New energy station primary frequency modulation capability evaluation method considering power grid structure and operation characteristics

Abstract

The invention discloses a new energy station primary frequency modulation capability evaluation method considering a power grid structure and operation characteristics. The method comprises the following steps: collecting data; the data comprises complex network data of a power system, historical operation data of the system, virtual inertia control parameters of a wind generating set, droop control parameters of a photovoltaic generating set and primary frequency modulation control performance data; quantifying the primary frequency modulation comprehensive capability of the new energy station according to the collected data; and training the new energy station primary frequency modulation capability evaluation model and obtaining the new energy station primary frequency modulation capability evaluation model with the highest prediction accuracy. According to the method, the new energy station primary frequency modulation capability evaluation model is established based on the XGBoost algorithm, compared with a traditional time domain simulation method, when the method is applied to different new energy stations, modeling does not need to be conducted again, and the method has obvious advantages in the aspects of universality and calculation efficiency.

Description

Technical Field

[0001] This application relates to the field of power system primary frequency regulation assessment technology, and in particular to a method for assessing the primary frequency regulation capability of new energy power plants that takes into account the grid structure and operating characteristics. Background Technology

[0002] Building a high-proportion renewable energy power system is the only way to achieve dual-carbon goals. The large-scale integration of renewable energy sources has led to low inertia and weak grid support characteristics in the power system, posing new challenges to primary frequency regulation. Quantitatively assessing the primary frequency regulation capability of renewable energy power plants is the foundation and prerequisite for guiding optimized frequency regulation control of these plants, analyzing the frequency response characteristics of new power systems, planning frequency regulation resources for power plants, and achieving interaction with the grid.

[0003] Currently, the primary frequency regulation capability assessment of new energy power systems mostly adopts the time-domain simulation method. By building a refined simulation model, the frequency regulation capability of the power station is simulated and analyzed. However, the system model is highly specific and lacks universality, making it difficult to meet the needs of the rapid development and integration of new energy. Summary of the Invention

[0004] In view of the technical problems mentioned in the background, the purpose of this invention is to provide a method for evaluating the primary frequency regulation capability of new energy power plants that takes into account the power grid structure and operating characteristics.

[0005] To achieve the objectives of this invention, the technical solution provided by this invention is as follows:

[0006] This application provides a method for evaluating the primary frequency regulation capability of new energy power plants that considers power grid structure and operating characteristics, including the following steps:

[0007] Step S1: Collect data;

[0008] The data includes complex power system network data, historical system operation data, virtual inertia control parameters of wind turbine generators, droop control parameters of photovoltaic generators, and primary frequency regulation control performance data; the primary frequency regulation control performance data includes: the maximum rate of change of frequency (ROCOF) after primary frequency regulation. max The maximum frequency deviation Δf after a single frequency modulation operation nadir,1 Steady-state frequency deviation Δf after one frequency modulation operation ss And the frequency deviation product IFD after one frequency modulation action;

[0009] Step S2: Based on the maximum frequency change rate ROCOF after one frequency modulation operation max The maximum frequency deviation Δf after a single frequency modulation operation nadir,1 Steady-state frequency deviation Δf after one frequency modulation operation ssAnd the frequency deviation product IFD after a single frequency regulation action, which quantifies the comprehensive primary frequency regulation capability of new energy power plants;

[0010] Step S3: Train the primary frequency regulation capability assessment model for new energy power plants and obtain the primary frequency regulation capability assessment model for new energy power plants with the highest prediction accuracy.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0012] (1) The present invention establishes a primary frequency regulation capability evaluation model for new energy power stations based on the XGBoost algorithm. Compared with the traditional time-domain simulation method, it does not require remodeling when applied to different new energy power stations, and has obvious advantages in terms of universality and computational efficiency.

[0013] (2) The input data of the XGBoost algorithm of this invention is "structural data + operational data". Compared with the traditional evaluation method based on "operational data", by analyzing the impact of "structural data" on frequency regulation capability, it can guide new energy power stations to make a prediction of frequency regulation capability at the planning level, which has stronger engineering practice significance. Attached Figure Description

[0014] Figure 1 A schematic diagram of the method flow for evaluating the primary frequency regulation capability of new energy power plants, taking into account the power grid structure and operating characteristics, provided in an embodiment of the present invention;

[0015] Figure 2 This is the input data type for the XGBoost algorithm in this embodiment of the invention.

[0016] Figure 3 This is a diagram illustrating the XGBoost algorithm framework in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] like Figure 1 As shown, this embodiment of the invention provides a method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics, including the following steps:

[0019] Step S1: Collect data;

[0020] Among them, such as Figure 2As shown, the data includes power system complex network data, system historical operation data, wind turbine virtual inertia control parameters, photovoltaic generator droop control parameters, and primary frequency regulation control performance data; the primary frequency regulation control performance data includes: the maximum rate of change of frequency (ROCOF) after primary frequency regulation action. max The maximum frequency deviation Δf after a single frequency modulation operation nadir,1 Steady-state frequency deviation Δf after one frequency modulation operation ss And the frequency deviation product IFD after one frequency modulation action;

[0021] The formula for the frequency deviation product IFD after the first frequency modulation operation is as follows:

[0022]

[0023] In the formula, T1 is the start time of the first frequency modulation, T2 is the end time of the first frequency modulation, and Δf(t) is the frequency change function.

[0024] The complex network data of the power system includes: the eigenvalue R of the largest adjacency matrix of the power network. ρ The average length R of the characteristic path from the new energy power station node to the load node L , New energy power station nodality value R k Node betweenness R of new energy power stations B Network line load rate R D ;

[0025] The system's historical operating data includes: photovoltaic power output data P on the statistical day. p Wind power output data P on the statistical day W Load distribution P on the statistical day L ;

[0026] The virtual inertia control parameters of wind turbine generator sets include: inertial response time constant H, and wind turbine generator set rotational inertia J;

[0027] The droop control parameters for photovoltaic power generation units include: the droop coefficient K of the photovoltaic power station. PV .

[0028] Among them, the average length R of the characteristic path from the new energy power station node to the load node L The calculation formula is:

[0029]

[0030] In the formula, N L d is the number of load nodes in the network, and d is the number of edges traversed by the shortest path from the new energy power station to all load nodes.

[0031] The formula for the inertial response time constant H is as follows:

[0032]

[0033] In the formula, w s S is the rated speed, and S is the apparent power.

[0034] This step also includes preprocessing the collected historical system operation data:

[0035] The photovoltaic power output data P collected on statistical days was analyzed using a Gaussian mixture model clustering algorithm. p Wind power output data P on the statistical day W Load distribution P on the statistical day L Outlier filtering and removal were performed on the three types of data, and missing data were supplemented based on mean imputation.

[0036] The outlier screening and removal based on the Gaussian mixture model clustering algorithm includes the following steps:

[0037] (1) Establish the dataset x of samples to be clustered q (q=1…N d ), x q =[P p P W P L ], where N d This represents the total number of data sets.

[0038] (2) Initialize the mean μ of f (10≤f≤15) Gaussian mixture distributions. j Covariance ε j Calculate x q The probability γ of belonging to the j-th (j=1…f) mixture Gaussian distribution q ,j;

[0039] in,

[0040] (3) Update the mean and covariance of the above f mixture Gaussian distributions;

[0041]

[0042] (4) If the Gaussian parameter converges, i.e. μ j -μ' j <ζ,ε j -ε j If '<ζ, then clustering is complete; otherwise, let μ j ←μ' j , ε j ←ε j ', return (2);

[0043] (5) Based on the Gaussian parameters obtained from the solution, x q (q=1…N d )Regression to probability γ q,j The largest class forms f clusters {ψ1,ψ2,…,ψ f};If ψ j The number of data sets in (j=1…f) ≤ 5% × N d , then ψ j The data in the data is abnormal and needs to be removed.

[0044] Step S2: Based on the maximum frequency change rate ROCOF after one frequency modulation operation max The maximum frequency deviation Δf after a single frequency modulation operation nadir,1 Steady-state frequency deviation Δf after one frequency modulation operation ss And the frequency deviation product IFD after a single frequency regulation action, which quantifies the comprehensive primary frequency regulation capability of new energy power plants;

[0045] The quantitative indicator of primary frequency regulation capability of new energy sources is defined as follows:

[0046]

[0047] In the formula, α, β, σ, and η are the weights of four primary frequency regulation capability indicators, and y i It is the quantized value of the primary frequency modulation integrated capability corresponding to the i-th set of data.

[0048] Step S3: Train the primary frequency regulation capability assessment model for new energy power plants and obtain the primary frequency regulation capability assessment model for new energy power plants with the highest prediction accuracy.

[0049] Step S3 specifically includes the following:

[0050] Step S31: In this invention, the XGBoost algorithm flowchart is as follows: Figure 3 As shown, a sample dataset D = (x i ,y i (|D|=N) d x i =(N L,i ,…,K PV,i Divide the dataset into training and testing datasets;

[0051] Step S32: Establish the objective function minL of the primary frequency regulation capability assessment model for new energy power plants based on the XGBoost algorithm; wherein, the objective function minL of the primary frequency regulation capability assessment model for new energy power plants is defined as:

[0052]

[0053] in:

[0054]

[0055] In the formula, These are predicted values, where K is the number of branch trees in the XGBoost algorithm, and f k (x i Let Ω(f) be the leaf node score of the k-th tree, F be the mapping space of the regression tree, and Ω(f) be the value of the leaf node of the k-th tree. k ) is the model complexity penalty function, γ and λ are the penalty coefficient and regularization coefficient, respectively, and T and ω are the number of leaves and leaf weights of the k-th tree, respectively;

[0056] Step S33: Minimize the objective function L based on incremental training; L is rewritten as:

[0057]

[0058] Step S34: In f k Perform a second-order Taylor expansion at the point = 0 to approximate the objective function L; the objective function can be approximated as:

[0059]

[0060] In the formula, g i with h i Represent the loss function respectively The first and second derivatives of the current model, where j is the leaf node number of the k-th tree, ω j Let I be the weight of the j-th leaf node. j Let j be the sample set of the j-th leaf node;

[0061] Step S35: Calculate L (k) For ω j Take the derivative of the derivative and set it equal to 0 to calculate the optimal solution. The training set is input into the primary frequency regulation capability assessment model of new energy power plants. By adjusting the maximum number of decision trees and the maximum tree depth, the primary frequency regulation capability assessment model of new energy power plants with the highest prediction accuracy is obtained.

[0062] Among them, the optimal solution Represented as:

[0063] Finally, it should be noted that the above embodiments are merely illustrative and explanatory of the present invention, and are not intended to limit the present invention to the scope of the described embodiments. Furthermore, those skilled in the art will understand that the present invention is not limited to the above embodiments, and many more variations and modifications can be made based on the teachings of the present invention, all of which fall within the scope of protection claimed by the present invention.

Claims

1. A method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operational characteristics, characterized in that, Includes the following steps: Step S1: Collect data; The data includes complex power system network data, historical system operation data, virtual inertia control parameters of wind turbine generators, droop control parameters of photovoltaic generators, and primary frequency regulation control performance data; the primary frequency regulation control performance data includes: the maximum rate of change of frequency (ROCOF) after primary frequency regulation. max The maximum frequency deviation Δf after a single frequency modulation operation nadir,1 Steady-state frequency deviation Δf after one frequency modulation operation ss And the frequency deviation product IFD after one frequency modulation action; Step S2: Based on the maximum frequency change rate ROCOF after one frequency modulation operation max The maximum frequency deviation Δf after a single frequency modulation operation nadir,1 Steady-state frequency deviation Δf after one frequency modulation operation ss And the frequency deviation product IFD after a single frequency regulation action, which quantifies the comprehensive primary frequency regulation capability of new energy power plants; Step S3: Train the primary frequency regulation capability assessment model for new energy power plants and obtain the primary frequency regulation capability assessment model for new energy power plants with the highest prediction accuracy.

2. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics according to claim 1, characterized in that, Step S3 specifically includes the following: Step S31: Establish the sample dataset D = (x i ,y i (|D|=N) d x i =(N L,i ,…,K PV,i Divide the dataset into training and testing datasets; Step S32: Establish the objective function minL of the primary frequency regulation capability evaluation model for new energy power stations based on the XGBoost algorithm; Step S33: Minimize the objective function L based on incremental training; Step S34: In f k Perform a second-order Taylor expansion at the point = 0 to approximate the objective function L; Step S35: Calculate L (k) For ω j Take the derivative of ω and set it equal to 0, then calculate the optimal solution ω. j * The training set is input into the primary frequency regulation capability assessment model of new energy power plants. By adjusting the maximum number of decision trees and the maximum tree depth, the assessment model of primary frequency regulation capability of new energy power plants with the highest prediction accuracy is obtained.

3. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics according to claim 2, characterized in that, In step S32, the objective function minL of the primary frequency regulation capability assessment model for new energy power plants is defined as: in: In the formula, These are predicted values, where K is the number of branch trees in the XGBoost algorithm, and f k (x i Let Ω(f) be the leaf node score of the k-th tree, F be the mapping space of the regression tree, and Ω(f) be the value of the leaf node of the k-th tree. k ) is the model complexity penalty function, γ and λ are the penalty coefficient and regularization coefficient, respectively, and T and ω are the number of leaves and leaf weights of the k-th tree, respectively; In step S33, the objective function L is minimized based on incremental training, and L is rewritten as: In step S34, the objective function can be approximated as: In the formula, g i with h i Represent the loss function respectively The first and second derivatives of the current model, where j is the leaf node number of the k-th tree, ω j Let I be the weight of the j-th leaf node. j Let j be the sample set of the j-th leaf node; In step S35, the optimal solution Represented as:

4. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics according to any one of claims 1-3, characterized in that, In step S1, the complex network data of the power system includes: the eigenvalue R with the largest adjacency matrix of the power network. ρ The average length R of the characteristic path from the new energy power station node to the load node L , New energy power station nodality value R k Node betweenness R of new energy power stations B Network line load rate R D ; The system's historical operating data includes: photovoltaic power output data P on the statistical day. p Wind power output data P on the statistical day W Load distribution P on the statistical day L ; The virtual inertia control parameters of wind turbine generator sets include: inertial response time constant H, and wind turbine generator set rotational inertia J; The droop control parameters for photovoltaic power generation units include: the droop coefficient K of the photovoltaic power station. PV .

5. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics according to claim 4, characterized in that, The average length R of the characteristic path from the new energy power station node to the load node L The calculation formula is: In the formula, N L d is the number of load nodes in the network, and d is the number of edges traversed by the shortest path from the new energy power station to all load nodes.

6. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics according to claim 4, characterized in that, The formula for the inertial response time constant H is as follows: In the formula, w s S is the rated speed, and S is the apparent power.

7. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics according to claim 1, characterized in that, The formula for the frequency deviation product IFD after the first frequency modulation operation is as follows: In the formula, T1 is the start time of the first frequency modulation, T2 is the end time of the first frequency modulation, and Δf(t) is the frequency change function.

8. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics according to claim 1, characterized in that, Step S1 includes a step of preprocessing the collected historical system operation data: The photovoltaic power output data P collected on statistical days was analyzed using a Gaussian mixture model clustering algorithm. p Wind power output data P on the statistical day W Load distribution P on the statistical day L Outlier filtering and removal were performed on the three types of data, and missing data were supplemented based on mean imputation.

9. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operation characteristics according to claim 8, characterized in that, Outlier screening and removal based on Gaussian mixture model clustering algorithm includes the following steps: (1) Establish the dataset x of samples to be clustered q (q=1…N d ), x q =[P p P W P L ], where N d This represents the total number of data sets. (2) Initialize the mean μ of f (10≤f≤15) Gaussian mixture distributions. j Covariance ε j Calculate x q The probability γ of belonging to the j-th (j=1…f) mixture Gaussian distribution q,j ; in, (3) Update the mean and covariance of the above f mixture Gaussian distributions; (4) If the Gaussian parameter converges, i.e. μ j -μ' j <ζ,ε j -ε j If '<ζ, then clustering is complete; otherwise, let μ j ←μ' j , ε j ←ε j ', return (2); (5) Based on the Gaussian parameters obtained from the solution, x q (q=1…N d )Regression to probability γ q,j The largest class forms f clusters {ψ1,ψ2,…,ψ f };If ψ j The number of data sets in (j=1…f) ≤ 5% × N d , then ψ j The data in the data is abnormal and needs to be removed.

10. The method for evaluating the primary frequency regulation capability of new energy power plants considering power grid structure and operating characteristics according to claim 1, characterized in that, In step S2, the quantitative index of the primary frequency regulation capability of new energy sources is defined as follows: In the formula, α, β, σ, and η are the weights of four primary frequency regulation capability indicators, and y i It is the quantized value of the primary frequency modulation integrated capability corresponding to the i-th set of data.

Images