Inverter parameter identification method based on multi-point crossing and related device

Abstract

The invention relates to the technical field of inverters, and discloses an inverter parameter identification method based on multi-point crossover and a related device, and the method comprises the steps: determining a first population corresponding to a model parameter of an inverter model under a current iteration number G; performing evolution processing on the first population by using a preset multi-point crossover rule to obtain a target population; determining target fitness according to a predetermined real response value of the inverter, an estimated response value of the inverter under the model parameter individuals in the target population and a preset criterion function; if the target fitness does not meet the preset convergence condition, G is made to be equal to G + 1, the step of determining the first population corresponding to the model parameters of the inverter model under the current iteration number is executed again, and G is the current iteration number; and if the target fitness meets a preset convergence condition, taking a model parameter corresponding to the optimal model parameter individual in the target population as an optimal identification result. Parameter identification is carried out through a multi-point crossing rule, so that a parameter identification result is more reliable and accurate.

Description

Technical Field

[0001] This invention relates to the field of inverter technology, and in particular to an inverter parameter identification method and related apparatus based on multi-point cross-connection. Background Technology

[0002] In the application of inverters, accurate model parameters play a crucial role in inverter performance evaluation, fault diagnosis, and control strategy optimization. Traditional methods for identifying inverter parameters often have many drawbacks.

[0003] On the one hand, some methods based on fixed models and simple calculations are difficult to adapt to the complex changes in inverters under different operating conditions, resulting in limited identification accuracy. For example, under conditions such as grid voltage fluctuations and sudden load changes, these methods cannot effectively capture the real changes in inverter parameters, leading to deviations in subsequent control and evaluation based on these parameters.

[0004] On the other hand, while some iterative optimization methods improve the accuracy of parameter identification to a certain extent, they are prone to getting trapped in local optima during the evolutionary process, resulting in slow convergence. For example, using an evolutionary algorithm with only single-point crossover results in insufficient population diversity, making it difficult to effectively explore the entire solution space, thus prolonging the parameter identification cycle and failing to meet the actual need for fast and accurate acquisition of inverter parameters.

[0005] Given the problems with traditional inverter parameter identification methods, there is an urgent need for a new inverter parameter identification method that can effectively improve identification accuracy, accelerate convergence speed, and has strong adaptability, so as to better serve the development and application of inverter-related technologies. Summary of the Invention

[0006] The main objective of this invention is to provide a method and related apparatus for inverter parameter identification based on multi-point cross-connection, which can solve the problem in the prior art of lacking a new method for inverter parameter identification that can effectively improve identification accuracy, accelerate convergence speed and has strong adaptability.

[0007] To achieve the above objectives, the first aspect of the present invention provides a method for inverter parameter identification based on multi-point crossover, the method comprising: Determine the first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes several individual model parameters; The first population is subjected to evolutionary processing using a preset multi-point crossover rule to obtain the target population; The target fitness is determined based on the predetermined actual response value of the inverter, the estimated response value of the inverter under the model parameter individuals in the target population, and the preset criterion function. If the target fitness does not meet the preset convergence condition, then let G = G + 1, and return to the step of determining the first group corresponding to the model parameters of the inverter model at the current iteration number, where G is the current iteration number; If the target fitness satisfies the preset convergence condition, then the model parameters corresponding to the individual with the best model parameters in the target population are taken as the best identification result.

[0008] To achieve the above objectives, a second aspect of the present invention provides an inverter parameter identification device based on multi-point crossover, the device comprising: Population determination module: used to determine the first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes a number of individual model parameters; Population evolution module: used to perform evolutionary processing on the first population using preset multi-point crossover rules to obtain the target population; Fitness determination module: used to determine the target fitness based on the predetermined actual response value of the inverter, the estimated response value of the inverter under the model parameter individuals in the target population, and the preset criterion function; Convergence Judgment Module: If the target fitness does not meet the preset convergence condition, then let G=G+1 and return to the step of determining the first group corresponding to the model parameters of the inverter model under the current iteration number, where G is the current iteration number; Parameter determination module: If the target fitness satisfies the preset convergence condition, the model parameters corresponding to the individual with the best model parameters in the target population are taken as the best identification result.

[0009] To achieve the above objectives, a third aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0010] To achieve the above objectives, a fourth aspect of the present invention provides a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0011] The embodiments of the present invention have the following beneficial effects: This invention provides a method for inverter parameter identification based on multi-point crossover. The method includes: determining a first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes several individual model parameters; performing evolutionary processing on the first population using a preset multi-point crossover rule to obtain a target population; determining the target fitness based on the predetermined true response value of the inverter, the estimated response value of the inverter under the individual model parameters in the target population, and a preset criterion function; if the target fitness does not meet the preset convergence condition, then let G = G + 1, and return to the step of determining the first population corresponding to the model parameters of the inverter model at the current iteration number, where G is the current iteration number; if the target fitness meets the preset convergence condition, then the model parameter corresponding to the best individual model parameter in the target population is taken as the best identification result. By introducing a multi-point crossover rule to perform evolutionary processing on the inverter model parameter population, the diversity of the population can be effectively enhanced, avoiding getting trapped in local optima, improving the global search capability of parameter identification, accelerating the convergence speed, broadening the scope of problem solving, and making the parameter identification results more reliable and accurate. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] in: Figure 1 This is a flowchart of an inverter parameter identification method based on multi-point crossover in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a photovoltaic inverter virtual synchronous generator system according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a virtual synchronous generator control strategy in an embodiment of the present invention; Figure 4 This is a structural block diagram of an inverter parameter identification device based on multi-point crossover in an embodiment of the present invention; Figure 5 This is a structural block diagram of a computer device in an embodiment of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] It should be noted that, in order to improve the accuracy of model parameter identification, this application uses an improved differential evolution algorithm to identify model parameters. Specifically, this invention provides a photovoltaic inverter modeling and parameter identification method based on virtual synchronous generator (VSG) technology, aiming to improve the voltage support capability of the power grid and solve the problem of modeling and simulating high-proportion new energy power systems. With the energy system transitioning towards cleaner and lower-carbon energy, distributed photovoltaic power generation systems, as a form of distributed generation, are increasingly being integrated into the power grid. However, these systems, connected to the grid via power electronic converters, lack sufficient inertia and damping support, leading to reduced grid stability. Therefore, this invention employs VSG control technology to enhance the grid support function of the power electronic converters by simulating the characteristics of synchronous generators. The new energy power system can be the aforementioned distributed photovoltaic power generation system, and the new energy power system includes at least power electronic converters such as inverters, which can be VSG-based inverters.

[0016] This invention first constructs a simulation model of a photovoltaic inverter in PSCAD software and verifies the effective voltage support capability of VSG technology for the power grid. Subsequently, a mathematical model of the inverter is established in the MATLAB environment, and an improved differential evolution algorithm is used to identify key parameters, significantly improving the accuracy of parameter identification and the algorithm's global search capability. By simulating different grid disturbances, such as voltage dips, the self-descriptive ability, generalization ability, and parameter stability of the established model are verified, demonstrating the model's stability and practicality under different disturbance conditions. The beneficial effect of this invention is that it provides a modeling method that can accurately simulate the external characteristics of photovoltaic inverters, which helps to promote the modeling and research of distributed generation with VSG, and has important practical application value for improving grid stability and the integration of new energy sources. Please refer to the following content for details.

[0017] Please see Figure 1 , Figure 1 This is a flowchart illustrating a multi-point crossover-based inverter parameter identification method in an embodiment of the present invention. This method can be applied to both terminals and servers. The terminal can be a desktop terminal or a mobile terminal; a mobile terminal can be at least one of a mobile phone, tablet computer, or laptop computer. The server can be a standalone server or a server cluster composed of multiple servers. This embodiment uses a terminal application as an example. Figure 1 The method shown includes the following steps: 101. Determine the first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes several individual model parameters; The inverter model can be a pre-built mathematical model of a VSG-based inverter. The first population is the population before the crossover operation, such as the initial population or the mutated population. Then, the crossover operation is used to further evolve the individual model parameters in the first population to optimize the model parameters of each individual.

[0018] In one feasible implementation, the first population includes a mutant population. Therefore, before determining the first population corresponding to the model parameters of the inverter model at the current iteration number G, the method further includes: determining an initial search space for the model parameters of the VSG-based inverter model, the initial search space reflecting the value range of the model parameters; generating an initial population based on the initial search space and a preset number of individuals Q; the initial population includes Q initial model parameter individuals. Then, determining the first population corresponding to the model parameters of the inverter model at the current iteration number G includes: performing a mutation operation using a preset mutation rule, the current iteration number G, and the initial population to obtain a mutated population.

[0019] In one feasible implementation, the mutation rule includes a preset variable scaling factor algorithm. Then, the step of performing mutation operation using the preset mutation rule, the current iteration number G, and the initial population to obtain a mutated population includes: using the minimum and maximum values ​​of the preset variable scaling factor, the maximum iteration number, G, and the variable scaling factor algorithm to obtain a variable scaling factor; and performing mutation operation based on the variable scaling factor and the initial population to obtain a mutated population.

[0020] Specifically, before the crossover operation, an initial search space for the model parameters of the VSG-based inverter model is determined, and the initial search space is used to reflect the value range of the model parameters. It should be noted that, in order to accurately identify the model parameters of the VSG-based inverter model, it is first necessary to determine the initial search space for the model parameters. This initial search space reflects the value range of the model parameters. Specifically, the initial search space refers to the space formed by the range of values ​​that each variable can take during the evolution process. For a D-dimensional optimization problem, the initial search space is a D-dimensional hyperrectangle, with each dimension corresponding to the value range of a variable.

[0021] The model parameters are the VSG parameters to be identified, such as the inertia coefficient J, damping coefficient D, and droop coefficient Kq. The VSG-based inverter model can be a mathematical model of the inverter built based on a VSG control strategy, such as establishing a VSG-based inverter mathematical model in the MATLAB environment. The inverter can be a photovoltaic inverter, for example. The range of model parameter values ​​can be preset or obtained through analysis of physical relationships, without limitation. For example, the range of the inertia coefficient J can be set to J∈[0.1,5.0], or the influence of key parameters (such as J, D, Kq, etc.) on the inverter output characteristics (such as voltage, frequency, power, etc.) can be studied through theoretical analysis of the established VSG inverter model. Through this analysis, the reasonable variation range of each parameter can be preliminarily determined.

[0022] Determining the initial search space may include: establishing a VSG-based inverter model; obtaining the actual response values ​​of the inverter under different voltage drops; and determining the initial search space based on the VSG-based inverter model and the actual response values.

[0023] For example, a simulation model of the photovoltaic inverter was pre-built in PSCAD software, and a VSG-based mathematical model of the inverter was established in the MATLAB environment. Furthermore, for parameter identification, relevant measurement data needs to be obtained from the simulation model, including the actual response values ​​of the inverter under various operating modes. This application collected actual response values ​​under different voltage drop amplitudes as sample data by setting a certain degree of voltage sag, where the actual response values ​​include voltage, active power, and reactive power data. For example, the data comes from a disturbance experiment of the PSCAD simulation model, with the following specific steps: 1. Set different grid disturbance scenarios in PSCAD (e.g., voltage sag of 20%–50%, frequency fluctuation). 2. Collect dynamic response data of voltage V, active power P, and reactive power Q at the inverter outlet through sensors. 3. Divide the data into three groups: training set (for parameter identification), validation set (for model generalization ability testing), and test set (for final performance evaluation).

[0024] Furthermore, based on the mathematical model and actual response values ​​of the VSG-based inverter, the initial search space for each model parameter is determined, and this space serves as a constraint for subsequent model parameter identification based on the improved differential evolution algorithm. The mathematical models include: the synchronous generator equivalent model, the power control model, and the voltage-current dual closed-loop control model.

[0025] For example, the construction of a photovoltaic inverter simulation model includes: building a simulation model of the photovoltaic inverter in PSCAD software. This model simulates a photovoltaic array connected to a DC bus via a DC / DC converter, then converting the DC current to AC current through a three-phase full-bridge inverter, and finally feeding it to the grid through an L-type filter circuit. The photovoltaic array in the model uses a boost circuit and is controlled with maximum power point tracking as the objective. See [reference needed]. Figure 2 , Figure 2 This is a schematic diagram of a photovoltaic inverter virtual synchronous generator system according to an embodiment of the present invention. Figure 2 In this diagram, PV stands for photovoltaic array, DC / DC is a boost DC-DC converter, VSG is a virtual synchronous generator control unit, PWM is a three-phase full-bridge inverter (PWM control), and L is an L-type filter (inductor filter). vi This is the input modulation wave for the photovoltaic inverter. Lf For filtering inductors, R l , L l They are used to simulate line impedance and inductance, respectively. i abc Let A be the instantaneous current at the output of the photovoltaic inverter, and B be the grid connection point of the photovoltaic power generation system.

[0026] The VSG control strategy implementation includes a virtual synchronous generator control strategy, which enables the inverter to simulate the characteristics of a synchronous generator. The fundamental voltage output by the inverter simulates the internal potential of the synchronous generator, the inductance on the inverter side simulates the synchronous reactance of the synchronous generator, and the three-phase output filter voltage of the inverter simulates the terminal voltage of the synchronous generator. (See also...) Figure 3 , Figure 3 This is a schematic diagram of a virtual synchronous generator control strategy in an embodiment of the present invention. Figure 3 middle, P ref This is a reference value for active power. Q ref This is a reference value for reactive power. Qe This is an estimate of the reactive power output of the photovoltaic inverter. T m For mechanical torque, T e For electromagnetic torque, D p The damping coefficient is... D q The droop coefficient is... K This is the excitation regulator coefficient. s For the Laplace operator, J The moment of inertia of the virtual synchronous generator. ω For the angular frequency of the virtual synchronous generator, ω n The rated angular frequency, θ The angle between the stator and rotor, V ref This is the voltage reference value. V rms This refers to the phase voltage amplitude output by the photovoltaic inverter. M f For the mutual inductance between the stator and rotor, I f This is the rotor excitation current. P com This is an estimated value of the active power output by the photovoltaic inverter. vi This is the input modulation wave for the photovoltaic inverter.

[0027] The mathematical model establishment and parameter identification include: establishing a VSG-based inverter mathematical model in the MATLAB environment so that the subsequent improved differential evolution algorithm can identify the key parameters of the model. These model parameters include at least the inertia coefficient J, mechanical torque, electromagnetic torque Ke, damping coefficient D, angular frequency, reactive power-voltage droop coefficient Kq, voltage reference value Vref, etc. The VSG-based inverter model can be the VSG-based inverter mathematical model itself.

[0028] This allows for the generation of an initial population based on the initial search space and the preset number of individuals Q; the initial population includes Q initial individuals, which are also model parameter individuals, with each individual corresponding to a set of candidate model parameters.

[0029] Understandably, the improved differential evolution algorithm needs to be initialized. Based on the initial search space and the preset number of individuals Q, an initial population is generated, consisting of Q initial individuals X. ij Each individual X ij A set of candidate model parameters is used, each set including at least the inertial time constant J (kg·m²), damping coefficient D (N·m·s / rad), reactive power-voltage droop coefficient Kq (V / Var), voltage reference value Vref (V), and electromagnetic torque coefficient Ke. These parameters are used to construct a complete mathematical model of the VSG inverter and are verified in MATLAB / PSCAD. Through the competition of simulated evolution operations among the various candidate model parameters, the best individual is selected, and the candidate model parameters corresponding to this best individual are the optimal candidate model parameters.

[0030] For example, initialization begins, which may include setting the population size (number of individuals Q), mutation factor F, and crossover rate CR of the differential evolution algorithm (DE). The improved differential evolution algorithm (DE) is combined with parameter identification through the following steps: Parameter encoding: The VSG parameters to be identified (such as inertia coefficient J, damping coefficient D, droop coefficient Kq, etc.) are encoded into individual vectors X=[x1,x2,...,xj], where the value of j is related to the dimension D of the model parameters. For example, if the model parameters include inertia time constant J, damping coefficient D, reactive power-voltage droop coefficient Kq, voltage reference value Vref, and electromagnetic torque coefficient Ke, then the dimension is 5, and D is 5. Each individual X has five dimensions of model parameters, thus forming Q individuals with five dimensions to create the initial population.

[0031] To improve the accuracy of the model parameters, the first population is a mutant population. Therefore, it is necessary to perform a mutation operation on the initial population. Specifically, the mutation operation is performed using the preset mutation rules, the current iteration number G, and the initial population to obtain the mutant population. Furthermore, in order to improve the accuracy of the optimal model parameters, this application does not directly conduct competition between individuals, but instead selects individuals through evolutionary means such as mutation, crossover, and out-of-population competition to obtain the target population, and then selects the best individuals from the target population.

[0032] Specifically, by using a preset mutation rule, the current iteration number G, and the initial population to perform mutation operations, a mutant population is obtained. The mutation rule may include a variable scaling factor algorithm, which calculates a variable scaling factor. The initial population is then mutated using the variable scaling factor to obtain a new population—the mutant population. This achieves the optimization of the model parameters in the initial population through evolutionary mutation.

[0033] In one feasible implementation, the mutation rule includes a preset variable-scale factor algorithm. The step of performing a mutation operation using the preset mutation rule, the current iteration number G, and the initial population to obtain a mutated population includes: using the minimum value F of the preset variable-scale factor. min and maximum value F max Maximum number of iterations G max The variable scaling factor F is obtained by using the G and the variable scaling factor algorithm; a mutation operation is performed based on the variable scaling factor F and the initial population to obtain the mutated population (i.e., the first population).

[0034] It should be noted that in traditional differential evolution algorithms, the scaling factor F is usually fixed, for example, taking a value in the range [0,2]. However, a fixed value of F may not be suitable for different optimization stages. To better balance population diversity and convergence, this invention proposes a method for dynamically adjusting F, the scaling factor adjustment: initially using a large F (e.g., F=0.8) to enhance the global search, and later reducing F (e.g., F=0.2) for a finer search.

[0035] Specifically, the calculation expression for the variable scaling factor algorithm can be defined as: ; Wherein: F min and F max These are the minimum and maximum values ​​of the scaling factor, respectively, which can be adjusted according to the complexity of the problem, for example, F. min =0.5, F max =1.5. G is the current iteration number, G max This represents the maximum number of iterations. The number of iterations indicates how many generations of evolution have occurred, and which generation the individual belongs to at this point.

[0036] Furthermore, mutation can be achieved by selecting two extreme individuals for each individual from among the many individuals in the population. The extreme individuals include the nearest and the least nearest individuals. Then, a comprehensive calculation is performed on the individual and its two extreme individuals based on a variable scaling factor to obtain the mutated individual.

[0037] 102. The first population is subjected to evolutionary processing using a preset multi-point crossover rule to obtain the target population; Furthermore, the first population undergoes crossover evolutionary processing to optimize model parameters. Specifically, this application selects a multi-point crossover rule for the crossover operation. Further, the mutant population is subjected to crossover operation in step 102 to obtain a crossover population, which can be regarded as the target population. The crossover rule may include a preset adaptive crossover probability algorithm. The crossover probability is calculated by the adaptive crossover probability algorithm, and the mutant population is subjected to crossover operation based on the crossover probability to obtain a new population—the crossover population. This achieves further optimization of the model parameters in the mutant population through crossover evolutionary means.

[0038] Specifically, step 102 includes the following steps A01 to A07: A01. Determine the crossover probability for each individual model parameter in the first population; In one feasible implementation, determining the crossover probability corresponding to each model parameter individual in the first population includes: Using the preset minimum and maximum crossover probabilities, the maximum number of iterations, G, and the adaptive crossover probability algorithm, the crossover probability corresponding to each individual model parameter in the first population is obtained.

[0039] For example, the crossover rule includes a preset adaptive crossover probability algorithm. The step of performing a crossover operation on the mutated population using the preset crossover rule to obtain a crossover population includes: using the minimum value of the preset crossover probability... and maximum value Maximum number of iterations The current iteration number G and the adaptive crossover probability algorithm are used to obtain the crossover probability. According to the crossover probability A crossover operation is performed on the mutant population to obtain a crossover population.

[0040] It should be noted that in traditional differential evolution algorithms, the crossover probability... It is fixed. However, the fixed value This approach may not be applicable at different optimization stages. To better balance population diversity and convergence and further improve identification accuracy, this invention improves the crossover operation by proposing an adaptive adjustment method. The method.

[0041] For example, the calculation formula for the adaptive crossover probability algorithm can be expressed as: ; in: and These are the minimum and maximum values ​​of the crossover probability, respectively, and can be adjusted according to the complexity of the problem, for example... , . This is the current iteration number. It represents the maximum number of iterations.

[0042] In one feasible implementation, crossover is a crucial step in differential evolution algorithms, used to increase population diversity by combining the target vector with the trial vector to generate a new offspring population. In this invention, the crossover operation employs an improved strategy to enhance the algorithm's global search capability and convergence speed. Specifically, this application provides a multi-point crossover evolutionary operation, where, in traditional single-point crossover, only one component originates from the trial vector. In multi-point intersection, multiple components can be randomly selected from the test vector.

[0043] For example, performing multi-point crossover operations on the mutant population based on the crossover probability to obtain a crossover population may include the following steps A02 to A07 to implement the multi-point crossover operation: A02. For the target vector corresponding to each model parameter individual in the first population, generate an experimental vector corresponding to each model parameter individual. The vector includes J components, and each component corresponds to a model parameter of one dimension. It should be noted that each individual in the population encodes the J model parameters to be identified (such as the inertia coefficient J, damping coefficient D, droop coefficient Kq, etc.) into an individual vector X=[x1,x2,...,xj], where j belongs to J. The target vector for each individual in the mutant population... Generate the test vector corresponding to each individual. Each component corresponds to a model parameter in one dimension.

[0044] This can be understood as follows: with the increase of the number of iterations, each iteration can obtain the descendants of the initial individual i, where... This can be understood as the target vector of individual i in the Gth generation. This can be understood as the experimental vector of individual i in the Gth generation, where the vector corresponding to each individual has J components.

[0045] A03. Using the preset random number generation rules, generate random numbers corresponding to each component; Furthermore, a corresponding random number is generated for each component. Specifically, a random number is generated for each component using a preset random number generation rule.

[0046] For example, for each target vector Generate the corresponding test vector Determine the current iteration number. and maximum number of iterations Calculate the adaptive crossover probability for the current iteration. .

[0047] The random number generation rule can use the rand algorithm to generate random numbers. Specifically, for each component... ( ), J=D, where, generate a in Uniformly distributed random numbers within a range .

[0048] A04. Using the dimension D of the model parameters as the range of values, randomly select k different integers; Furthermore, random selection A number of distinct integers, in Random selection within the range A number of distinct integers, ensuring that these components come from the test vector. .in, It can be adjusted according to the complexity of the problem, for example or .

[0049] A05. If the target vector contains the random number... Less than or equal to the crossover probability CR, or j equal to the integer. If the j-th component is found in the target vector, then the j-th component in the target vector is replaced with the j-th component in the test vector to obtain the cross vector corresponding to the individual, where j ∈ J and j is an integer; otherwise, if the above conditions are not met, then the cross vector is retained. .

[0050] Crossover operations introduce randomness to the target vector. With test vector Combine them to generate new descendant vectors. .

[0051] After obtaining random numbers and integers, the offspring vector is generated by combining the crossover probability. That is, cross vector Specifically, for each component Generate descendant vectors based on the formula for the crossover operation. : ; in: It is the first The target vector at the th th ... The value of the algebra, It is the first The j-th component of the target vector is in the... The value of the abbreviation.

[0052] It is the first The test vector at the th test vector in ... The value of the algebra, It is the first The j-th component of the test vector is in the... The value of the abbreviation.

[0053] It is the first The descendant vector at the th generation... The value of the algebra, It is the first The j-th component of the descendant vector is in the... The value of the abbreviation.

[0054] Is Random numbers that are uniformly distributed within a range.

[0055] It is the crossover probability, and its typical value range is... It is used to control the intensity of cross operations.

[0056] It is the first of the vectors Each component.

[0057] Is A random integer is selected from the range, ensuring that at least one component comes from the trial vector. .

[0058] A06. Based on the cross vector of the individual model parameter and the target vector, evaluate the fitness to determine the first fitness of the cross vector and the second fitness of the target vector; A07. If the first fitness is better than the second fitness, then the crossover vector is replaced with the target vector to obtain a second population, wherein the target population includes the second population.

[0059] Furthermore, evaluate the descendant vectors: evaluate the generated descendant vectors. Perform an evaluation and calculate its fitness value. Selection operation: Select the descendant vector... With the target vector The vectors with better fitness are compared and selected to enter the next generation of the population.

[0060] Using the cross vectors of each individual With the target vector Fitness comparisons are performed to identify superior individuals. Specifically, fitness is evaluated based on the crossover vector and the target vector of each individual to determine a first fitness of the crossover vector and a second fitness of the target vector. If the first fitness is better than the second fitness, the crossover vector is replaced with the target vector, resulting in a crossover population. Conversely, if the first fitness is not better than the second fitness, the original target vector is retained. This results in a crossover population. The best individuals are then retained from the final crossover population.

[0061] If a higher fitness indicates a better individual, then the first fitness is superior to the second fitness. This is determined by whether the first fitness is greater than the second fitness; if it is, the individual is superior, and if it is not, the individual is not superior. Conversely, if a lower fitness indicates a better individual, then the first fitness is superior to the second fitness. This is determined by whether the first fitness is less than the second fitness; if it is less, the individual is superior, and if it is not, the individual is not superior.

[0062] In other words, the crossover operation proposed in this application employs an improved strategy, which not only increases population diversity but also enhances the global search capability and convergence speed of the differential evolution algorithm, achieving more accurate and efficient solution of model parameters. By introducing adaptive crossover probability and a multi-point crossover strategy, the improved crossover operation can better balance population diversity and convergence. The adaptive crossover probability can dynamically adjust the crossover strength according to different stages of the optimization process, thereby increasing population diversity in the early stages of optimization and improving convergence accuracy in the later stages. The multi-point crossover strategy further increases population diversity, helping to avoid the algorithm getting trapped in local optima. The second population can be the crossover population.

[0063] In one feasible implementation, further, out-of-population competition can be conducted using a preset out-of-population competition rule and the crossover population to obtain a target population. That is, replacing the crossover vector with the target vector yields a second population. This further includes: randomly selecting K model parameter individuals from the second population to obtain a random population, where the random population includes K model parameter individuals, K is less than Q, and Q is the total number of model parameter individuals in the second population; if the random population contains a target model parameter individual with a third fitness greater than the fourth fitness of the best model parameter individual in the second population, then the best model parameter individual in the second population is replaced with the target model parameter individual to obtain a third population, where the target population includes the third population. The third population is the population after the competition.

[0064] Furthermore, by using the preset out-of-population competition rules and the crossover population to conduct out-of-population competition, a target population is obtained. The out-of-population competition rules are then used to conduct competition operations on the mutated population to obtain a new population—the target population. This achieves the optimization of model parameters in the crossover population through out-of-population competition as an evolutionary means.

[0065] It should be noted that during each generation of evolution, a number of individuals are randomly selected from the population to form a "random population," and the fitness of these individuals is compared with that of the individuals in the current evolutionary population. If an individual in the random population is better than the best individual in the current population, then the best individual in the current population is replaced by the random individual. This process helps to break the limitation of local optima and increase global search capabilities.

[0066] Specifically, the crossover population is randomly selected. K individuals are selected to obtain a random population. The random population comprises K individuals, where K is less than Q. If there exists a target individual in the random population with a third fitness greater than the fourth fitness of the best individual in the crossover population, then the best individual in the crossover population is replaced by the target individual to obtain the target population. Otherwise, if the fitness is not greater than the target individual, the original individuals are retained.

[0067] For example, it may specifically include the following process: initialization: In the current population In, randomly select Individuals form a random population. .in It is a relatively small integer, typically ranging from 10% to 20% of the population size Q; For example, if the population size You can take or .

[0068] Fitness assessment: For random populations Each individual in the dataset undergoes a fitness assessment, and its fitness value is calculated. ; At the same time, for the current population Each individual in the process also undergoes a fitness assessment, and its fitness value is calculated. .

[0069] Comparison and Replacement: For random populations Each individual in Its fitness value Compared with the current population The fitness value of the best individual in the system Compare; if Then use replace ; This process can be repeated until all individuals in the random population have been evaluated.

[0070] Update population: After the above comparison and replacement operations, update the current population. ; The process repeats itself as the next generation of evolution begins.

[0071] By introducing out-of-population competition, the algorithm effectively avoids getting trapped in local optima. The introduction of a randomized population increases population diversity, allowing the algorithm to explore more of the solution space during the global search phase. This strategy is particularly suitable for complex optimization problems, especially those prone to getting trapped in local optima during the optimization process.

[0072] 103. Determine the target fitness based on the predetermined actual response value of the inverter, the estimated response value of the inverter under the model parameter individuals in the target population, and the preset criterion function; In one feasible implementation, the response values ​​include the inverter's voltage, active power, and reactive power. Then, determining the target fitness based on the predetermined true response values ​​of the inverter, the estimated response values ​​of the inverter under individual model parameters in the target population, and a preset criterion function includes: determining the errors between the estimated and true response values ​​of the voltage V, active power P, and reactive power Q; and determining the target fitness based on the errors and the preset criterion function Fitness.

[0073] 104. If the target fitness does not meet the preset convergence condition, let G = G + 1, and return to the step of determining the first group corresponding to the model parameters of the inverter model at the current iteration number, where G is the current iteration number; 105. If the target fitness satisfies the preset convergence condition, then the model parameters corresponding to the individual with the best model parameters in the target population are taken as the best identification result.

[0074] Finally, after each generation of evolution, a criterion function is calculated based on the estimated and actual response values. Then, a convergence check is performed: if the criterion function meets the convergence condition, the identification process ends and the model parameters are output; otherwise, the mutation operation is returned.

[0075] The process involves determining the estimated response values ​​of the inverters under the candidate model parameters in the evolved target population, comparing these estimated response values ​​with the actual response values ​​to assess fitness and determine convergence. Fitness is calculated using a criterion function, as follows: ; In the formula, For fitness; This represents the estimated response value for the i-th data point. is the true response value of the i-th data point; N is the number of data points, which is related to the sampling frequency and sampling duration, and is not limited here. , and These are the estimated response values ​​of voltage, active power, and reactive power at the i-th data point, respectively. , and These are the actual response values ​​of voltage, active power, and reactive power at the i-th data point, respectively.

[0076] In this application, all fitness calculations can be obtained using criterion functions, enabling the evaluation of better individuals.

[0077] Specifically, when the fitness function value is less than a preset threshold or the number of iterations reaches the upper limit, the convergence condition is considered met, the evolution is terminated, and the optimal parameters are output. The optimal parameters are the model parameters corresponding to the best individual with the best fitness in the final population. For example, when the fitness value is <1e −4 The iteration may terminate after 500 iterations.

[0078] Finally, parameter identification and model validation: Parameter identification was performed using the second set of sample data, yielding key parameter identification values ​​for the VSG inverter. These identification values ​​were then used to fit the first and third sets of samples, and the goodness of fit between the original response and the identified response was observed. By comparing the fitting curves under different voltage drop amplitudes, the adaptability and practicality of the proposed synchronous inverter model under different sample conditions were verified.

[0079] Parameter stability test: To test the stability of the model parameters, the first set of samples was used for three repeated identifications. The results were very close to the true values ​​and the variance was almost zero, which proved that the model has good parameter stability.

[0080] In summary, the present invention adopts the following technical solution: 1. Construction of photovoltaic inverter simulation model: A photovoltaic inverter simulation model based on VSG technology is constructed in PSCAD software. The photovoltaic array is simulated to be connected to the DC bus through a DC / DC converter, and then the DC current is converted to AC current through a three-phase full-bridge inverter and sent to the grid through an L-type filter circuit.

[0081] 2. VSG Control Strategy Implementation: The VSG control algorithm is adopted to enable power electronic converters, such as inverters, to possess the characteristics of synchronous generators. By calculating the synchronous generator model in real time, the output of the synchronous generator in this state is obtained, and this output is used as a command to control the converter to track the command.

[0082] 3. Mathematical Model Establishment and Parameter Identification: A VSG-based inverter mathematical model is established in the MATLAB environment, and key parameters, including inertia coefficient, mechanical torque, electromagnetic torque, damping coefficient, angular frequency, reactive power-voltage droop coefficient, and voltage reference value, are identified through an improved differential evolution algorithm.

[0083] 4. Algorithm Improvement: The differential evolution algorithm is improved by adjusting the calculation method of the variable scale factor and introducing out-of-group competition operation to improve global search capability and search accuracy.

[0084] 5. Model Validation: By simulating different power grid disturbances, such as voltage drops, the self-descriptive ability, generalization ability, and parameter stability of the established model are verified.

[0085] The photovoltaic inverter modeling and parameter identification method based on VSG technology proposed in this invention can accurately describe the external characteristics of photovoltaic inverters, which helps to promote the modeling and research of distributed generation with VSG. The application of the improved differential evolution algorithm in photovoltaic inverter parameter identification improves the accuracy of parameter identification and the global search capability of the algorithm, which helps to build a more accurate photovoltaic inverter model, thereby improving the stability of the power grid and the efficiency of new energy integration.

[0086] This invention provides a method for inverter parameter identification based on multi-point crossover. The method includes: determining a first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes several individual model parameters; performing evolutionary processing on the first population using a preset multi-point crossover rule to obtain a target population; determining the target fitness based on the predetermined true response value of the inverter, the estimated response value of the inverter under the individual model parameters in the target population, and a preset criterion function; if the target fitness does not meet the preset convergence condition, then let G = G + 1, and return to the step of determining the first population corresponding to the model parameters of the inverter model at the current iteration number, where G is the current iteration number; if the target fitness meets the preset convergence condition, then the model parameter corresponding to the best individual model parameter in the target population is taken as the best identification result. By introducing a multi-point crossover rule to perform evolutionary processing on the inverter model parameter population, the diversity of the population can be effectively enhanced, avoiding getting trapped in local optima, improving the global search capability of parameter identification, accelerating the convergence speed, broadening the scope of problem solving, and making the parameter identification results more reliable and accurate.

[0087] Please see Figure 4 , Figure 4 This is a structural block diagram of an inverter parameter identification device based on multi-point crossover in an embodiment of the present invention, as shown below. Figure 4 The apparatus shown includes: Population determination module 401: used to determine the first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes a number of individual model parameters; Population evolution module 402: used to perform evolutionary processing on the first population using preset multi-point crossover rules to obtain the target population; Fitness determination module 403: used to determine the target fitness based on the predetermined actual response value of the inverter, the estimated response value of the inverter under the model parameter individuals in the target population, and the preset criterion function; Convergence judgment module 404: If the target fitness does not meet the preset convergence condition, then let G=G+1 and return to the step of determining the first group corresponding to the model parameters of the inverter model under the current iteration number, where G is the current iteration number; Parameter determination module 405: If the target fitness satisfies the preset convergence condition, the model parameters corresponding to the best model parameter individual in the target population are taken as the best identification result.

[0088] It should be noted that, Figure 4 The contents of each module in the device shown are... Figure 1 The steps in the method shown are similar, and will not be repeated here to avoid repetition. For details, please refer to [reference needed]. Figure 1 The content of each step in the method shown.

[0089] This invention provides an inverter parameter identification device based on multi-point crossover. The device includes: a population determination module for determining a first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes a plurality of model parameter individuals; a population evolution module for performing evolutionary processing on the first population using a preset multi-point crossover rule to obtain a target population; a fitness determination module for determining a target fitness based on a preset true response value of the inverter, an estimated response value of the inverter at the model parameter individuals in the target population, and a preset criterion function; a convergence judgment module for setting G=G+1 and returning to the step of determining the first population corresponding to the model parameters of the inverter model at the current iteration number if the target fitness does not meet a preset convergence condition, where G is the current iteration number; and a parameter determination module for taking the model parameter corresponding to the best model parameter individual in the target population as the best identification result if the target fitness meets the preset convergence condition. By introducing multi-point crossover rules to perform evolutionary processing on the inverter model parameter population, we can effectively enhance population diversity, avoid getting trapped in local optima, improve the global search capability of parameter identification, accelerate the convergence speed, broaden the scope of problem solving, and make the parameter identification results more reliable and accurate.

[0090] Figure 5 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 5 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. The internal memory may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. Those skilled in the art will understand that… Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0091] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform actions such as... Figure 1 The steps of the method shown.

[0092] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following actions: Figure 1 The steps of the method shown.

[0093] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROG), programmable ROG (PROG), electrically programmable ROG (EPROG), electrically erasable programmable ROG (EEPROG), or flash memory. Volatile memory may include random access memory (RAG) or external cache memory. By way of illustration and not limitation, RAGs are available in a variety of forms, such as static RAG (SRAG), dynamic RAG (DRAG), synchronous DRAG (SDRAG), dual data rate SDRAG (DDRSDRAG), enhanced SDRAG (ESDRAG), synchronous link DRAG (SLDRAG), direct memory bus RAG (RDRAG), direct memory bus dynamic RAG (DRDRAG), and memory bus dynamic RAG (RDRAG), etc.

[0094] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for identifying inverter parameters based on multi-point crossover, characterized in that, The method includes: Determine the first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes several individual model parameters; The first population is subjected to evolutionary processing using a preset multi-point crossover rule to obtain the target population; The target fitness is determined based on the predetermined actual response value of the inverter, the estimated response value of the inverter under the model parameter individuals in the target population, and the preset criterion function. If the target fitness does not meet the preset convergence condition, then let G = G + 1, and return to the step of determining the first group corresponding to the model parameters of the inverter model at the current iteration number, where G is the current iteration number; If the target fitness satisfies the preset convergence condition, then the model parameters corresponding to the individual with the best model parameters in the target population are taken as the best identification result.

2. The method according to claim 1, characterized in that, The step of performing evolutionary processing on the first population using a preset multi-point crossover rule to obtain the target population includes: Determine the crossover probability for each individual model parameter in the first population; For each model parameter individual in the first population, generate an experimental vector for each model parameter individual. The vector includes J components, each component corresponding to a model parameter of one dimension. Using preset random number generation rules, generate random numbers corresponding to each component; Using the dimension of the model parameters as the range of values, randomly select k different integers; If there exists a random number in the target vector that is less than or equal to the crossover probability, or j is equal to the j-th component of the integer, then the j-th component in the target vector is replaced with the j-th component in the test vector to obtain the crossover vector corresponding to the individual model parameter, j∈J, where j is an integer; Fitness evaluation is performed based on the cross vector of the individual and the target vector of the model parameters to determine the first fitness of the cross vector and the second fitness of the target vector; If the first fitness is better than the second fitness, then the crossover vector is replaced with the target vector to obtain a second population, wherein the target population includes the second population.

3. The method according to claim 2, characterized in that, The process of replacing the cross vector with the target vector to obtain the second group further includes: K model parameter individuals are randomly selected from the second population to obtain a random population, which includes K model parameter individuals, where K is less than Q, and Q is the total number of model parameter individuals in the second population; If there exists a target model parameter individual in the random population with a third fitness greater than the fourth fitness of the best model parameter individual in the second population, then the best model parameter individual in the second population is replaced by the target model parameter individual to obtain a third population, wherein the target population includes the third population.

4. The method according to claim 2, characterized in that, Determining the crossover probability for each model parameter individual in the first population includes: Using the preset minimum and maximum crossover probabilities, the maximum number of iterations, G, and the adaptive crossover probability algorithm, the crossover probability corresponding to each individual model parameter in the first population is obtained.

5. The method according to claim 1, characterized in that, The first population includes a variant population. Therefore, the process of determining the first population corresponding to the model parameters of the inverter model at the current iteration number G also includes: Determine the initial search space for the model parameters of the VSG-based inverter model, wherein the initial search space is used to reflect the range of values ​​for the model parameters; Based on the initial search space and the preset number of individuals Q, an initial population is generated; the initial population includes Q individuals with initial model parameters. The first group corresponding to the model parameters of the inverter model at the current iteration number G includes: Mutation operations are performed using preset mutation rules, the current iteration number G, and the initial population to obtain a mutated population.

6. The method according to claim 5, characterized in that, The mutation rule includes a preset variable scaling factor algorithm. The step of performing mutation operations using the preset mutation rule, the current iteration number G, and the initial population to obtain the mutated population includes: The variable scaling factor is obtained by using the preset minimum and maximum values ​​of the variable scaling factor, the maximum number of iterations, G, and the variable scaling factor algorithm. Based on the variable scaling factor and the initial population, a mutation operation is performed to obtain a mutated population.

7. The method according to claim 1, characterized in that, The response values ​​include the inverter's voltage, active power, and reactive power. The step of determining the target fitness based on the predetermined actual response value of the inverter, the estimated response value of the inverter under individual model parameters in the target population, and a preset criterion function includes: Determine the errors between the estimated and actual response values ​​of the voltage, active power, and reactive power, respectively; The target fitness is determined based on the error and the preset criterion function.

8. An inverter parameter identification device based on multi-point crossover, characterized in that, The device includes: Population determination module: used to determine the first population corresponding to the model parameters of the inverter model at the current iteration number G, wherein the population includes a number of individual model parameters; Population evolution module: used to perform evolutionary processing on the first population using preset multi-point crossover rules to obtain the target population; Fitness determination module: used to determine the target fitness based on the predetermined actual response value of the inverter, the estimated response value of the inverter under the model parameter individuals in the target population, and the preset criterion function; Convergence Judgment Module: If the target fitness does not meet the preset convergence condition, then let G=G+1 and return to the step of determining the first group corresponding to the model parameters of the inverter model under the current iteration number, where G is the current iteration number; Parameter determination module: If the target fitness satisfies the preset convergence condition, the model parameters corresponding to the individual with the best model parameters in the target population are taken as the best identification result.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

Images