Capacity optimization method for virtual power plant participating in power grid primary frequency modulation auxiliary service

Abstract

The invention belongs to the technical field of power grid regulation and control, and particularly relates to a capacity optimization method for a virtual power plant participating in power grid primary frequency modulation auxiliary service, which comprises the following steps: S0, constructing a frequency response model of a traditional unit in a power grid; s1, constructing a frequency response model of the power-type virtual power plant; s2, combining the frequency response model of the traditional unit in the step S0 with the frequency response model of the power type virtual power plant in the step S1, and constructing an improved system frequency response model considering the power type virtual power plant; s3, taking the lowest frequency modulation cost and the fastest frequency modulation speed as objective functions, and establishing a capacity optimization model of the virtual power plant participating in the primary frequency modulation auxiliary service; and S4, solving the capacity optimization model in the S3 to obtain a Pareto solution set formed by a plurality of non-dominated capacity schemes, and selecting one capacity scheme from the Pareto solution set as a final capacity scheme based on a preset robustness evaluation index. According to the method, the optimal capacity configuration scheme with economical efficiency, rapidity and reliability can be obtained.

Description

Technical Field

[0001] This invention belongs to the field of power grid control technology, and in particular relates to a capacity optimization method for virtual power plants participating in primary frequency regulation ancillary services of the power grid. Background Technology

[0002] With the accelerated global energy transition, the penetration rate of fluctuating and intermittent renewable energy sources such as wind power and solar power in the power system continues to rise. These new energy sources are generally connected to the grid through power electronic converters, lacking the physical rotational inertia of traditional synchronous generator sets. This makes them unable to provide effective inertial response and primary frequency regulation support when system frequency is disturbed. Their large-scale integration leads to a significant decrease in the overall equivalent rotational inertia of the power grid, weakening the system's ability to withstand disturbances, and making frequency fluctuations more severe and frequent. This results in a sharp increase in the demand for primary frequency regulation resources, while the supply of traditional frequency regulation resources faces a structural shortage, exacerbating the supply-demand imbalance.

[0003] Faced with this challenge, relying solely on the construction or renovation of traditional frequency regulation power sources (such as the flexible transformation of thermal power plants and pumped storage power stations) not only results in high investment costs and long construction cycles, but also contradicts the development direction of promoting green and low-carbon transformation under the "dual carbon" goals. Meanwhile, a massive amount of distributed regulation resources are emerging on the grid load side, including electrochemical energy storage, distributed photovoltaic / wind power, adjustable loads, and flexible loads with interruptibility capabilities. Although these resources are small in scale, dispersed, and diverse in characteristics, they can be integrated into a virtual power plant (VPP) with observable, measurable, and controllable capabilities through advanced information communication, collaborative control, and aggregation optimization technologies. In recent years, as the mechanism for virtual power plants to participate in the electricity ancillary services market has gradually improved, the technical feasibility of their participation in primary frequency regulation has been initially verified, and the industry's focus is shifting from "whether they can participate" to "how to participate efficiently and economically."

[0004] However, the regulation capacity of virtual power plants is significantly limited and time-varying: energy storage systems are constrained by state of charge (SOC), and their regulation capacity changes dynamically with the charging and discharging process; mobile resources such as electric vehicles have access uncertainties; and the regulation depth and duration of interruptible loads are limited by user willingness and equipment characteristics. Simply treating virtual power plants as fixed-capacity frequency-regulating resources and including them in system scheduling can easily lead to model distortion, insufficient response, and even default risks.

[0005] Building upon this foundation, the key to enhancing the market competitiveness and operational reliability of virtual power plants lies in how to scientifically and accurately optimize the available frequency regulation capacity while comprehensively considering multiple objectives such as frequency regulation cost and response speed, all while meeting system frequency security constraints. Existing research largely focuses on single-objective optimization (e.g., minimizing cost only) or static capacity setting, neglecting the trade-off between frequency regulation performance and economy, and lacking systematic evaluation of the robustness of the proposed solutions. Furthermore, since primary frequency regulation requires extremely high response speeds (typically on the order of seconds), the model must balance computational efficiency and physical accuracy, further increasing the complexity of the optimization problem.

[0006] Therefore, how to obtain the optimal capacity configuration scheme that combines economy, speed and reliability has become an urgent problem to be solved. Summary of the Invention

[0007] To address the shortcomings of the existing technologies, this invention provides a capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services, which can obtain an optimal capacity configuration scheme that combines economy, speed, and reliability.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0009] A capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services includes the following steps:

[0010] S0. Construct a frequency response model for traditional generating units in the power grid;

[0011] S1. Power-type regulating resources are used to form a power-type virtual power plant to participate in the primary frequency regulation of the power grid; a frequency response model of the power-type virtual power plant is constructed; wherein, the power-type regulating resources include energy storage, distributed new energy sources, and interruptible loads;

[0012] S2. Combining the frequency response model of the traditional unit in S0 and the frequency response model of the power-type virtual power plant in S1, an improved system frequency response model considering the power-type virtual power plant is constructed.

[0013] S3. Based on the improved system frequency response model constructed by S2, a capacity optimization model for virtual power plants participating in primary frequency regulation ancillary services is established with the objective functions of minimizing frequency regulation cost and maximizing frequency regulation speed.

[0014] S4. Solve the capacity optimization model of S3 to obtain a Pareto solution set consisting of multiple non-dominant capacity schemes, and select a capacity scheme from the Pareto solution set as the final capacity scheme based on a preset robustness evaluation index.

[0015] Compared with the prior art, the present invention has the following advantages:

[0016] 1. Improved accuracy and representativeness of the frequency response model. Traditional methods often simplify virtual power plants to a fixed capacity or ignore their dynamic characteristics, making it difficult to accurately reflect their actual regulation capabilities under frequency disturbances. This scheme constructs refined frequency response models for traditional generating units and power-type virtual power plants respectively, and integrates the two to form an improved system frequency response model. This model can more accurately depict the dynamic behavior of power systems with a high proportion of distributed resources during primary frequency regulation, providing a reliable foundation for subsequent optimization.

[0017] 2. This solution achieves synergistic optimization of frequency modulation economy and response performance. Existing capacity allocation methods often focus on a single objective (such as minimizing cost only), which can easily lead to insufficient response speed or resource redundancy. This solution establishes a multi-objective capacity optimization model with "lowest frequency modulation cost" and "fastest frequency modulation speed" as dual objectives, effectively balancing the contradiction between economy and dynamic performance. Under the premise of meeting system frequency security, it avoids the problems of over-configuration or response lag.

[0018] 3. Enhanced robustness and practicality of capacity configuration schemes. For multiple feasible solutions existing in the Pareto optimal solution set, this scheme introduces a pre-defined robustness evaluation index for secondary screening, comprehensively considering the impact of factors such as load fluctuations and resource availability uncertainty on the stability of the scheme. Compared to directly selecting extreme solutions (such as pure cost optimal or pure speed optimal), this strategy significantly improves the adaptability and reliability of the final capacity scheme in actual operating environments.

[0019] 4. This approach promotes the transformation of virtual power plants from "participation-oriented" to "highly efficient participation." Traditional virtual power plant participation in frequency regulation often relies on experience-based settings or static rules, lacking deep coupling with the overall system frequency dynamics. This method embeds the dynamic adjustment capabilities of virtual power plants into the system frequency response framework and performs capacity optimization based on this framework, enabling virtual power plants to truly become quantifiable, dispatchable, and reliable frequency regulation entities, providing technical support for their large-scale and commercial application in the ancillary services market.

[0020] In summary, this method can achieve scientific, efficient, and robust configuration of virtual power plant frequency regulation capacity. While ensuring grid frequency stability, it also takes into account economy and response performance, significantly improving the overall efficiency and market competitiveness of distributed resource aggregates participating in primary frequency regulation ancillary services.

[0021] Preferably, in S1, energy storage and distributed new energy sources participate in primary frequency regulation using droop control; interruptible loads support the frequency by being temporarily disconnected or having their power reduced in the event of a frequency drop.

[0022] Preferably, in S1, the frequency response model of the power-type virtual power plant includes:

[0023]

[0024] In the formula, Δf is the frequency deviation; s is the Laplace operator used for frequency domain modeling, representing the complex frequency variable; This represents the active power increment of the j-th energy storage unit; and These are the frequency regulation coefficient and time constant of the j-th energy storage unit, respectively; This represents the increase in active power of the k-th distributed renewable energy source. and These are the frequency regulation coefficient and time constant of the k-th distributed renewable energy source, respectively; The power reduction due to load l; Interruptible power set for load l; T l IL The interruption delay for load l.

[0025] Preferably, in S0, the frequency response model of traditional generating units in the power grid is:

[0026]

[0027] In the formula, ΔP i G (s) represents the active power increment of the i-th conventional unit; and These are the frequency regulation coefficient and time constant of the i-th conventional unit, respectively; Let be the high-pressure cylinder power generation ratio of the i-th conventional unit.

[0028] This setup achieves two key advantages: 1) Differentiated modeling and collaborative representation of different types of regulatory resources. Traditional methods often treat virtual power plants as a single aggregate, ignoring the dynamic differences in internal resources. This solution, by establishing separate frequency response models for energy storage, distributed renewable energy, and interruptible loads (e.g., using first-order inertial elements to describe the frequency regulation behavior of energy storage and renewable energy, and using transfer functions with delay terms to characterize the response characteristics of interruptible loads), more realistically reflects the response mechanisms and time characteristics of various resources under frequency disturbances, improving the physical rationality and engineering applicability of the models.

[0029] 2. Enhanced adaptability and accuracy of the system frequency response model. Existing system frequency response models are mainly designed for traditional synchronous generator units and are difficult to adapt to new power systems with a high proportion of power electronic interface power sources. This solution introduces a power-type virtual power plant frequency response model and couples it with the traditional generator unit model to construct an improved system frequency response model that includes various heterogeneous resources. This effectively compensates for the shortcomings of traditional models in terms of inertia loss and response speed variation, and improves the prediction accuracy for complex power grid dynamic behavior.

[0030] 3. Energy storage and distributed renewable energy sources participate in frequency regulation using droop control, while interruptible loads are triggered to disconnect or reduce power when the frequency drops. This design aligns with actual operating logic. This strategy enables virtual power plants to have "plug-and-play" frequency support capabilities similar to traditional generating units, while also taking into account the flexibility and controllability of resources, thus improving their reliability and practicality as frequency regulation resources.

[0031] Preferably, in S2, the improved system frequency response model of the power-type virtual power plant is constructed as follows:

[0032]

[0033] In the formula, ΔP represents the power deficit of the power grid; H sys D and T represent the equivalent inertia and damping of the power grid, respectively; IL K is the time constant for aggregating interruptible loads. G T G F G These are the frequency regulation coefficient, time constant, and high-pressure cylinder power generation ratio of the aggregated traditional unit, respectively; K E T E These are the frequency regulation coefficient and time constant of aggregated energy storage, respectively; K R T R These are the frequency regulation coefficient and time constant of aggregated distributed new energy sources, respectively; This refers to the rated capacity of traditional unit i; The rated capacity for energy storage j to participate in primary frequency regulation auxiliary services; The rated capacity for distributed renewable energy k to participate in primary frequency regulation auxiliary services; The rated capacity for interruptible load l to participate in primary frequency regulation ancillary services; N G N E N R N IL These represent the number of traditional generating units participating in primary frequency regulation of the power grid, the amount of energy storage, the amount of distributed renewable energy, and the amount of interruptible loads; H i Let be the inertial constant of the i-th conventional unit.

[0034] With this setup, the method can achieve accurate modeling of the dynamic frequency behavior of power systems containing virtual power plants, fully integrate the dynamic characteristics of various regulation resources, improve the accuracy and completeness of system frequency response analysis, and provide solid technical support for the efficient configuration and operation of virtual power plants in primary frequency regulation ancillary services.

[0035] Preferably, in S3, the objective function of the capacity optimization model is:

[0036] minF(f1,f2);

[0037]

[0038] In the formula, C represents the frequency modulation cost; C VPP and C G These are the frequency regulation costs of aggregated resources and traditional generating units, respectively; c G c E c R c IL These are the unit frequency regulation costs for traditional generating units, energy storage, distributed renewable energy, and interruptible loads, respectively; t lim For frequency modulation speed; Δf lim This represents the maximum allowable steady-state frequency deviation of the power grid.

[0039] This setup addresses two issues: 1. Traditional capacity configuration methods typically focus on a single objective (such as minimizing operating costs), which can easily lead to insufficient response speed or resource redundancy. This solution introduces a dual-objective function, quantifying both the total cost of frequency regulation and the system frequency change rate separately. By integrating economic efficiency and response performance into the optimization framework, it can balance cost control and rapid adjustment capabilities while meeting frequency security requirements, thereby improving overall operating efficiency.

[0040] 2. The objective function assigns unit frequency regulation cost coefficients to different resource types (traditional generating units, energy storage, distributed renewable energy, and interruptible loads), and performs weighted calculations based on their rated capacity. This approach accurately reflects the economic differences among various resources in frequency regulation services. This differentiated modeling method makes the optimization process more closely aligned with market mechanisms, helping to guide rational resource investment and improve resource allocation efficiency.

[0041] Preferably, in S3, the constraints of the capacity optimization model include:

[0042]

[0043] In the formula, Nad max Maximum frequency deviation threshold; RoC max The threshold for the rate of change of frequency; and These are the upper and lower limits of the energy storage capacity, respectively. and These are the upper and lower limits of distributed renewable energy capacity, respectively. and These represent the upper and lower limits of the interruptible load capacity.

[0044] Preferably, in S4, the preset robustness evaluation index is determined in the following way:

[0045] For each capacity scheme in the Pareto solution set, multiple uncertainty scenarios are generated. The disturbance capacity schemes under each scenario are input into the improved system frequency response model to calculate the corresponding frequency deviation and frequency change rate.

[0046] The robustness evaluation index is the maximum comprehensive deviation R between the frequency response and the benchmark response under each uncertainty scenario.

[0047] This setup has several advantages: 1. It enhances the adaptability of capacity optimization results to real-world operational uncertainties. Traditional optimization methods typically solve for the optimal solution based on deterministic assumptions, neglecting real-world uncertainties such as fluctuations in energy storage SOC, fluctuations in renewable energy output, and delays in load interruption response. This approach generates multiple uncertainty scenarios for each Pareto solution, simulating the impact of resource availability changes on the system's frequency response. This effectively identifies capacity schemes that perform stably in complex dynamic environments, significantly enhancing the anti-interference capability and operational reliability of the selected scheme.

[0048] 2. This approach achieves a decision-making upgrade from "static optimality" to "dynamic robustness." While solutions in the Pareto solution set are independent under ideal conditions, their actual performance can vary significantly due to external disturbances. This approach introduces a robustness evaluation mechanism, linking the theoretically optimal solution with actual operational risks. This avoids selecting "fragile solutions" that perform well only under ideal conditions but fail under disturbances, thus shifting decision-making from pursuing the optimal single performance indicator to considering both stability and robustness.

[0049] Preferably, the maximum comprehensive deviation R is:

[0050]

[0051] In the formula, N U r represents the number of uncertain scenarios. x Let be the comprehensive deviation under the x-th uncertainty scenario; Δf and The frequency deviation and its rate of change under undisturbed conditions, Δf x and This is the corresponding value under the x-th uncertainty scenario;

[0052] The capacity scheme with the smallest R value in the Pareto solution set is selected as the final capacity scheme.

[0053] This setup enables a refined robustness assessment of virtual power plant frequency regulation capacity schemes under various disturbance scenarios. By introducing a comprehensive deviation index that takes into account both steady-state and dynamic responses, it scientifically selects capacity configurations that still have excellent performance under the worst conditions.

[0054] Preferably, in S4, the capacity optimization model is solved using the heuristic multi-objective particle swarm optimization algorithm. Attached Figure Description

[0055] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0056] Figure 1 This is a flowchart of the method;

[0057] Figure 2 This is a flowchart illustrating the heuristic multi-objective particle swarm optimization algorithm used in this embodiment to solve the capacity optimization model. Detailed Implementation

[0058] The following detailed explanation illustrates the specific implementation methods:

[0059] Example:

[0060] like Figure 1 As shown in the figure, this embodiment discloses a capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services, including the following steps:

[0061] S0. Construct a frequency response model for traditional generating units in the power grid.

[0062] As a stable source of frequency regulation resources and the fundamental source of system inertia, traditional generating units have laid a solid foundation for primary frequency regulation of the power grid, and their position remains irreplaceable. In practical implementation, the frequency response model of traditional generating units in the power grid is as follows:

[0063]

[0064] In the formula, ΔP i G (s) represents the active power increment of the i-th conventional unit; and These are the frequency regulation coefficient and time constant of the i-th conventional unit, respectively; Let be the high-pressure cylinder power generation ratio of the i-th conventional unit.

[0065] S1. Power-type regulating resources are used to form a power-type virtual power plant to participate in the primary frequency regulation of the power grid; a frequency response model of the power-type virtual power plant is constructed; wherein, the power-type regulating resources include energy storage, distributed new energy sources, and interruptible loads; energy storage and distributed new energy sources participate in primary frequency regulation using droop control; interruptible loads support the frequency by being temporarily disconnected or having their power reduced in the event of a frequency drop.

[0066] The core capability of a virtual power plant lies in aggregating and coordinating the control of massive, dispersed flexible resources (distributed renewable energy, energy storage, controllable loads, electric vehicles, etc.), making it appear externally as a unified and controllable "giant power plant." This provides the foundation for its participation in primary frequency regulation of the power grid. Given the rapid and short-term nature of primary frequency regulation, this method employs power-type regulating resources (fast response, small capacity): energy storage, distributed renewable energy, and interruptible loads, to form a virtual power plant participating in primary frequency regulation—that is, a power-type virtual power plant.

[0067] In specific implementation, the frequency response model of the power-type virtual power plant includes:

[0068]

[0069] In the formula, Δf is the frequency deviation; s is the Laplace operator used for frequency domain modeling, representing the complex frequency variable; This represents the active power increment of the j-th energy storage unit; and These are the frequency regulation coefficient and time constant of the j-th energy storage unit, respectively; This represents the increase in active power of the k-th distributed renewable energy source. and These are the frequency regulation coefficient and time constant of the k-th distributed renewable energy source, respectively; The power reduction due to load l; Interruptible power set for load l; T l IL The interruption delay for load l.

[0070] Traditional methods often treat virtual power plants as a single aggregate, ignoring the dynamic differences in the internal resources. This solution establishes separate frequency response models for energy storage, distributed renewable energy, and interruptible loads (e.g., using first-order inertial elements to describe the frequency regulation behavior of energy storage and renewable energy, and using transfer functions with delay terms to characterize the response characteristics of interruptible loads). This more realistically reflects the response mechanisms and time characteristics of various resources under frequency disturbances, improving the physical rationality and engineering applicability of the models. Furthermore, existing system frequency response models are mainly geared towards traditional synchronous generator units and are difficult to adapt to new power systems with a high proportion of power electronic interface power sources. This solution introduces a power-type virtual power plant frequency response model and couples it with a traditional generator unit model to construct an improved system frequency response model that includes various heterogeneous resources. This effectively compensates for the shortcomings of traditional models in terms of inertia and response speed variations, improving the prediction accuracy for complex grid dynamic behavior. Moreover, energy storage and distributed renewable energy participate in frequency regulation using droop control, while interruptible loads are triggered to disconnect or reduce power when the frequency drops. This design aligns with actual operating logic. This strategy enables virtual power plants to have "plug-and-play" frequency support capabilities similar to traditional generating units, while taking into account the flexibility and controllability of resources, thus improving their reliability and practicality as frequency regulation resources.

[0071] S2. Combining the frequency response model of the traditional unit in S0 and the frequency response model of the power-type virtual power plant in S1, an improved system frequency response model considering the power-type virtual power plant is constructed.

[0072] In practical implementation, the dynamic change process of the grid frequency, taking into account the frequency regulation of power-type virtual power plants, can be described by a first-order oscillation equation, as follows:

[0073]

[0074] In the formula, H sys D and ΔP represent the equivalent inertia and damping of the power grid, respectively; ΔP represents the power deficit of the power grid; N G N E N R N IL These are the number of traditional generating units participating in the primary frequency regulation of the power grid, the number of energy storage units, the number of distributed renewable energy sources, and the number of interruptible loads.

[0075] Performing a frequency domain transformation on the above equation yields the improved system frequency response model considering the power-type virtual power plant:

[0076]

[0077] Substituting the frequency response models of power-type virtual power plants and traditional generating units into the above equation, and aggregating various frequency regulation resources, we obtain the final improved system frequency response model:

[0078] Substituting the frequency response models of the power-type virtual power plant and traditional generating units into the above equation, and aggregating various frequency regulation resources, we obtain the improved system frequency response model considering the power-type virtual power plant as follows:

[0079]

[0080]

[0081] In the formula, ΔP represents the power deficit of the power grid; H sys D and T represent the equivalent inertia and damping of the power grid, respectively; IL K is the time constant for aggregating interruptible loads. G T G F G These are the frequency regulation coefficient, time constant, and high-pressure cylinder power generation ratio of the aggregated traditional unit, respectively; K E T E These are the frequency regulation coefficient and time constant of aggregated energy storage, respectively; K R T R These are the frequency regulation coefficient and time constant of aggregated distributed new energy sources, respectively; This refers to the rated capacity of traditional unit i; The rated capacity for energy storage j to participate in primary frequency regulation auxiliary services; The rated capacity for distributed renewable energy k to participate in primary frequency regulation auxiliary services; The rated capacity for interruptible load l to participate in primary frequency regulation ancillary services; N G N E N R N IL These represent the number of traditional generating units participating in primary frequency regulation of the power grid, the amount of energy storage, the amount of distributed renewable energy, and the amount of interruptible loads; H i Let be the inertial constant of the i-th conventional unit.

[0082] This enables accurate modeling of the dynamic frequency behavior of power systems containing virtual power plants, fully integrates the dynamic characteristics of various regulation resources, improves the accuracy and completeness of system frequency response analysis, and provides solid technical support for the efficient configuration and operation of virtual power plants in primary frequency regulation ancillary services.

[0083] S3. Based on the improved system frequency response model constructed in S2, a capacity optimization model for virtual power plants participating in primary frequency regulation ancillary services is established with the objective functions of minimizing frequency regulation cost and maximizing frequency regulation speed.

[0084] In practical implementation, the objective function of the capacity optimization model is:

[0085] minF(f1,f2);

[0086]

[0087] In the formula, C represents the frequency modulation cost; C VPP and C G These are the frequency regulation costs of aggregated resources and traditional generating units, respectively; c G c E c R c IL These are the unit frequency regulation costs for traditional generating units, energy storage, distributed renewable energy, and interruptible loads, respectively; t lim This represents the frequency modulation speed; the smaller this value, the faster the frequency recovery, and the better the frequency modulation effect. Δf lim The maximum allowable steady-state frequency deviation of the power grid (corresponding to the minimum quasi-steady-state frequency), whose derivative is greater than 0, avoids t lim Δf is reached during the frequency drop phase. lim The timer stops when the time is up.

[0088] The constraints of the capacity optimization model include:

[0089]

[0090] In the formula, Nad max Maximum frequency deviation threshold; RoC max The threshold for the rate of change of frequency; and These are the upper and lower limits of the energy storage capacity, respectively. and These are the upper and lower limits of distributed renewable energy capacity, respectively. and These represent the upper and lower limits of the interruptible load capacity.

[0091] Traditional capacity allocation methods typically focus on a single objective (such as minimizing operating costs), which can easily lead to insufficient response speed or resource redundancy. This solution introduces a dual-objective function, quantifying both the total cost of frequency regulation and the system frequency change rate separately. This integrates economic efficiency and response performance into the optimization framework, enabling cost control and rapid adjustment capabilities while meeting frequency security requirements, thus improving overall operational efficiency. Furthermore, the objective function assigns unit frequency regulation cost coefficients to different resource types (traditional generating units, energy storage, distributed renewable energy, and interruptible loads), and performs weighted calculations based on their rated capacity, accurately reflecting the economic differences of various resources in frequency regulation services. This differentiated modeling approach makes the optimization process closer to market mechanisms, helping to guide rational resource allocation and improve resource allocation efficiency.

[0092] S4. Solve the capacity optimization model of S3 to obtain a Pareto solution set consisting of multiple non-dominant capacity schemes, and select a capacity scheme from the Pareto solution set as the final capacity scheme based on a preset robustness evaluation index.

[0093] In practice, the capacity optimization model is solved using a heuristic multi-objective particle swarm optimization algorithm. For example... Figure 2 As shown, the solution process includes:

[0094] 1) The multi-objective particle swarm optimization algorithm is used to initialize the population for encoding the capacity of each frequency regulation resource in the virtual power plant participating in primary frequency regulation ancillary services.

[0095] 2) Run the improved system frequency response model of the virtual power plant and determine whether Δf and dΔf / dt satisfy the constraints. If the constraints are not satisfied, proceed to step 3). If the constraints are satisfied, calculate the objective function frequency regulation cost f1 and frequency regulation speed f2.

[0096] 3) Determine whether the algorithm has converged / iterated completely. If so, present the Pareto solution set; otherwise, proceed to step 4).

[0097] 4) Update the particle's velocity and position, update each particle's individual historical best position and global best position, and return to step 2).

[0098] The Pareto solution set presented by the multi-objective particle swarm optimization algorithm includes multiple mutually non-substitutable optimal solutions (capacity schemes). This invention proposes a robustness criterion to select the most robust solution as a compromise solution.

[0099] The optimized capacity scheme for each frequency regulation resource in the virtual power plant to participate in primary frequency regulation ancillary services. It may change due to various uncertainties; assuming the capacity changes by [u] min [1], a 1×N can be obtained through random sampling. U The uncertainty magnitude matrix [u1,…,u] x ,…,u NU If ], then the actual capacity scheme under the x-th uncertain scenario is

[0100] Using the actual capacity scheme under the x-th uncertain scenario as input to the improved system frequency response model, the obtained Δf x and dΔf x The magnitude of the deviation of / dt from the original values ​​Δf and dΔf / dt represents the robustness of the optimal capacity scheme in the x-th uncertain scenario.

[0101] In specific implementation, the preset robustness evaluation index is determined in the following way:

[0102] For each capacity scheme in the Pareto solution set, multiple uncertainty scenarios are generated. The disturbance capacity schemes under each scenario are input into the improved system frequency response model to calculate the corresponding frequency deviation and frequency change rate.

[0103] The robustness evaluation index is the maximum comprehensive deviation R between the frequency response and the benchmark response under each uncertainty scenario.

[0104] The maximum overall deviation R is:

[0105]

[0106] In the formula, N U r represents the number of uncertain scenarios. x Let be the comprehensive deviation under the x-th uncertainty scenario; Δf and The frequency deviation and its rate of change under undisturbed conditions, Δf x and This is the corresponding value under the x-th uncertainty scenario;

[0107] The capacity scheme with the smallest R value in the Pareto solution set is selected as the final capacity scheme.

[0108] Traditional optimization methods typically solve for optimal solutions based on deterministic assumptions, neglecting real-world uncertainties such as fluctuations in energy storage SOC, renewable energy output, and load interruption response delays. This approach generates multiple uncertainty scenarios for each Pareto solution, simulating the impact of resource availability changes on system frequency response. This effectively identifies capacity schemes that perform stably under complex dynamic environments, significantly enhancing the anti-interference capability and operational reliability of the selected schemes. Furthermore, it achieves a decision-making upgrade from "static optimality" to "dynamic robustness." While schemes in the Pareto solution set are independent under ideal conditions, their actual performance can vary significantly due to external disturbances. This approach introduces a robustness evaluation mechanism, linking the theoretically optimal solution with actual operational risks. This avoids selecting "fragile solutions" that perform well only under ideal conditions but fail under disturbances, driving a shift in decision-making from pursuing the best single performance indicator to considering both stability and robustness.

[0109] Traditional methods often simplify virtual power plants to fixed capacity or ignore their dynamic characteristics, making it difficult to accurately reflect their actual regulation capabilities under frequency disturbances. This scheme constructs refined frequency response models for traditional generating units and power-type virtual power plants separately, and integrates the two to form an improved system frequency response model. This model can more accurately depict the dynamic behavior of power systems with a high proportion of distributed resources during primary frequency regulation, providing a reliable foundation for subsequent optimization. Furthermore, existing capacity allocation methods often focus on a single objective (such as minimizing cost), which can easily lead to insufficient response speed or resource redundancy. This scheme establishes a multi-objective capacity optimization model with "lowest frequency regulation cost" and "fastest frequency regulation speed" as dual objectives, effectively balancing the contradiction between economy and dynamic performance. While ensuring system frequency security, it avoids over-configuration or response lag. In addition, traditional virtual power plant participation in frequency regulation often relies on experience-based settings or static rules, lacking deep coupling with the overall system frequency dynamics. This method embeds the dynamic regulation capability of the virtual power plant into the system frequency response framework and optimizes the capacity based on this framework, making the virtual power plant a truly quantifiable, dispatchable, and reliable frequency regulation entity, thus providing technical support for its large-scale and commercial application in the ancillary services market.

[0110] This method enables the scientific, efficient, and robust configuration of virtual power plant frequency regulation capacity. While ensuring grid frequency stability, it also takes into account economic efficiency and response performance, significantly improving the overall effectiveness and market competitiveness of distributed resource aggregates participating in primary frequency regulation ancillary services.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services, characterized in that, Includes the following steps: S0. Construct a frequency response model for traditional generating units in the power grid; S1. Power-type regulating resources are used to form a power-type virtual power plant to participate in the primary frequency regulation of the power grid; a frequency response model of the power-type virtual power plant is constructed; wherein, the power-type regulating resources include energy storage, distributed new energy sources, and interruptible loads; S2. Combining the frequency response model of the traditional unit in S0 and the frequency response model of the power-type virtual power plant in S1, an improved system frequency response model considering the power-type virtual power plant is constructed. S3. Based on the improved system frequency response model constructed by S2, a capacity optimization model for virtual power plants participating in primary frequency regulation ancillary services is established with the objective functions of minimizing frequency regulation cost and maximizing frequency regulation speed. S4. Solve the capacity optimization model of S3 to obtain a Pareto solution set consisting of multiple non-dominant capacity schemes, and select a capacity scheme from the Pareto solution set as the final capacity scheme based on a preset robustness evaluation index.

2. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 1, characterized in that: In S1, energy storage and distributed new energy sources participate in primary frequency regulation using droop control; interruptible loads support the frequency by being temporarily disconnected or having their power reduced in the event of a frequency drop.

3. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 1, characterized in that: In S1, the frequency response model of the power-type virtual power plant includes: In the formula, Δf is the frequency deviation; s is the Laplace operator used for frequency domain modeling, representing the complex frequency variable; This represents the active power increment of the j-th energy storage unit; and These are the frequency regulation coefficient and time constant of the j-th energy storage unit, respectively; This represents the increase in active power of the k-th distributed renewable energy source. and These are the frequency regulation coefficient and time constant of the k-th distributed renewable energy source, respectively; The power reduction due to load l; Interruptible power set for load l; T l IL The interruption delay for load l.

4. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 3, characterized in that: In S0, the frequency response model of traditional generating units in the power grid is: In the formula, ΔP i G (s) represents the active power increment of the i-th conventional unit; and These are the frequency regulation coefficient and time constant of the i-th conventional unit, respectively; Let be the high-pressure cylinder power generation ratio of the i-th conventional unit.

5. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 4, characterized in that: In S2, the improved system frequency response model considering the power-type virtual power plant is constructed as follows: In the formula, ΔP represents the power deficit of the power grid; H sys D and T represent the equivalent inertia and damping of the power grid, respectively; IL K is the time constant for aggregating interruptible loads. G T G F G These are the frequency regulation coefficient, time constant, and high-pressure cylinder power generation ratio of the aggregated traditional unit, respectively; K E T E These are the frequency regulation coefficient and time constant of aggregated energy storage, respectively; K R T R These are the frequency regulation coefficient and time constant of aggregated distributed new energy sources, respectively; This refers to the rated capacity of traditional unit i; The rated capacity for energy storage j to participate in primary frequency regulation auxiliary services; The rated capacity for distributed renewable energy k to participate in primary frequency regulation auxiliary services; The rated capacity for interruptible load l to participate in primary frequency regulation ancillary services; N G N E N R N IL These represent the number of traditional generating units participating in primary frequency regulation of the power grid, the amount of energy storage, the amount of distributed renewable energy, and the amount of interruptible loads; H i Let be the inertial constant of the i-th conventional unit.

6. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 5, characterized in that: In S3, the objective function of the capacity optimization model is: minF(f1,f2); In the formula, C represents the frequency modulation cost; C VPP and C G These are the frequency regulation costs of aggregated resources and traditional generating units, respectively; c G c E c R c IL These are the unit frequency regulation costs for traditional generating units, energy storage, distributed renewable energy, and interruptible loads, respectively; t lim For frequency modulation speed; Δf lim This represents the maximum allowable steady-state frequency deviation of the power grid.

7. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 6, characterized in that: In S3, the constraints of the capacity optimization model include: In the formula, Nad max Maximum frequency deviation threshold; RoC max The threshold for the rate of change of frequency; and These are the upper and lower limits of the energy storage capacity, respectively. and These are the upper and lower limits of distributed renewable energy capacity, respectively. and These represent the upper and lower limits of the interruptible load capacity.

8. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 1, characterized in that: In S4, the preset robustness evaluation index is determined in the following way: For each capacity scheme in the Pareto solution set, multiple uncertainty scenarios are generated. The disturbance capacity schemes under each scenario are input into the improved system frequency response model to calculate the corresponding frequency deviation and frequency change rate. The robustness evaluation index is the maximum comprehensive deviation R between the frequency response and the benchmark response under each uncertainty scenario.

9. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 8, characterized in that: The maximum overall deviation R is: In the formula, N U r represents the number of uncertain scenarios. x Let be the comprehensive deviation under the x-th uncertainty scenario; Δf and The frequency deviation and its rate of change under undisturbed conditions, Δf x and This is the corresponding value under the x-th uncertainty scenario; The capacity scheme with the smallest R value in the Pareto solution set is selected as the final capacity scheme.

10. The capacity optimization method for virtual power plants participating in grid primary frequency regulation ancillary services as described in claim 1, characterized in that: In S4, the capacity optimization model is solved using the heuristic multi-objective particle swarm optimization algorithm.

Images