Static var generator fault ride-through parameter identification method based on measured data

Abstract

The application discloses a static var generator fault ride-through parameter identification method based on measured data, and relates to the technical field of power electronic equipment modeling and power grid simulation. The scheme can quantize the simulation and measurement deviations of multiple working conditions and multiple fault stages by dividing the fault ride-through multiple time sequence stages, constructing a three-dimensional deviation target function fusing a multi-working-condition variance robust term, and can automatically complete the SVG fault ride-through parameter identification, is favorable to improving the matching accuracy of the electromechanical transient model and the field measured dynamic response, and simultaneously improving the stability of the identification result in the multi-working-condition scene.

Description

Technical Field

[0001] This application relates to the field of power electronic equipment modeling and power grid simulation technology, and in particular to a method for identifying fault ride-through parameters of a static var generator based on measured data. Background Technology

[0002] The large-scale grid connection of new energy sources is driving the transformation of the power system towards a higher proportion of power electronics. Static var generators (SVA), as core reactive power support equipment in new energy power plants, directly determine the transient voltage stability level of the power grid due to their high and low voltage ride-through dynamic characteristics. In accordance with the relevant management documents on power grid simulation modeling, new energy power plants need to improve their electromechanical transient parameter database for power electronic equipment and complete refined modeling based on on-site measured waveform data.

[0003] Currently, the electromechanical transient parameters of Static Var Generators (SVG) in engineering projects mostly adopt typical parameters from manufacturers or rely on manual fitting of measured data for correction. This method is highly dependent on human experience, can only identify a small number of sensitive parameters, is difficult to handle nonlinear coupling relationships between equipment parameters, and the fitting results are prone to getting trapped in local optima. Conventional optimization methods lack a unified deviation evaluation system for multiple operating conditions, resulting in large fluctuations in error evaluation under different high- and low-voltage fault ride-through conditions. Furthermore, they do not distinguish between feasible and infeasible solutions for hierarchical sorting, making it easy for the population optimization direction to deviate. This makes it impossible to achieve automated and high-precision identification of SVG fault ride-through parameters under multiple operating conditions, and it is difficult to meet the needs of refined simulation modeling of new power systems. Summary of the Invention

[0004] The purpose of this application is to provide a method for identifying fault ride-through parameters of SVG based on measured data, aiming to solve the technical problem of poor stability in the identification of fault ride-through parameters of SVG in the prior art.

[0005] To achieve the above objectives, this application provides a method for identifying fault ride-through parameters of a static var generator based on measured data, comprising: Construct an electromechanical transient parameter optimization model and an electromechanical transient simulation model for a static var generator, and initialize the multi-objective optimization population corresponding to the electromechanical transient parameter optimization model; The fault ride-through time phase of the static var generator is divided, a three-dimensional deviation objective function with a robust term for multi-condition variance is established, and a multi-objective optimization criterion is formed. Based on the multi-objective optimization criterion, a constraint priority ranking mechanism is configured in the multi-objective optimization framework of the electromechanical transient parameter optimization model to perform hierarchical ranking of feasible solutions for individuals in the multi-objective optimization population. The parameters to be identified carried by the individuals in the multi-objective optimization population are input into the electromechanical transient simulation model to generate power flow files and stability calculation files, and simulation calculations are performed to obtain simulation data. The parameters to be identified carried by the multi-objective optimization population are iteratively updated based on the deviation between the simulation data and the measured data. After the closed-loop optimization converges, the parameters to be identified of the electrical transient model are output.

[0006] In one embodiment, an electromechanical transient parameter optimization model and an electromechanical transient simulation model of a static var generator are constructed, and a multi-objective optimization population corresponding to the electromechanical transient parameter optimization model is initialized, including: Construct an electromechanical transient mathematical model of a static var generator, and determine the set of parameters to be identified carried by the multi-objective optimization population and the value range of each parameter to be identified. Construct an electromechanical transient parameter optimization model with each of the parameters to be identified as optimization variables; Configure the power grid operating conditions and equipment parameters, and establish an electromechanical transient simulation model that is linked to the electromechanical transient parameter optimization model; Multiple sets of variable vectors are generated within the value range of each parameter to be identified, and a multi-objective optimization population corresponding to the electromechanical transient parameter optimization model is obtained.

[0007] In one embodiment, the fault ride-through time phase of the static var generator is divided, a three-dimensional deviation objective function with a robust term for multi-condition variance is established, and a multi-objective optimization criterion is formed, including: The entire fault ride-through process of the static var generator is divided into multiple time stages; wherein, the multiple time stages include the pre-fault steady state, the inter-fault transient state, the inter-fault steady state, the post-fault transient state, and the post-fault steady state; Calculate the average and maximum interval deviations between the simulation data and the measured data for each time series stage; A multi-condition variance robust term is introduced to correct the staged deviation, and a three-dimensional deviation objective function with a robust term is constructed. Using the three-dimensional deviation objective function as the optimization criterion, a multi-objective optimization criterion is generated.

[0008] In one embodiment, the formula for calculating the interval average deviation is: , In the formula, The average deviation over the interval; These correspond to the pre-fault steady state, inter-fault transient state, inter-fault steady state, post-fault transient state, and post-fault steady state, respectively. This is the starting index of the simulation data range; This is the termination number of the simulation data range; This is the starting number of the measured interval; This is the end number of the measured interval; Per-unit values ​​for electromechanical transient simulation data; Per-unit values ​​are the data measured on-site or in a semi-physical test.

[0009] In one embodiment, the formula for calculating the maximum deviation of the interval is: , In the formula, The maximum deviation in the interval; Identify the target dimension; A, B2, and C2 correspond to the steady state before the fault, the steady state between faults, and the steady state after the fault, respectively.

[0010] In one embodiment, the expression for the three-dimensional deviation objective function is: , In the formula, For the first k Dimensional bias objective function; The total number of high and low voltage ride-through conditions is preset; For the first The first working condition Dimensional stage deviation value; The robustness coefficient; The variance calculation for the corresponding dimensional deviations under multiple operating conditions constitutes the multi-operating-condition variance robustness term.

[0011] In one embodiment, based on the multi-objective optimization criterion, a constraint priority ranking mechanism is configured in the multi-objective optimization framework of the electromechanical transient parameter optimization model to perform hierarchical ranking of feasible solutions for individuals within the multi-objective optimization population, including: When the number of iterations is the same as the initial iteration, the initial population is taken as the merged population. When the number of iterations is not the initial iteration, the parent population and the child population are merged to obtain the merged population. The upper and lower limits of each deviation objective function are pre-defined, and the constraint violation degree of each individual in the merged population is quantified by the constraint violation degree formula. The feasible solution set and the infeasible solution set are divided according to the degree of constraint violation. The feasible solution set has a higher priority than the infeasible solution set. The infeasible solution set is arranged in ascending order of constraint violation degree. Perform a non-dominated sort on the feasible solution set, dividing the feasible solution set into multiple dominated levels; Generate uniform reference points that fit the target constraint range and complete the target normalization process; The feasible solutions after stratification are assigned to the nearest reference point, and elite individuals are selected based on vertical distance to generate a new generation of parent population; Selective, crossover, and mutation genetic operations are performed on the new generation of parent populations to generate offspring populations.

[0012] In one embodiment, the formula for calculating the constraint violation degree is: , In the formula, For individual constraint violation degree; Identify the target dimension; This is the lower bound constraint for the k-th dimension deviation objective function; For the first Upper bound constraint on the dimensionality deviation objective function; For the first Dimensional bias objective function.

[0013] In one embodiment, the parameters to be identified carried by individuals in the multi-objective optimization population are input into the electromechanical transient simulation model to generate power flow files and stability calculation files, and simulation operations are performed to obtain simulation data, including: Write the parameters to be identified carried by individual populations into the simulation template file; Generate power flow analysis files and stability calculation analysis files based on the template files after filling in the parameters; Call the power flow calculation program to parse the power flow analysis file and output the power flow result file; The stability calculation program is invoked to read the power flow result file, perform transient stability calculations, and output the original stability result file. The stable original result file is converted to a different format to obtain the simulation data.

[0014] In one embodiment, the parameters to be identified carried by the multi-objective optimization population are iteratively updated based on the deviation between the simulation data and the measured data. After the closed-loop optimization converges, the parameters to be identified of the electromechanical transient model are output, including: Based on the multi-condition deviation objective function, the comprehensive objective value of the current population individual is calculated by combining the multi-condition variance robustness term; Elite individuals are selected to form the next generation of parent population based on a constraint-priority hierarchical ranking mechanism. Perform selection, crossover, and mutation genetic operations on the new generation of parent population to generate offspring populations carrying updated parameters to be identified; The simulation, target calculation, and population update steps are executed repeatedly until the iteration termination condition is met. After the optimization process converges, the optimal individual is selected from the final population based on the key objective priority rule, and the optimal parameters to be identified are extracted and output.

[0015] The above-mentioned technical solution of this application has at least the following beneficial technical effects: The technical solution of this application divides the fault traversal into multiple time-series stages and constructs a three-dimensional deviation objective function that integrates multi-condition variance robust terms. This function can uniformly quantify the simulation and measured deviations of multiple conditions and fault stages. The variance robust terms offset the evaluation deviations caused by a single condition, thereby improving the adaptability of parameter identification to the conditions from the perspective of evaluation criteria. By configuring a constraint priority sorting mechanism to process individual populations hierarchically, feasible solutions that meet error constraints are retained first, while invalid and infeasible individuals are filtered out, reducing optimization interference and lowering the probability of getting trapped in local optima. Combined with the automatic simulation file generation process, the simulation data of individual populations are calculated in batches. The parameters to be identified are updated by iterative closed-loop of multi-objective populations. No manual fitting and parameter tuning are required throughout the process. This can automatically complete the identification of SVG fault traversal parameters, which is conducive to improving the matching accuracy between electromechanical transient models and field measured dynamic responses, while also improving the stability of identification results under multiple conditions. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating one implementation of the method for identifying fault ride-through parameters of a static var generator based on measured data provided in this application. Figure 2 This is a flowchart of a unified simulation framework based on BPA-PSD according to a specific embodiment provided in this application; Figure 3 This is a flowchart illustrating the overall steps of an improved multi-objective optimization model according to a specific embodiment provided in this application. Figure 4 This is a schematic diagram of the five-stage timing division and deviation calculation logic for SVG fault traversal according to a specific embodiment provided in this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to specific embodiments and accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of this application. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0018] The embodiments described in this application are only some, not all, of the embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments described herein without inventive effort are within the scope of protection of this application.

[0019] The large-scale grid connection of new energy sources is driving the transformation of the power system towards a higher proportion of power electronics. Static var generators (SVA), as core reactive power support equipment in new energy power plants, directly determine the transient voltage stability level of the power grid due to their high and low voltage ride-through dynamic characteristics. In accordance with the relevant management documents on power grid simulation modeling, new energy power plants need to improve their electromechanical transient parameter database for power electronic equipment and complete refined modeling based on on-site measured waveform data.

[0020] Currently, the electromechanical transient parameters of Static Var Generators (SVG) in engineering projects mostly adopt typical parameters from manufacturers or rely on manual fitting of measured data for correction. This method is highly dependent on human experience, can only identify a small number of sensitive parameters, is difficult to handle nonlinear coupling relationships between equipment parameters, and the fitting results are prone to getting trapped in local optima. Conventional optimization methods lack a unified deviation evaluation system for multiple operating conditions, resulting in large fluctuations in error evaluation under different high- and low-voltage fault ride-through conditions. Furthermore, they do not distinguish between feasible and infeasible solutions for hierarchical sorting, making it easy for the population optimization direction to deviate. This makes it impossible to achieve automated and high-precision identification of SVG fault ride-through parameters under multiple operating conditions, and it is difficult to meet the needs of refined simulation modeling of new power systems.

[0021] To address the aforementioned technical issues, this application provides a method for identifying fault ride-through parameters of a static var generator (SVG) based on measured data. The technical solution of this application divides fault ride-through into multiple time-series stages and constructs a three-dimensional deviation objective function that integrates multi-condition variance robust terms. This function can uniformly quantify the simulation-measurement deviations across multiple conditions and fault stages. The variance robust terms offset the evaluation deviations caused by a single condition, improving the adaptability of parameter identification to the evaluation criteria. By configuring a constraint priority ranking mechanism to process individual populations hierarchically, feasible solutions that satisfy error constraints are prioritized, while invalid and infeasible individuals are filtered out, reducing optimization interference and lowering the probability of getting trapped in local optima. Combined with an automatic simulation file generation process, the simulation data of individual populations is calculated in batches. The parameters to be identified are updated through a multi-objective population iterative closed-loop process, eliminating the need for manual fitting and parameter tuning throughout the process. This automates the identification of SVG fault ride-through parameters, improving the matching accuracy between the electromechanical transient model and the measured dynamic response, while also enhancing the stability of the identification results under multiple conditions.

[0022] In one implementation, please refer to Figure 1 The method for identifying fault ride-through parameters of a static var generator based on measured data includes the following steps: Step S1: Construct an electromechanical transient parameter optimization model and an electromechanical transient simulation model for the static var generator (SVM), and initialize the multi-objective optimization population corresponding to the electromechanical transient parameter optimization model. In one embodiment, step S1 further includes the following steps: constructing an electromechanical transient mathematical model for the SVM, determining the set of parameters to be identified carried by the multi-objective optimization population and the value range of each parameter to be identified; constructing an electromechanical transient parameter optimization model with each parameter to be identified as an optimization variable; configuring the grid operating conditions and equipment parameters, and establishing an electromechanical transient simulation model linked to the electromechanical transient parameter optimization model; generating multiple sets of variable vectors within the value range of each parameter to be identified, and obtaining the multi-objective optimization population corresponding to the electromechanical transient parameter optimization model.

[0023] The specific implementation process of step S1 is as follows: First, build an electromechanical transient mathematical model of the static var generator (SVM), and sort out all control parameters such as voltage droop ratio coefficient, transient reactive current regulation ratio coefficient, and voltage slope adjustment coefficient. From this, select parameters with higher weights affecting the dynamic response of high and low voltage ride-through as the set of parameters to be identified. At the same time, define the independent upper and lower limit ranges of each parameter to be identified based on the operating limits of the equipment, which can accurately lock the range of optimization variables, avoid meaningless parameter optimization, and reduce computational overhead. Then, use all parameters to be identified as optimization variables to build an electromechanical transient parameter optimization model to achieve multi-parameter synchronous joint optimization. This can simultaneously handle the nonlinear coupling relationship between parameters and overcome the deficiency that manual fitting can only identify a small number of parameters. Subsequently, configure equipment parameters corresponding to various SVG topologies and control strategies, as well as a total of 90 sets of parameters such as no-load low ride-through, capacitive high-power low ride-through, capacitive high-power high ride-through, and inductive high-power low ride-through. For high-voltage and low-voltage power grid operation, an electromechanical transient simulation model is built that can be linked with the optimization model in real time. This model can adapt to the simulation needs of equipment in multiple scenarios and expand the applicability of parameter identification. Finally, a Latin hypercube sampling method is used to generate multiple sets of variable vectors to construct a multi-objective optimization population. First, the value range of each parameter to be identified is evenly divided into several non-overlapping sub-intervals. In each sub-interval, a uniformly random number between 0 and 1 is generated. Combined with the interval boundary, the sampling points of each parameter are calculated. All parameter sampling points are combined to form a dimension-matched variable vector. The initial population matrix is ​​obtained by combining all vectors. After completion, boundary verification is performed on all sampled parameters. If a parameter exceeds the preset upper and lower limits, it is directly corrected to the interval boundary value to ensure that all individual parameters in the population are legal and valid. Latin hypercube sampling can evenly cover the entire parameter value range. Compared with random sampling, it improves the diversity of the initial population sample and reduces the probability of the algorithm getting trapped in local optima early.

[0024] Step S2: Divide the fault ride-through time phase of the static var generator, establish a three-dimensional deviation objective function with a robust term for multi-condition variance, and form a multi-objective optimization criterion; in one embodiment, step S2 further includes the following steps: Step S21: Divide the entire fault ride-through process of the static var generator into multiple time stages; among which, the multiple time stages include the pre-fault steady state, the inter-fault transient state, the inter-fault steady state, the post-fault transient state, and the post-fault steady state; Step S22: Calculate the average deviation and maximum deviation between the simulation data and the measured data for each time series stage; in one embodiment, the formula for calculating the average deviation is: , In the formula, The average deviation over the interval; These correspond to the pre-fault steady state, inter-fault transient state, inter-fault steady state, post-fault transient state, and post-fault steady state, respectively. This is the starting index of the simulation data range; This is the termination number of the simulation data range; This is the starting number of the measured interval; This is the end number of the measured interval; Per-unit values ​​for electromechanical transient simulation data; Per-unit values ​​are the data measured on-site or in a semi-physical test.

[0025] In one embodiment, the formula for calculating the maximum deviation of the interval is: , In the formula, The maximum deviation in the interval; Identify the target dimension; A, B2, and C2 correspond to the steady state before the fault, the steady state between faults, and the steady state after the fault, respectively.

[0026] Step S23: Introduce a multi-condition variance robust term to correct the staged deviation, and construct a three-dimensional deviation objective function containing the robust term; in one embodiment, the expression of the three-dimensional deviation objective function is: , In the formula, For the first k Dimensional bias objective function; The total number of high and low voltage ride-through conditions is preset; For the first The first working condition Dimensional stage deviation value; The robustness coefficient; The variance calculation for the corresponding dimensional deviations under multiple operating conditions constitutes the multi-operating-condition variance robustness term.

[0027] Step S24: Use the three-dimensional deviation objective function as the optimization criterion to generate a multi-objective optimization criterion.

[0028] The specific implementation process of step S2 is as follows: The entire fault traversal process of SVG is divided into five time-series stages: pre-fault steady state, inter-fault transient state, inter-fault steady state, post-fault transient state, and post-fault steady state. This fully covers the entire dynamic process before and after the fault occurs, ensuring that the deviation between simulation and measured data can be quantified throughout the entire time period, without omitting any error in the transient or steady state stage. For the three key electrical physical quantities of voltage, reactive power, and reactive current, the interval average deviation is uniformly calculated for all five time-series stages. The interval average deviation is obtained by taking the absolute value of the difference between the simulation segmented mean and the measured segmented mean, which can quantify the overall data offset of each fault stage. For the three steady-state stages of pre-fault steady state, inter-fault steady state, and post-fault steady state, the maximum interval deviation is additionally calculated by extracting the absolute maximum value of the difference between simulation and measured data within the same time period, accurately capturing instantaneous extreme errors under steady-state conditions. To address the shortcomings of mean indicators in reflecting single-point exceedances, independent weighting coefficients are assigned to five time-series stages to differentiate the impact of different fault stages on equipment dynamic characteristics. Weighted averages yield a comprehensive segmented deviation under a single operating condition. A robustness coefficient ranging from 0.1 to 0.5 is then introduced to construct a multi-operating-condition variance robustness term. A larger robustness coefficient emphasizes suppressing multi-operating-condition deviation fluctuations. The mean of the segmented deviations for all 90 sets of high-low crossing operating conditions is calculated, and the variance robustness term is superimposed to construct a three-dimensional deviation objective function. The variance robustness term can offset evaluation biases caused by anomalies in single-operating-condition data, significantly improving the stability of parameter evaluation in multi-operating-condition scenarios. Finally, the three-dimensional deviation objective function is used as a unified optimization criterion to generate multi-objective optimization criteria, achieving integrated quantitative evaluation of multiple electrical quantities, multiple fault stages, and multiple operating conditions. This solves the problems of one-sided evaluation and poor adaptability of traditional single-indicator evaluations.

[0029] Step S3: Based on the multi-objective optimization criterion, a constraint priority ranking mechanism is configured in the multi-objective optimization framework of the electromechanical transient parameter optimization model to perform hierarchical ranking of feasible solutions for individuals within the multi-objective optimization population; in one embodiment, step S3 further includes the following steps: Step S31: When the number of iterations is the initial iteration, take the initial population as the merged population; when the number of iterations is not the initial iteration, merge the parent population and the child population to obtain the merged population.

[0030] Step S32: Pre-set the upper and lower limits of the objective function for each deviation, and quantify the constraint violation degree of each individual in the merged population using the constraint violation degree formula; in one embodiment, the constraint violation degree calculation formula is: , In the formula, For individual constraint violation degree; Identify the target dimension; This is the lower bound constraint for the k-th dimension deviation objective function; For the first Upper bound constraint on the dimensionality deviation objective function; For the first Dimensional bias objective function.

[0031] Step S33: Divide the feasible solution set into feasible and infeasible solution sets according to the degree of constraint violation. The feasible solution set has a higher priority than the infeasible solution set. The infeasible solution set is arranged in ascending order of constraint violation degree. Step S34: Perform a non-dominated sort on the feasible solution set to divide the feasible solution set into multiple dominance levels; Step S35: Generate uniform reference points that fit the target constraint range and complete the target normalization process; Step S36: Assign the stratified feasible solutions to the nearest reference points, and select elite individuals based on vertical distance to generate a new generation of parent population; Step S37: Perform selection, crossover, and mutation genetic operations on the new generation of parent population to generate offspring population.

[0032] The specific implementation process of step S3 is as follows: During the first iteration, the initial population generated in step S1 is directly used as the merged population. When the number of iterations is greater than or equal to 1, the parent population generated in the previous iteration is merged with the child population to obtain the merged population, fully preserving high-quality parameter individuals from each generation to avoid losing high-quality solutions. Independent upper and lower bound constraints are pre-set for each objective dimension of the three-dimensional deviation objective function. The constraint violation degree of each individual in the merged population is quantified using the constraint violation degree formula. The constraint violation degree is obtained by summing the exceedance values ​​of the objective parameters in each dimension. A constraint violation degree of 0 indicates that all indicators of that individual satisfy the error constraints and are a feasible solution; a value greater than... A value of 0 indicates that the index exceeds the limit, making the solution infeasible. This allows for quick differentiation between valid and invalid parameter individuals, reducing computational waste caused by invalid individuals participating in subsequent calculations. The solution set is divided according to the constraint violation degree. The overall priority of the feasible solution set is higher than that of all infeasible solutions. Within the infeasible solution set, it is sorted in ascending order of constraint violation degree, prioritizing the selection of parameter individuals that meet the error requirements and reducing interference from invalid individuals on the optimization direction. Non-dominated hierarchical sorting is performed only on the feasible solution set, dividing it into multiple dominance levels. The first level is the optimal level where no other individual can dominate it, and subsequent levels can be dominated by solutions in higher levels. This hierarchical approach clearly distinguishes the superiority and inferiority levels of individuals. For subsequent elite selection; generate a uniform set of reference points adapted to the target constraints of each dimension, scale and adapt the original reference points according to the target upper and lower limits, and perform target normalization processing on all hierarchical feasible solutions to eliminate the differences in the dimensions of different target dimensions, ensure that the weights of each deviation index are equal, and avoid a single index dominating the optimization result; starting from the optimal dominance level, allocate feasible solutions to the nearest reference points layer by layer, and determine the quality of individuals by the degree of matching of the vertical distance between individuals and reference points. When multiple individuals correspond to the same reference point, only the individual with the smallest vertical distance is retained as an elite. Select elites layer by layer until the preset population size is reached to form a new generation of parent population, which can The system stably retains high-quality individuals with balanced parameters across multiple dimensions. A binary tournament selection process is used to select and cross-reference the next generation of elite parent populations. Simulated binary cross-linking is employed with fixed cross-linking probabilities and exponents. After the cross-linking operation, parameter value boundaries are verified, and parameters exceeding the limits are forcibly corrected to a preset range. Then, a polynomial mutation operation is used, with the mutation probability set to the reciprocal of the total dimensions of the parameters to be identified, along with a fixed mutation exponent. After mutation, parameter boundaries are verified again. This entire genetic process of selection, cross-linking, and mutation generates a new offspring population. Through hierarchical elite selection combined with standardized genetic operations, the quality of population parameters is continuously optimized, reducing the probability of the algorithm converging to a local optimum.

[0033] Step S4: Input the parameters to be identified carried by individuals in the multi-objective optimization population into the electromechanical transient simulation model, generate power flow files and stability calculation files, and perform simulation calculations to obtain simulation data; in one embodiment, step S4 further includes the following steps: Step S41: Write the parameters to be identified carried by individuals in the population into the simulation template file; Step S42: Generate power flow analysis file and stability calculation analysis file based on the template file after filling parameters; Step S43: Call the power flow calculation program to parse the power flow analysis file and output the power flow result file; Step S44: Call the stability calculation program to read the power flow result file, perform transient stability calculation, and output the original stability result file; Step S45: Convert the format of the stable original result file to obtain simulation data.

[0034] The specific implementation process of step S4 is as follows: Extract the complete set of parameters to be identified carried by each individual in the multi-objective optimization population, and write each parameter into a pre-constructed electromechanical transient simulation blank template file to complete parameter filling. This achieves batch automatic import of population parameters into the simulation model, eliminating the need for manual modification of simulation files and significantly improving batch simulation efficiency. Based on the template file with filled parameters, two types of simulation input files are automatically generated: power flow analysis files and stability calculation analysis files. A unified and standardized simulation input format is adopted for reading by the corresponding calculation program. The power flow calculation program is called to parse the power flow analysis file and perform power flow calculations, outputting a power flow result file with the .BSE format extension, providing basic power grid operation data for transient stability calculations. The stability calculation program is called to read the .BSE power flow result file and perform transient stability simulation calculations, outputting a raw stability result file with the .SWX format extension, completely recording the electrical time-series data of the entire fault process. The .SWX raw result file is converted to a standardized tabular format for outputting simulation time-series data. This unified data format facilitates subsequent point-by-point comparison with measured data to calculate various deviations, avoiding the problems of incompatibility between the original simulation file format and the inability to directly calculate deviations.

[0035] Step S5: Iteratively update the parameters to be identified carried by the multi-objective optimization population based on the deviation between the simulation data and the measured data. After the closed-loop optimization converges, output the parameters to be identified for the electromechanical transient model. In one embodiment, step S5 further includes the following specific steps: Calculate the comprehensive objective value of the current population individuals based on the multi-condition deviation objective function and the multi-condition variance robustness term; select elite individuals to form a new generation of parent population based on the constraint priority hierarchical sorting mechanism; perform selection, crossover, and mutation genetic operations on the new generation of parent population to generate a offspring population carrying the updated parameters to be identified; cyclically execute the simulation, objective calculation, and population update steps until the iteration termination condition is met; after the optimization process converges, select the optimal individuals from the final population based on the key objective priority rule, extract the optimal parameters to be identified, and output them.

[0036] The specific implementation process of step S5 is as follows: Read the standardized simulation data and field measured / semi-physical test data output in step S4; based on the three-dimensional deviation objective function constructed in step S2, superimpose the multi-condition variance robustness term to calculate the comprehensive objective value corresponding to each parameter individual in the current population, and fully quantify the overall matching degree between simulation and actual measurement under multiple conditions, multiple fault stages, and multiple electrical quantities; substitute the comprehensive objective values ​​and constraint violation degrees of all individuals into the constraint priority hierarchical sorting mechanism in step S3 to complete the solution set partitioning, hierarchical, and elite screening to obtain a new generation of parent population, and continuously eliminate parameter individuals with excessive errors and poor stability under multiple conditions; perform a complete set of genetic operations of selection, crossover, and mutation on the new generation of parent population to generate a offspring population carrying updated parameters to be identified, and complete parameter iterative optimization; repeat the steps in a loop. The entire process of S4 batch simulation, comprehensive target value calculation, and population hierarchical update is set with two types of iteration termination conditions. The optimization convergence is determined when the number of iterations reaches the preset maximum number of iterations or when there is no significant change in the frontier of multiple dominance levels for several consecutive rounds. The dual termination conditions take into account both computational efficiency and optimization sufficiency, avoiding premature stopping that would lead to insufficient parameter accuracy or meaningless continuous iteration. After optimization convergence, the optimal individual is extracted from the final converged population using a key target priority screening rule. The complete set of parameters to be identified carried by this individual is decomposed and output, resulting in accurate identification parameters of SVG electromechanical transients that are adapted to all fault time series and multiple high and low crossing conditions. The parameter identification is completed automatically throughout the entire process without the need for repeated manual parameter tuning and fitting, which greatly reduces the workload of manual modeling and improves the matching accuracy between the electromechanical transient model and the actual dynamic response of the equipment.

[0037] In one specific embodiment, please refer to Figures 2 to 4 , Figure 2 This document provides a complete demonstration of the entire process of generating power flow and stability calculation files from parameter template files, calling the accompanying calculation program to complete transient simulations, and converting simulation data. It is used for batch solving of SVG electromechanical transient simulation output data under multiple operating conditions. Figure 3 This invention fully demonstrates the entire process of population initialization, multi-condition target calculation, constraint priority hierarchical sorting, reference point normalization, elite selection, genetic operations, and iterative convergence determination, serving as the core optimization logic carrier. Figure 4The system divides the time series into five intervals: pre-fault steady state, inter-fault transient state, inter-fault steady state, post-fault transient state, and post-fault steady state. These intervals correspond to the calculation logic for the average and maximum deviations of three physical quantities: voltage, reactive power, and reactive current. This provides fundamental data for multi-objective evaluation. This embodiment relies on the electromechanical transient modeling scenario of a single-unit SVG at a new energy power station for identification. During implementation, measured waveform data of SVG high and low voltage ride-through and test data from a semi-physical platform are collected first. A set of operating conditions containing 90 high and low voltage ride-through scenarios is then built, covering no-load low-voltage ride-through, capacitive high-power low-voltage ride-through, capacitive high-power high-voltage ride-through, and inductive high-power ride-through. In typical operating scenarios such as low-voltage ride-through, the least squares method is used to fit basic control parameters such as voltage droop ratio, transient reactive current regulation ratio, and voltage slope adjustment coefficient. Then, a local sensitivity analysis method based on single-factor parameter disturbance scanning is employed to observe the dynamic output changes of voltage, reactive power, and reactive current within each disturbance model. The set of key parameters with high weighting on the dynamic response of high and low voltage ride-through is selected based on the sensitivity values ​​as the dimensions of parameters to be identified in subsequent optimization. Subsequently, a single-machine electromechanical transient simulation template file adapted to multiple types of SVG topologies and control strategies is built, relying on... Figure 2 The BPA-PSD unified simulation framework is used for batch simulations. First, the parameters to be identified carried by individuals in the optimization population are written into a blank simulation template file to complete parameter filling. Based on the filled template file, two types of input files are automatically generated: a power flow analysis file and a stability calculation analysis file. The power flow calculation program PFNT.exe is called to read the power flow analysis file, perform calculations, and output a power flow result file with the .BSE extension. Then, the stability calculation program SWNT.exe is called to read the .BSE power flow result file to perform transient stability simulation, outputting a raw stability result file with the .SWX extension. The .SWX raw result file is format-converted to obtain standardized tabular simulation data for deviation comparison, simultaneously relying on... Figure 4 A three-dimensional deviation evaluation system for multiple time-series stages was constructed. The complete fault-crossing process of SVG was divided into five time-series intervals: pre-fault steady state, inter-fault transient state, inter-fault steady state, post-fault transient state, and post-fault steady state. The average deviation of each time-series interval was calculated for three physical quantities: voltage, reactive power, and reactive current. The maximum deviation was calculated only for the three steady-state intervals: pre-fault steady state, inter-fault steady state, and post-fault steady state. Independent weighting coefficients were assigned to each of the five time-series intervals. A single-condition deviation index was obtained by weighted summation of all time-series deviations under a single operating condition. A robust balance coefficient ranging from 0.1 to 0.5 was introduced to integrate the fluctuation characteristics of all deviation indices under multiple operating conditions, constructing a comprehensive evaluation objective that considers the stability of multiple operating conditions and forming a complete multi-objective optimization criterion. The larger the robust balance coefficient value, the more prioritized the algorithm is in ensuring the consistency of identification results under different operating conditions. Then, according to... Figure 3The process involves building and improving a multi-objective optimization model and initializing the population. The total dimension D of the parameters to be identified is set, and the population size N is fixed. Independent upper and lower limit ranges are defined for each parameter. A random sampling seed is configured to ensure reproducibility. Latin hypercube sampling is used for population initialization. First, the value range of each parameter is evenly divided into M non-overlapping sub-intervals. Within each sub-interval, a random number between 0 and 1 is generated and evenly distributed. The sampling points for that parameter are calculated based on the sub-interval boundaries. These D-dimensional parameter sampling points are combined to form N groups of D-dimensional individuals, constituting the initial population matrix. Boundary checks are then performed. If a sampled parameter exceeds the preset upper and lower limits, it is directly corrected to the corresponding interval boundary value to ensure the validity of all individual parameters. Population initialization is then complete. Subsequently, each parameter individual within the population is sequentially fed into the BPA-PSD simulation framework, traversing all 90 high- and low-voltage driving conditions to perform simulations. The time-series deviation indices corresponding to voltage, reactive power, and reactive current under each condition are obtained. For each target dimension, the average value of all condition deviation indices is first calculated, and then a multi-condition fluctuation balance term is superimposed to obtain the comprehensive target value for that dimension. Independent upper and lower limits are pre-set for each evaluation target. Based on the constraint boundaries, the constraint violation degree of each individual is quantified. A constraint violation degree value of 0 indicates that the individual is a feasible solution, while a value greater than 0 indicates that the individual exceeds the target constraint range and is an infeasible solution. Subsequently, a constraint-priority hierarchical sorting process is performed on the population individuals. The initial iteration directly uses the initial population as the merged population. When the number of iterations is greater than or equal to 1, the parent and offspring populations from the previous round are merged to obtain a merged population. Based on the constraint violation degree, two solution sets are divided: all individuals with a constraint violation degree of 0 are classified as feasible solutions, and the rest as infeasible solutions. The overall priority of feasible solutions is higher than that of all infeasible solutions. Within the infeasible solution set, they are arranged in ascending order of constraint violation degree. Non-dominated hierarchical processing is performed only on feasible solutions, dividing them layer by layer into multiple dominance levels. The first level is the optimal level with no other solutions dominating it; each subsequent level can be dominated by solutions in the upper level. After hierarchical processing, a uniformly distributed set of reference points adapted to the target constraint range is generated. The original reference points are scaled and adapted based on the upper and lower limits of the target in each dimension. Target normalization processing is then performed on all hierarchical feasible solutions. To eliminate the dimensional differences between different evaluation objectives, each feasible solution is assigned to the nearest reference point layer by layer, starting from the first dominance level. The degree of matching is determined by the vertical distance between the solution and the reference point. When multiple individuals are assigned to the same reference point, only the individual with the smallest vertical distance is retained as the elite. Elite individuals are selected sequentially from each dominance level until the preset population size N is reached, forming a new generation of parent populations. A binary tournament selection operation is performed on the new generation of parent elite populations to screen crossover parents. Simulated binary crossover operation is used with a fixed crossover probability and crossover exponent. After crossover, the parameter boundaries are checked, and parameters exceeding the limits are forcibly corrected to within the interval. Then, a polynomial mutation operation is used, with the mutation probability set to the reciprocal of the total parameter dimension D, and a fixed mutation exponent is used.After mutation, the parameter value range is re-verified. A new offspring population is generated through selection, crossover, and mutation, and then re-entered into the multi-condition simulation stage to start a new round of iterations. Two termination criteria are set during the iteration process; the loop stops when either criterion is met: the number of iterations reaches the preset maximum number of iterations, and the multi-objective dominance level front shows no significant change for several consecutive iterations, indicating convergence. After iteration terminates, the optimal individual is selected from the final converged population using a key objective priority selection rule. All the parameters to be identified carried by this individual are extracted as key identification parameters for fault passage in the SVG electromechanical transient model, completing the entire identification process.

[0038] This application aims to protect a method for identifying fault ride-through parameters of a static var generator (SVG) based on measured data. The technical solution divides the fault ride-through process into multiple time-series stages and constructs a three-dimensional deviation objective function that integrates a robust variance term across multiple operating conditions. This function can uniformly quantify the simulation-measurement deviations across multiple operating conditions and fault stages. The robust variance term offsets the evaluation deviations caused by a single operating condition, improving the adaptability of parameter identification to operating conditions from the perspective of evaluation criteria. By configuring a constraint priority ranking mechanism, the method processes individual populations hierarchically, prioritizing feasible solutions that meet error constraints and filtering out invalid and infeasible individuals, reducing optimization interference and lowering the probability of getting trapped in local optima. Combined with an automatic simulation file generation process, the method batch-calculates the simulation data of individual populations and relies on a multi-objective population iterative closed-loop update of the parameters to be identified. The entire process requires no manual fitting or parameter tuning, enabling automated identification of SVG fault ride-through parameters. This improves the matching accuracy between the electromechanical transient model and the measured dynamic response, while also enhancing the stability of the identification results under multiple operating conditions.

[0039] It should be understood that the specific embodiments described above are merely illustrative or explanatory of the principles of this application and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of this application should be included within the protection scope of this application. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

Claims

1. A method for identifying fault ride-through parameters of a static var generator based on measured data, characterized in that, include: Construct an electromechanical transient parameter optimization model and an electromechanical transient simulation model for a static var generator, and initialize the multi-objective optimization population corresponding to the electromechanical transient parameter optimization model; The fault ride-through time phase of the static var generator is divided, a three-dimensional deviation objective function with a robust term for multi-condition variance is established, and a multi-objective optimization criterion is formed. Based on the multi-objective optimization criterion, a constraint priority ranking mechanism is configured in the multi-objective optimization framework of the electromechanical transient parameter optimization model to perform hierarchical ranking of feasible solutions for individuals in the multi-objective optimization population. The parameters to be identified carried by the individuals in the multi-objective optimization population are input into the electromechanical transient simulation model to generate power flow files and stability calculation files, and simulation calculations are performed to obtain simulation data. The parameters to be identified carried by the multi-objective optimization population are iteratively updated based on the deviation between the simulation data and the measured data. After the closed-loop optimization converges, the parameters to be identified of the electromechanical transient model are output.

2. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 1, characterized in that, Construct an electromechanical transient parameter optimization model and an electromechanical transient simulation model for a static var generator, and initialize the multi-objective optimization population corresponding to the electromechanical transient parameter optimization model, including: Construct an electromechanical transient mathematical model of a static var generator, and determine the set of parameters to be identified carried by the multi-objective optimization population and the value range of each parameter to be identified. Construct an electromechanical transient parameter optimization model with each of the parameters to be identified as optimization variables; Configure the power grid operating conditions and equipment parameters, and establish an electromechanical transient simulation model that is linked to the electromechanical transient parameter optimization model; Multiple sets of variable vectors are generated within the value range of each parameter to be identified, and a multi-objective optimization population corresponding to the electromechanical transient parameter optimization model is obtained.

3. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 1, characterized in that, The fault ride-through time sequence of the static var generator (SVM) is divided, and a three-dimensional deviation objective function with a robust term for multi-condition variance is established to form a multi-objective optimization criterion, including: The entire fault ride-through process of a static var generator is divided into multiple time-series stages; wherein, the multiple time-series stages include the pre-fault steady state, the inter-fault transient state, the inter-fault steady state, the post-fault transient state, and the post-fault steady state. Calculate the average and maximum interval deviations between the simulation data and the measured data for each time series stage; A multi-condition variance robust term is introduced to correct the staged deviation, and a three-dimensional deviation objective function with a robust term is constructed. Using the three-dimensional deviation objective function as the optimization criterion, a multi-objective optimization criterion is generated.

4. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 3, characterized in that, The formula for calculating the average deviation of the interval is: ， In the formula, The average deviation over the interval; These correspond to the pre-fault steady state, inter-fault transient state, inter-fault steady state, post-fault transient state, and post-fault steady state, respectively. This is the starting index of the simulation data range; This is the termination number of the simulation data range; This is the starting number of the measured interval; This is the end number of the measured interval; Per-unit values ​​for electromechanical transient simulation data; Per-unit values ​​are the data measured on-site or in a semi-physical test.

5. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 4, characterized in that, The formula for calculating the maximum deviation of the interval is: ， In the formula, The maximum deviation in the interval; Identify the target dimension; A, B2, and C2 correspond to the steady state before the fault, the steady state between faults, and the steady state after the fault, respectively.

6. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 5, characterized in that, The expression for the three-dimensional deviation objective function is: ， In the formula, For the first k Dimensional bias objective function; The total number of high and low voltage ride-through conditions is preset; For the first The first working condition Dimensional stage deviation value; The robustness coefficient; The variance calculation for the corresponding dimensional deviations under multiple operating conditions constitutes the multi-operating-condition variance robustness term.

7. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 6, characterized in that, Based on the aforementioned multi-objective optimization criterion, a constraint priority ranking mechanism is configured in the multi-objective optimization framework of the electromechanical transient parameter optimization model to perform hierarchical ranking of feasible solutions for individuals within the multi-objective optimization population, including: When the number of iterations is the same as the initial iteration, the initial population is taken as the merged population. When the number of iterations is not the initial iteration, the parent population and the child population are merged to obtain the merged population. The upper and lower limits of each deviation objective function are pre-defined, and the constraint violation degree of each individual in the merged population is quantified by the constraint violation degree formula. The feasible solution set and the infeasible solution set are divided according to the degree of constraint violation. The feasible solution set has a higher priority than the infeasible solution set. The infeasible solution set is arranged in ascending order of constraint violation degree. Perform a non-dominated sort on the feasible solution set, dividing the feasible solution set into multiple dominated levels; Generate uniform reference points that fit the target constraint range and complete the target normalization process; The feasible solutions after stratification are assigned to the nearest reference point, and elite individuals are selected based on vertical distance to generate a new generation of parent population; Selective, crossover, and mutation genetic operations are performed on the new generation of parent populations to generate offspring populations.

8. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 7, characterized in that, The formula for calculating the constraint violation degree is as follows: ， In the formula, For individual constraint violation degree; Identify the target dimension; This serves as the lower bound constraint for the k-th dimension deviation objective function; For the first Upper bound constraint on the dimensionality deviation objective function; For the first Dimensional bias objective function.

9. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 8, characterized in that, The parameters to be identified carried by the individuals in the multi-objective optimization population are input into the electromechanical transient simulation model to generate power flow files and stability calculation files, and simulation operations are performed to obtain simulation data, including: Write the parameters to be identified carried by individual populations into the simulation template file; Generate power flow analysis files and stability calculation analysis files based on the template files after filling in the parameters; Call the power flow calculation program to parse the power flow analysis file and output the power flow result file; Call the stability calculation program to read the power flow result file, perform transient stability calculation, and output the original stability result file; The stable original result file is converted to a different format to obtain the simulation data.

10. The method for identifying fault ride-through parameters of a static var generator based on measured data according to claim 9, characterized in that, The parameters to be identified carried by the multi-objective optimization population are iteratively updated based on the deviation between the simulation data and the measured data. After the closed-loop optimization converges, the parameters to be identified of the electromechanical transient model are output, including: Based on the multi-condition deviation objective function, the comprehensive objective value of the current population individual is calculated by combining the multi-condition variance robustness term; Elite individuals are selected to form the next generation of parent population based on a constraint-priority hierarchical ranking mechanism. Perform selection, crossover, and mutation genetic operations on the new generation of parent population to generate offspring populations carrying updated parameters to be identified; The simulation, target calculation, and population update steps are executed repeatedly until the iteration termination condition is met. After the optimization process converges, the optimal individual is selected from the final population based on the key objective priority rule, and the optimal parameters to be identified are extracted and output.

Images