's AGC signal. This signal is provided by the grid dispatch center and indicates the required power adjustment. This AGC signal serves as input for subsequent power allocation and optimization steps. The AGC signal serves as an input parameter for top-level control, guiding the overall power regulation of the PV plant. The AGC signal can take various forms, such as step signals and continuous step signals. After receiving the AGC signal, the PV plant calculates the regulation capability of each cluster group based on the current inverter operating status and regulation capability.

[0061] Step 4: Calculate the regulation capability of each inverter at the current moment to determine the maximum and minimum power output ranges it can adjust in response to the AGC signal. These capabilities are limited by the inverter's current operating state and device limitations, such as maximum and minimum power output ranges.

[0062] The regulation capability of each inverter is as follows:

[0063]

[0064] ΔP in (k) = P in (k)-P in (k-1) (9)

[0065] Among them, P in,iw (k-1) is the current input power of the inverter at time (k-1); ΔP in (k) represents the total power regulation. Using the above formula, we can calculate the regulation capability of each inverter and use this to allocate power appropriately. This approach ensures that PV power plants can quickly and accurately adjust their output power as grid demand fluctuates, thereby maintaining grid stability and reliability.

[0066] Step 5: Summarize the regulation capabilities of all inverters in each cluster group and obtain the regulation capability of the entire group by summing the regulation capabilities of each inverter, as shown in Equations (10) and (11). These aggregated regulation capabilities are used for power allocation in the top-level control to ensure that the power allocation of each group is reasonable and can meet the overall regulation requirements of the system.

[0067]

[0068] where ΔP out,i (k) represents the total power regulation capability of the i-th cluster group at time k. out,i (k) represents the total reactive power regulation capability of the i-th cluster group at time k. i is the number of inverters in the i-th cluster group.

[0069] Step 6: Use the FATA algorithm to optimize in the top-level control and initialize the algorithm parameters. This step aims to optimize the power deviation and reactive power reserve of the photovoltaic power station. Set the population size and maximum number of iterations, initialize the light propagation parameters Para1 and Para2, and select the mirage light filtering strategy to control the balance between global search and local search. Use the definite integral method to calculate the integral area S of the population quality and decide whether each individual enters the next step.

[0070] One-step search process. Execute the light propagation strategy and calculate the individual quality factor P and individual fitness factor q, as shown in the following formula:

[0071]

[0072] Among them, P measures the quality of individuals in the light propagation group. In the power allocation optimization of large-scale photovoltaic power plants, it represents the relative performance of a power allocation scheme in the optimization objective. S is the area of ​​the fitness curve calculated based on the definite integral principle, which represents the overall performance of the power allocation scheme. best

[0073] and S worst Represent the integrated areas of the best and worst solutions respectively; q represents the individual fitness factor, which measures the performance of the power allocation scheme of each cluster group in responding to the grid AGC signal;

[0074] fit i It represents the fitness of the i-th cluster group, reflecting the effectiveness of the group in meeting the goals of power deviation and reactive power reserve. best and fit worst are the best and worst fitness values ​​in the population, respectively.

[0075] Light propagation strategies adjust individual positions through refraction and reflection processes to optimize power distribution.

[0076] In the optimization of power allocation in large-scale photovoltaic power plants, the light propagation strategy represents an iterative adjustment process of power allocation. By exploring a better allocation scheme, it ensures that the photovoltaic power plant achieves a balance between different objectives, as shown in the following formula:

[0077] x next =x best +x·Para1 (14)

[0078] x next =x rand +x f ·Para2 (15)

[0079] Para1=tan(θ) (16)

[0080]

[0081] The refraction strategy is used for global search, parameter Para1 is responsible for global search, and parameter Para2

[0082] Responsible for local search. x best is the current best individual; x randrepresents a random individual in the population, represents the reference power allocation scheme, and tan(θ) represents the refraction angle. These parameters play a key role in optimizing power allocation in PV power plants, helping to strike a balance between global and local search, ensuring that the inverter group can effectively respond to AGC signals and optimize power regulation on the grid.

[0083] Step 7: The optimal solution obtained in the top-level control is transferred to the bottom-level control, and the AGC signal is proportionally distributed to each inverter in the group. For the bottom-level control model, the bottom-level control further distributes the AGC signal allocated to each cluster group to each inverter in the group based on the output of the top-level control. The allocation strategy is based on the regulation capability of the inverter, as shown in formula (18):

[0084]

[0085] where ΔP in,iw (k) is the power input allocated to the w-th inverter in the i-th cluster group, ΔP max,iw (k) and ΔP min,iw (k) are the maximum and minimum power regulation ranges of the inverter respectively.

[0086] Step 8: Output the optimal regulated power input for each inverter, monitor the inverter response in real time, and make necessary adjustments and optimizations in the next control cycle based on the feedback data. In each control cycle, the power deviation is evaluated in real time based on the power output and feedback data of each inverter. If the system detects that the inverter output deviates too much from the set power, the top-level control will optimize by updating the power allocation plan for the next cycle. The specific optimization goal is to minimize the power deviation and ensure reactive power reserve, as shown in the following formula

[0087]

[0088] Where ΔP in (k) represents the deviation of the input power of the photovoltaic power station at time step k; ΔP out,iw (k) represents the output power deviation of the wth inverter in the ith cluster at time step k. n represents the number of clusters. This optimization objective ensures that each inverter achieves more accurate power regulation in the next cycle based on feedback data.

[0089] The specific model diagram of the above steps is as follows Figure 2 shown.

[0090] The present invention has the following advantages and effects compared to the prior art:

[0091] (1) This paper proposes a clustered hierarchical power control method based on the mirage algorithm, which can simultaneously optimize the AGC response performance and reactive power reserve. Compared with traditional methods, this method has better optimization effects and provides greater adjustment flexibility for the dynamic operation of the power station.

[0092] (2) Compared with the existing optimization scheduling method of large-scale photovoltaic power stations, by introducing the FCM clustering algorithm based on adjusting performance parameters, the present invention reduces the optimization complexity of the underlying control and significantly improves the computational efficiency of AGC scheduling of large-scale power stations.

[0093] (3) The present invention proposes an optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency regulation of the power grid. By finding a balance between the two optimization objectives, it can provide a larger reactive power reserve while minimizing the power deviation.

[0094] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency modulation of the power grid, characterized in that: The following steps are involved: 1) Use the FCM algorithm to cluster inverters in all large-scale photovoltaic power plants and divide inverters with similar regulation characteristics into multiple groups; 2) Based on the inverter clustering results, the large-scale photovoltaic power station problem is decomposed into top-level control and bottom-level control. The top-level control is responsible for power distribution between multiple cluster groups, while the bottom-level control is responsible for power distribution within each group. Top-level and bottom-level control models are established. The task of the top-level control is to optimize power distribution to meet the dual-objective optimization model of minimizing power deviation and maximizing reactive power reserve. 3) Obtaining the AGC signal of the current photovoltaic power station, which is provided by the power grid dispatching center and is used to indicate the required power adjustment amount; 4) The photovoltaic power station calculates the regulation capacity of each inverter based on the current operating status and regulation capacity of the inverter; 5) Summarize the regulation capabilities of all inverters in each cluster group, obtain the regulation capability of the entire group by summing the regulation capabilities of each inverter, and allocate power in the top-level control based on the summarized regulation capability of the entire group to ensure that the power allocation of each group is reasonable and can meet the overall regulation requirements of the system; 6) Use the FATA algorithm to optimize in the top-level control and initialize the parameters of the FATA algorithm; 7) The optimal solution obtained in the top-level control is transmitted to the bottom-level control, and the AGC signal is proportionally distributed to each inverter in the group; 8) Output the optimal regulated power input for each inverter, monitor the inverter response in real time, and make adjustments and optimizations in the next control cycle based on the feedback data.

2. The optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency modulation of the power grid according to claim 1 is characterized in that: In step 1), the FCM algorithm collects characteristic data of all inverters in the PV power station: （1）； （2）； Among them, in formula (1) Indicates the membership degree of the i-th inverter to the j-th cluster; represents the Euclidean distance between the i-th inverter and the j-th cluster center; 𝑚 is the fuzzification coefficient, which is used to control the fuzziness of the membership. Usually 𝑚>

1. When 𝑚 increases, the fuzziness of the membership also increases. Formula (2) is used to calculate the distance between the inverter and the cluster center. , is used to measure the similarity between the inverter and the cluster center. The smaller the distance, the more similar they are. In formula (2), D is the dimension of the feature vector; c represents the number of clusters. represents the d-th eigenvalue of the i-th inverter; represents the dth eigenvalue of the jth inverter; the membership matrix of the inverter is calculated by formula (1) and formula (2), thereby determining the cluster group to which each inverter belongs.

3. The optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency modulation of the power grid according to claim 1, characterized in that: In step 2), the optimization problem of top-level control is expressed as: （3）； （4）； （5）； In formula (3) Represents the sum of power deviations. The goal is to minimize this value so that the output power of the PV plant is as close as possible to the AGC signal requirement. represents the AGC signal input to the entire PV power station at time k, which also represents the total power adjustment required by the grid dispatching center; n represents the total number of cluster groups in the PV power station; k represents the current time step; h represents the time step index, ranging from k to k+τ; in formula (4), It represents the sum of the maximum reactive power reserves of all cluster groups. The goal is to maximize this value to provide sufficient reactive power regulation capability. is the maximum reactive power that the i-th cluster group can provide, which means the amount of reactive power that each group can contribute under the current circumstances; in formula (5) and are weight factors used to adjust the importance of the two objectives. These weight factors reflect the relative importance of the PV power plant to minimizing power deviation and maximizing reactive power reserve under specific operating conditions.

4. The optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency modulation of the power grid according to claim 1, characterized in that: In step 3), the AGC signal is a step signal and a continuous step signal.

5. The optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency modulation of the power grid according to claim 1, characterized in that: In step 5), the regulation capabilities of all inverters in each cluster group are summarized, and the regulation capability of the entire group is obtained by summing the regulation capabilities of each inverter, as shown in formula (10) and formula (11): （10）； （11）； in represents the total power regulation capability of the i-th cluster group at time k, represents the total reactive power regulation capability of the i-th cluster group at time k, is the number of inverters in the i-th cluster group.

6. The optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency modulation of the power grid according to claim 1, characterized in that: Step 6) specifically includes: setting the population size and the maximum number of iterations, initializing the light propagation parameters Para1 and Para2, selecting the mirage light filtering strategy to control the balance between global search and local search; using the definite integral method to calculate the integral area S of the population quality to determine whether each individual enters the next search process; executing the light propagation strategy to calculate the individual quality factor P and the individual fitness factor q, as shown in the following formula: （12）； （13） P measures the quality of individuals in a light propagation group. In the optimization of power allocation in large-scale photovoltaic power plants, it indicates the relative performance of a power allocation scheme in terms of the optimization objective. S is the area of ​​the fitness curve calculated based on the definite integral principle and represents the overall performance of the power allocation scheme. and Represent the integrated areas of the best and worst solutions respectively; q represents the individual fitness factor, which measures the performance of the power allocation scheme of each cluster group in responding to the grid AGC signal; represents the fitness of the i-th cluster group, reflecting the effectiveness of the group in meeting the goals of power deviation and reactive power reserve; and are the best and worst fitness values ​​in the population, respectively. The light propagation strategy adjusts individual positions through refraction and reflection to optimize power distribution. In the optimization of power distribution in large-scale photovoltaic power plants, the light propagation strategy represents an iterative adjustment process of power distribution. By exploring a better distribution scheme, it ensures that the photovoltaic power plant achieves a balance between different goals, as shown in the following formula: （14）； （15）； （16）； （17）； The refraction strategy is used for global search, parameter Para1 is responsible for global search, and parameter Para2 is responsible for local search; Is the best individual at the moment; represents a random individual in the population, represents the reference power allocation scheme; Represents the refraction angle.

7. The optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency modulation of the power grid according to claim 1, characterized in that: In step 7), for the bottom-level control model, the bottom-level control further distributes the AGC signal assigned to each cluster group to each inverter in the group according to the output of the top-level control. The distribution strategy is based on the regulation capability of the inverter, as shown in formula (18): （18）； in is the power input allocated to the w-th inverter in the ith cluster group, and are the maximum and minimum power adjustment ranges of the inverter respectively.

8. The optimal power allocation method for a large-scale photovoltaic power station in response to secondary frequency modulation of the power grid according to claim 1, characterized in that: In step 8), within each control cycle, the power deviation is evaluated in real time based on the power output and feedback data of each inverter. If the system detects that the inverter output deviates too much from the set power, the top-level control optimizes the power allocation plan by updating the next cycle. The specific optimization goal is to minimize the power deviation and ensure reactive power reserve, as shown in the following formula: （19）； in, Indicates that at time step The deviation of the input power of the photovoltaic power station at the moment; Indicates the The first The inverter is in the time step The output power deviation at the moment, represents the number of clustering groups. This optimization objective ensures that each inverter achieves more accurate power regulation in the next cycle based on the feedback data.

Images