Method for identifying reactive power support control parameters of grid-following type current converter
By dividing voltage operating conditions, reactive current clustering and genetic particle swarm algorithms for mesh-type inverters, the problem of low accuracy and efficiency of reactive support control parameters identification in the prior art is solved, and efficient and accurate multi-dimensional parameter identification is achieved.
Patent Information
- Application Number
- CN202510359209.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-25
- Publication Date
- 2025-06-24
AI Technical Summary
The existing reactive support control parameter identification method of mesh-type inverter is not accurate and efficient, and cannot meet the identification requirements of multi-dimensional input and multi-parameters.
By dividing multiple voltage conditions, the inverter operation test is carried out, the reactive current is clustered, the reactive current reaches the current limit amplitude is identified, the prediction error is corrected, and the genetic particle swarm algorithm is used to find the optimal solution to the control parameters.
It improves the accuracy and efficiency of reactive support control parameters identification, can identify multi-dimensional control parameters, reduces the error rate of manual judgment, and enhances the accuracy and convergence of iteration.
Smart Images
Figure CN120200328A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the technical field of new energy power generation, and more specifically, relates to a method for identifying reactive power support control parameters of a grid-connected converter. Background Art
[0002] With the continuous expansion of the scale of new energy connected to the power grid, the safe and stable operation of the power system faces many challenges. The growth of new energy has a more significant impact on the power grid. To address these challenges, it is particularly important to build electromechanical transient and electromagnetic transient models that can accurately reflect the characteristics of new energy power generation. Most existing new energy sources are connected to the grid using grid-connected converters. According to national standards, new energy equipment needs to have the ability of reactive power support under multiple voltage conditions. Since the requirement for reactive power support is relatively high when the voltage drops and rises significantly, in order to meet the national standard requirements, an independent reactive power support strategy applicable to multiple voltage conditions is usually adopted. However, new energy manufacturers usually provide a whole-machine black-box model, and it is impossible to directly obtain its relevant control parameters and control strategies. If the modeling is completely carried out according to the typical fixed reactive power control strategy, a large error will be introduced into the model, affecting the correctness of the simulation analysis conclusion. Therefore, it is necessary to identify the reactive power support strategy based on measured data.
[0003] Currently, the existing methods for identifying reactive power support control parameters of grid-connected converters rely relatively heavily on manual identification, which is prone to errors and has low efficiency; the existing methods that do not require manual participation in parameter identification can identify parameters in a relatively single dimension and cannot meet the identification requirements of multi-dimensional input and multi-parameters. Summary of the Invention
[0004] In view of the above deficiencies or improvement requirements of the prior art, this application provides a method for identifying reactive power support control parameters of a grid-connected converter, aiming to solve the technical problem of low accuracy and efficiency of the existing methods for identifying reactive power support control parameters.
[0005] To achieve the above objective, this application provides a method for identifying reactive power support control parameters of a grid-connected converter, including: Dividing multiple voltage conditions for the operation test of the converter, and grouping the obtained test data according to the voltage conditions; Performing density-based clustering on the reactive current in the test data, comparing the clustering and grouping results, and finding the reactive current that reaches the current limiting amplitude A or the reactive current that may reach the current limiting amplitude B ; If there is reactive current A , then delete the reactive current A from the test data, and use the maximum difference between the predicted value and the actual value of the reactive power support model as the prediction error; if there is reactive current B, then based on the reactive current B Define a clipping risk factor and use the clipping risk factor to correct the prediction error; Regard the parameters to be identified in the reactive power support model as particles, take minimizing the prediction error as the optimization goal, and use the genetic particle swarm optimization algorithm to find the optimal solution of the control parameters.
[0006] Preferably, divide multiple voltage conditions for the operation test of the converter, and group the obtained test data according to the voltage conditions, specifically: Conduct multiple groups of tests for each voltage condition. In each group of tests, make the initial reactive power zero and measure the initial reactive current; when the voltage drops or rises to the set voltage condition, measure the reactive current and the port voltage; Record all test data and group the obtained test data according to the voltage conditions.
[0007] Preferably, perform density-based clustering on the reactive current in the test data, compare the clustering and grouping results, and find the reactive current that reaches the current limiting amplitude A or the reactive current that is at risk of reaching the current limiting amplitude B ; specifically: If the clustering parameters can be adjusted so that the clustering results and the grouping results are all in agreement, then set the reactive current in the voltage condition group at the farthest end from the rated voltage as the reactive current at risk of reaching the current limiting amplitude B ; If by adjusting the clustering parameters, only the clustering results between the rated voltage and the threshold voltage can be made to match the grouping results, and all the reactive currents outside the threshold voltage are clustered into the same category and cannot match the grouping results, then set the reactive current outside the threshold voltage as the reactive current that reaches the current limiting amplitude A .
[0008] Preferably, take the maximum difference between the predicted value and the actual value of the reactive power support model as the prediction error. The prediction error is specifically:
[0009] where is the th reactive current predicted by the reactive power support model; is the th reactive current in the test data; is the total amount of reactive current in the test data; is to find the maximum value; .
[0010] Preferably, based on the reactive current B Define a clipping risk factor and use the clipping risk factor to correct the prediction error, specifically: According to the reactive currentB Judge the reactive current based on the relationship between the mean value of B and the magnitude of 1. Determine whether there is a reactive current reaching the current limiting amplitude in B . If so, multiply the difference between the mean value of the reactive current and 1 by a preset current limiting risk assessment sensitivity to obtain a current limiting risk factor, and use the current limiting risk factor to correct the prediction error of the reactive current B part.
[0011] Preferably, the corrected prediction error is:
[0012] where ; The th reactive current predicted by the reactive power support model; is the th reactive current in the test data; is the reactive current B The serial number range of the reactive current within; is to find the maximum value; where ; is the current limiting risk factor:
[0013] is the mean value of the reactive current B ; is the preset current limiting risk assessment sensitivity.
[0014] Preferably, with minimizing the prediction error as the optimization goal, a genetic particle swarm optimization algorithm is used to find the optimal solution of the control parameters. Specifically: set the prediction error as the fitness of the genetic particle swarm optimization algorithm. For each particle, the position with the minimum fitness is the individual optimal solution; for all particles, the particle with the minimum fitness is the global optimal solution. Use the genetic particle swarm optimization algorithm to perform the optimization iteration of the particles until the iteration stop condition is reached, and then output the global optimal particle as the optimal solution of the control parameters.
[0015] Preferably, the initial value of each particle in the particle swarm is defined by the method of initial value mutation.
[0016] Preferably, the initial value of the particle is specifically:
[0017]
[0018] where is the preset initial value reference; and is the preset upper and lower limits of the initial value; is a Gaussian distribution random number with a mean of zero and a standard deviation of 0.3.
[0019] Preferably, the reactive power support model is bounded by the rated voltage and is specifically divided into a low-voltage reactive power support model and a high-voltage reactive power support model. Among them, the low-voltage reactive power support model is:
[0020] The high-voltage reactive power support model is:
[0021] Among them, is the reactive current, is the terminal voltage of the machine, is the initial reactive current; is the low-voltage strategy voltage reference value, is the reactive power compensation coefficient, is the initial reactive current influence coefficient, is the reactive current set value; The parameters to be identified are: 、 、 and .
[0022] Generally speaking, compared with the prior art through the above technical solutions conceived by the present application, the following beneficial effects are obtained: (1) Compared with the traditional manual parameter identification, the method of the present application regards the parameters to be identified as particles and uses the genetic particle swarm algorithm to find the optimal solution of the control parameters, avoiding the link of human judgment, reducing the error rate, and greatly improving the efficiency. By adding the idea of genetics to the particle swarm, the optimization ability of the method is enhanced, which can better avoid falling into the local optimum of parameter identification and enhance the iteration accuracy and convergence. In addition, the dimension of the particle can be set to represent various dimensions of the parameters to be identified. Therefore, compared with the existing methods, the method of the present application does not limit the number of dimensions of the identified parameters and can identify various dimensions of control parameters.
[0023] (2) The method of the present application considers the influence of the converter reaching the current limiting amplitude. By clustering the test data, two scenarios of reaching the current limiting amplitude are intelligently identified, and the optimization target of the genetic particle swarm algorithm can be intelligently adjusted according to different scenarios where the current limiting amplitude appears. Therefore, the control parameter identification method of the present application has higher accuracy and rationality.
[0024] (3) The reactive power support model disclosed by the parameter identification method of the present application has a clearer physical meaning. By identifying specific parameters around this model and adding strong constraints to the identification strategy, the identification is more well-founded and more persuasive than the pure data model.
[0025] (4) This application uses the method of initial value mutation to regulate the initial distribution of particles, making it more adaptable to the characteristics of the reactive power support control parameters of the converter, and by enhancing the diversity of the initial values, avoiding the local optimal trap, and finally improving the global convergence and robustness of the genetic particle swarm optimization algorithm. Description of the Drawings
[0026] Figure 1 It is the flow chart of the identification method for the reactive power support control parameters of the network-connected converter provided by the embodiment of this application.
[0027] Figure 2 It is the simulation data curve when the voltage sag is 20% provided by the embodiment of this application.
[0028] Figure 3 It is the simulation data curve when the voltage sag is 40% provided by the embodiment of this application.
[0029] Figure 4 It is the simulation data curve when the voltage sag is 60% provided by the embodiment of this application.
[0030] Figure 5 It is the simulation data curve when the voltage sag is 70% provided by the embodiment of this application. Detailed Embodiment
[0031] In order to make the purpose, technical solution and advantages of this application more clear, the following further details this application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.
[0032] The terms "first" and "second" etc. in the specification and claims of this application are used to distinguish different objects, rather than to describe a specific order of the objects. For example, the first voltage condition and the second voltage condition etc. are used to distinguish different voltage conditions, rather than to describe the specific order of the voltage conditions.
[0033] In the embodiments of this application, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, using words such as "exemplary" or "for example" aims to present relevant concepts in a specific way.
[0034] In the description of the embodiments of this application, unless otherwise specified, the meaning of "a plurality of" refers to two or more. For example, a plurality of voltage conditions refers to two or more voltage conditions, etc.; a plurality of groups refers to two or more groups, etc.
[0035] The embodiments of the present application will be described below with reference to the accompanying drawings in the embodiments of the present application. The embodiments of the present application are for identifying the control parameters of low-voltage reactive power support for voltages below the rated voltage, and the same process method is also used for identifying the control parameters of high-voltage reactive power support.
[0036] As Figure 1 shown, the method for identifying the reactive power support control parameters of the grid-connected converter disclosed in the embodiments of the present application includes the following steps: Step 1: Collect and group data on the low-voltage strategy of the converter: Pre-divide multiple voltage conditions for on-site measurement of the equipment, such as voltage conditions with voltage drops of 20%, 40%, 60%, and 70%. In other embodiments, other multiple voltage conditions can also be selected for on-site measurement of the equipment.
[0037] For each voltage condition, multiple groups of tests are carried out. In each group of tests, the initial reactive power is set to zero, which can float slightly up and down, and the initial reactive current at this time is measured Iq0 . When the voltage drops or rises to the set voltage condition, the reactive current Iq and the port voltage Vt are measured.
[0038] All the data obtained from the tests are grouped according to the voltage conditions, and thus the test data of the converter under each voltage condition are obtained.
[0039] Next, the reactive power support control parameters of the converter are identified based on the test data.
[0040] Step 2: Identify the reactive current that is at risk of reaching the current limiting amplitude through clustering: The DBSCAN clustering method is used to classify the reactive current in the test data. The DBSCAN clustering method is a density-based clustering method. In other embodiments, other density-based clustering methods can also be used.
[0041] The specific clustering method is as follows: Select a neighborhood radius and a minimum number of samples, calculate the number of neighbor sample points within the neighborhood radius for each sample point. If the number of neighbor sample points is greater than the minimum number of samples, then the sample point is a valid sample, and it forms a category with the neighbor sample points within the neighborhood radius. Similar processing is performed on all sample points, thereby obtaining the DBSCAN clustering result.
[0042] Adjust the appropriate neighborhood radius and minimum number of samples. If it is possible to make the clustering result match all the grouping results in Step 1 by adjusting the clustering parameters, then the reactive current in the voltage condition group at the farthest end from the rated voltage is set as the reactive current that is at risk of reaching the current limiting amplitude B , and proceed to Step 6; otherwise, proceed to Step 3.
[0043] In this embodiment, the neighborhood radius is selected as 0.1 and the minimum number of samples is 3. The DBSCAN clustering result coincides with the grouping result obtained by grouping according to voltage conditions in step 1.
[0044] Step 3: Identify the reactive current reaching the current limiting amplitude through clustering: If it is found that the clustering result and the grouping result cannot completely coincide, further adjust the neighborhood radius and the minimum number of samples until the clustering result between the rated voltage and the threshold voltage coincides with the grouping result, and all the reactive currents outside the threshold voltage are clustered into the same category and cannot coincide with the grouping result. Set the reactive current outside the threshold voltage as the reactive current reaching the current limiting amplitude. A .
[0045] Here, the threshold voltage is a voltage value obtained through clustering. The clustering result of the reactive current between this voltage value and the rated voltage coincides with the grouping result, and all the reactive currents outside this voltage value are clustered into one category and cannot coincide with the grouping result.
[0046] Step 4: Based on the reactive current reaching the current limiting amplitude A Calculate the current limiting amplitude:
[0047] Among them, is the current limiting amplitude, is the reactive current A the th reactive current value in is the reactive current A the total number of reactive current values in
[0048] Step 5: Eliminate the reactive current A from the original test data: After the converter reaches the limit, the converter outputs the reactive current reaching the current limiting amplitude A , and the reactive current A will affect the identification of model parameters, so it is eliminated from the test data.
[0049] This embodiment considers the influence of the converter reaching the current limiting amplitude, and intelligently identifies two scenarios of reaching the current limiting amplitude through the method of clustering the test data in steps 2 - 5. Eliminate the reactive current A that will definitely affect the identification of model parameters from the test data to avoid invalid data.
[0050] In subsequent steps, the optimization objective of the genetic particle swarm algorithm will be intelligently adjusted according to different scenarios where the current limiting amplitude appears, making the identification result more reasonable and accurate.
[0051] Step 6: Determine the reactive power support control model and control parameters of the converter: Based on the requirements for reactive current output in the national standard, determine the low-voltage reactive power support control model and the high-voltage reactive power support control model, and determine the parameters to be identified.
[0052] The low-voltage reactive power support model is:
[0053] The high-voltage reactive power support model is:
[0054] Among them, is the reactive current, output as the dependent variable; is the terminal voltage, is the initial reactive current, and are the input variables, and the corresponding values are directly input into the model from the test data; is the voltage reference value, is the reactive power compensation coefficient, is the influence coefficient of the initial reactive current, is the reactive current set value; the parameters to be identified are: 、 、 and 。
[0055] The reactive power support model disclosed by the parameter identification method in this embodiment is based on the national standard requirements, has a clearer physical meaning, and specific parameters are identified around this model. Strong constraints are added to the identification strategy to make the identification more well-founded and more persuasive than the pure data model.
[0056] Step 7: Initialization of parameter identification: In the embodiment of this application, the genetic particle swarm algorithm is used to iterate the parameters. The number of selected particles is N, the maximum number of iterations is T generations, and each particle has 4-dimensional variables, corresponding to the four parameters to be identified in Step 6 、 、 and 。Set the initial iteration reference values of the four parameters to 、 、 、 ,The upper and lower limits of the four parameters are respectively 、 、 、 、 、 、 、 , the initial values of the parameters of each particle are defined by the method of initial value mutation, where As the first parameter of the th particle The initial value is shown as follows:
[0057]
[0058] where is a random number of a Gaussian distribution with a mean of zero and a standard deviation of 0.3. For the second to fourth parameters, namely , and the initial values of the particle iteration are also set in a similar manner.
[0059] In this embodiment, 4D particles are defined to represent 4 types of model parameters. In this application, the dimension of the particle can be set to represent the parameters to be identified in multiple dimensions. Therefore, compared with the existing method, the method of this application does not limit the number of dimensions of the identified parameters and can identify control parameters in multiple dimensions.
[0060] This application uses the method of initial value mutation to define the initial values of the parameters of each particle, making its initial values more adaptable to the characteristics of the converter reactive power support control parameters, and by increasing the diversity of the initial values, avoiding the local optimal trap, and finally improving the global convergence and robustness of the genetic particle swarm algorithm.
[0061] Step 8: Determine different fitness according to different scenarios where the current limiting amplitude appears: If there is a reactive current A reaching the current limiting amplitude, then the maximum difference between the predicted value and the actual value of the reactive power support model is taken as the prediction error:
[0062] where is the th reactive current predicted by the reactive power support model; is the th reactive current in the test data; is the total amount of reactive current in the test data; is to find the maximum value; . The prediction error is used as the fitness function in the genetic particle swarm algorithm:
[0063] If there is a reactive current B at risk of reaching the current limiting amplitude, then based on the reactive current BDefine the clipping risk factor, correct the prediction error with the clipping risk factor, and use the prediction error as the fitness function in the genetic particle swarm optimization algorithm, specifically as follows: Calculate the reactive current B The central value, and the calculation formula is as follows:
[0064] Wherein, Is the central output value of the reactive current B ; Is the number of samples in the reactive current B ; Is the reactive current B In the Th reactive current data
[0065] Define the clipping risk factor :
[0066] Wherein, Is the preset clipping risk assessment sensitivity
[0067] The corrected prediction error is:
[0068] Wherein, ; ; Use the prediction error as the fitness function in the genetic particle swarm optimization algorithm:
[0069] In this step, the optimization objective of the genetic particle swarm optimization algorithm is intelligently adjusted according to different scenarios where the current limiting amplitude appears, so that the identification result is more reasonable and accurate
[0070] Step 9: Determine the iteration direction of the particle swarm: Update the particle positions; for the particle swarm, select the particle with the minimum fitness as the global optimal solution, and for each particle, select the position with the minimum fitness of itself as the individual optimal solution, and carry out the iteration of the particle swarm algorithm
[0071] Step 10: Incorporate genetic ideas during the particle swarm iteration: Select half of the particles as the parent generation, select the crossover probability P1 and the mutation probability P2, perform genetic algorithm crossover and mutation on the parent generation particles to generate offspring particles. According to the definition of fitness in step 8, compare the fitness of the offspring particles and the parent generation particles. If the offspring particles have a smaller fitness than the parent generation particles, replace the corresponding parent generation particles
[0072] This application incorporates the idea of genetic algorithm into the ordinary particle swarm parameter identification algorithm, enhancing the optimization ability of the algorithm, better avoiding falling into the local optimum of parameter identification, and enhancing the iteration accuracy and algorithm convergence.
[0073] Step 11: Determine whether the iteration terminates: If the upper limit T of the iteration times is not reached, return to Step 9 to start a new round of iteration. If the upper limit of the iteration steps is reached, end the iteration and enter Step 12.
[0074] Step 12: The iteration terminates, and output the identified parameters: Select the global optimal solution of the last round of iteration as the output to obtain the identified parameters. Judge whether the errors of all samples are within the expected range based on the national standard. If all are within the expected range, take this parameter as the final output, and do not output the specific value of current limiting.
[0075] If, in addition to the samples in the reactive current B mentioned in Step 8, the errors of the remaining samples are all within the expected range, take this parameter as the final output, and the output current limiting value is the .
[0076] If there are cases where the errors of the remaining samples exceed the expected range, return to Step 7, reselect appropriate particle numbers and iteration times, and start the calculation again.
[0077] In this embodiment, the parameter identification results are: ; ; ; ; Select one group of samples from each of the data groups with voltage dips of 20%, 40%, 60%, and 80% respectively to conduct simulation verification. The verification results are as shown in the appendix Figures 2 - 5 . It can be seen from the figure that the simulation data and the measured data are basically completely fitted, and the effect of the parameter identification method of this application is good.
[0078] It is easy for those skilled in the art to understand that the above are only the preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent replacements, and improvements made within the spirit and principle of this application shall be included within the protection scope of this application.
Claims
1. A method for identifying reactive support control parameters of a grid-following converter, characterized in that: include: Divide the converter into multiple voltage conditions for operation test, and group the obtained test data according to the voltage conditions; Perform density-based clustering on the reactive current in the test data, compare the clustering and grouping results, and find the reactive current that reaches the current limiting amplitude A Reactive current that may risk reaching the current limit B ; If there is reactive current A , then the reactive current A Delete it from the test data, and take the maximum difference between the reactive support model prediction value and the actual value as the prediction error; If there is reactive current B , then based on the reactive current B defining a clipping risk factor, and using the clipping risk factor to correct the forecast error; The parameters to be identified in the reactive support model are regarded as particles, and the optimal solution of the control parameters is found by using the genetic particle swarm algorithm with minimizing the prediction error as the optimization goal.
2. The identification method according to claim 1, characterized in that: Divide the converter into multiple voltage conditions for operation test, and group the obtained test data according to the voltage conditions, specifically: Multiple groups of tests are performed for each voltage condition. In each test, the initial reactive power is set to zero and the initial reactive current is measured. When the voltage is reduced or increased to the set voltage condition, the reactive current and the port voltage are measured. Record all test data and group the obtained test data by voltage conditions.
3. The identification method according to claim 1, characterized in that: Perform density-based clustering on the reactive current in the test data, compare the clustering and grouping results, and find the reactive current that reaches the current limiting amplitude A Reactive current that may risk reaching the current limit B ; Specifically: If the clustering results and grouping results can be made consistent by adjusting the clustering parameters, the reactive current in the voltage condition group farthest from the rated voltage is set as the reactive current that has the risk of reaching the current limit amplitude. B ; If the clustering parameters are adjusted to make only the clustering results between the rated voltage and the threshold voltage match the grouping results, and all reactive currents outside the threshold voltage are clustered into the same category and cannot match the grouping results, then the reactive current outside the threshold voltage is set to the reactive current that reaches the current limiting amplitude. A .
4. The identification method according to claim 1, characterized in that: The maximum difference between the reactive support model predicted value and the actual value is taken as the prediction error. The prediction error is specifically: in, The first Reactive current; For the test data Reactive current; is the total amount of reactive current in the test data; To find the maximum value; .
5. The identification method according to claim 1, characterized in that: Based on reactive current B Define a limiting risk factor, and use the limiting risk factor to correct the prediction error, specifically: according to the reactive current B The relationship between the mean value and 1 determines the reactive current B Is there a reactive current that reaches the current limit amplitude in the current? If so, the reactive current B The difference between the mean and 1 is multiplied by the preset limit risk assessment sensitivity to obtain the limit risk factor, and the limit risk factor is used to correct the reactive current B Part of the prediction error.
6. The identification method according to claim 1 or 5, characterized in that: The corrected prediction error is: in, ; The first Reactive current; For the test data Reactive current; Reactive current B The serial number range of internal reactive current; To find the maximum value; in, ; The limiting risk factor is: Reactive current B Mean; Evaluate the sensitivity of the risk for the preset clipping.
7. The identification method according to claim 1, characterized in that: Taking minimizing the prediction error as the optimization goal, the genetic particle swarm algorithm is used to find the optimal solution of the control parameters. Specifically, the prediction error is set as the fitness of the genetic particle swarm algorithm. For each particle, the position with the minimum fitness is the individual optimal solution; for all particles, the particle with the minimum fitness is the global optimal solution. The genetic particle swarm algorithm is used to iterate the particle optimization until the iteration stop condition is reached, and the global optimal particle is output as the optimal solution of the control parameters.
8. The identification method according to claim 1 or 7, characterized in that: The initial value mutation method is used to define the initial value of each particle in the particle swarm.
9. The identification method according to claim 8, characterized in that: The initial value of the particle is specifically: in, is the preset initial reference value; and are the preset upper and lower limits of the initial value; is a Gaussian distributed random number with mean zero and standard deviation 0.
3.
10. The identification method according to claim 1, characterized in that: The reactive power support model is divided into a low voltage reactive power support model and a high voltage reactive power support model based on the rated voltage, wherein the low voltage reactive power support model is: The high voltage reactive support model is: in, is the reactive current, is the terminal voltage, is the initial reactive current; is the voltage reference value for the low voltage strategy, is the reactive power compensation coefficient, is the initial reactive current influence coefficient, is the reactive current setting value; the parameters to be identified are: , , and .