An intelligent control method for buck converters based on Kriging surrogate model-assisted genetic algorithm

Abstract

The present invention provides an intelligent control method for a buck converter based on a Kriging surrogate model-assisted genetic algorithm, including: Step 1, collecting historical power grid data; Step 2, constructing a Kriging surrogate model; Step 3, evaluating the global optimality of parameter combinations; Step 4, using a genetic algorithm to achieve directional search in the parameter space; Step 5, updating the Kriging surrogate model; Step 6, substituting the updated Kriging surrogate model into the genetic algorithm to implement a closed-loop feedback mechanism; Step 7, iteratively executing Steps 4 to 6 to find the minimum extreme point; Step 8, obtaining the controller parameters corresponding to the optimal individual and the fitness function value under the controller parameters; Step 9, applying the controller parameters corresponding to the optimal fitness to the actual buck converter system. The present invention adds a gradient penalty function to the genetic algorithm to enhance the generalization ability of the updated surrogate model to input voltage fluctuations, and significantly improves the robustness of the system.

Description

Technical Field

[0001] The present invention belongs to the technical field of power electronic converters, and particularly relates to an intelligent control method for a buck converter based on a Kriging surrogate model-assisted genetic algorithm. Background Art

[0002] Surrogate models, also known as approximation models, are a computational method widely used in complex system modeling, simulation, and optimization. They approximate a more complex or difficult-to-directly-compute system using a simplified model, thereby reducing computational costs and time. The Kriging surrogate model constructs an approximate mapping relationship between controller parameters and performance indicators through Gaussian process regression, replacing time-consuming electromagnetic transient simulations, and is particularly suitable for cases where the computational cost is high, there are many parameters, or the system behavior is not easily directly modeled. The genetic algorithm performs parameter search based on the surrogate model and gradually approaches the optimal solution through selection, crossover, and mutation operations. The Buck converter, also known as a step-down converter, is a single-switch non-isolated DC converter with an output voltage lower than the input voltage. Traditional Buck circuit simulations usually require a large amount of computational resources and time, especially when performing optimization design, the computational cost is high. In traditional methods, the surrogate model is only used to replace circuit simulation calculations, and the optimization algorithm mechanically calls the output of the surrogate model. There is a lack of closed-loop feedback and dynamic cooperation between the two. This method has several main problems:

[0003] The surrogate model cannot be dynamically updated based on the new samples of the genetic algorithm, and the model accuracy gradually degrades with iteration. This fragmented cooperation mode makes it difficult to balance optimization efficiency and global convergence.

[0004] The genetic algorithm cannot adjust the search strategy using the prediction variance provided by the surrogate model during the iteration process, resulting in the potential optimal solutions in the high-uncertainty regions being ignored.

[0005] Rigid constraint handling: Traditional methods use a fixed penalty coefficient to handle constraints (such as kp range limitations or non-negativity requirements for performance indicators), and cannot dynamically adjust the penalty intensity, resulting in the optimization results deviating from the actual engineering feasibility.

[0006] To solve the above problems, researchers began to explore combining surrogate models and algorithms to dynamically optimize the control strategy of energy storage converters. Among them, using surrogate models for approximately complex, computationally expensive, or difficult-to-evaluate models can significantly reduce the optimization time, and the combination of evolutionary algorithms has been widely studied due to its advantages in solving complex optimization problems. Evolutionary algorithms can find approximate optimal solutions in multi-objective optimization problems by simulating natural selection and genetic mechanisms. By using surrogate models, the computational requirements for complex physical models or detailed simulation models are avoided, which significantly reduces the computational complexity in each optimization iteration while maintaining sufficient accuracy. However, there are also some challenges when applying the combination of evolutionary algorithms and surrogate models to the optimization of buck converters:

[0007] 1) High-fidelity data acquisition cost: Surrogate models rely on high-quality simulation or experimental data, but the high-frequency switching characteristics of buck converters require fine-grid simulations (such as finite element analysis), and data generation is time-consuming, conflicting with the original intention of reducing computational costs.

[0008] 2) Trade-off of computational efficiency: When optimizing multiple variables such as controller parameters (such as PI gains) and circuit element parameters (inductors, capacitors), the generalization ability of surrogate models in high-dimensional spaces significantly decreases, and it is necessary to balance the model complexity and training cost.

[0009] 3) Limitations of evolutionary algorithms: Traditional evolutionary algorithms may prematurely converge to local optima, especially in multi-modal or high-dimensional problems.

[0010] In summary, the present invention combines a surrogate model and a genetic algorithm to propose an intelligent control method for a buck converter based on a Kriging surrogate model-assisted genetic algorithm. This method not only shortens the optimization time and improves the computational efficiency, but also provides an efficient and accurate control optimization method, reducing the computational burden, which is of great significance for promoting the development of power systems. Summary of the Invention

[0011] Object of the Invention: Aiming at the deficiencies of the prior art, the present invention provides an intelligent control method for a buck converter based on a Kriging surrogate model-assisted genetic algorithm. The method includes the following steps:

[0012] Step 1: Use the optimal Latin hypercube method to collect historical power grid data to form a training set of the model, and preprocess the historical power grid data;

[0013] Step 2: Construct a Kriging surrogate model, and fit the mapping relationship between controller parameters and performance indicators through Gaussian process regression;

[0014] Step 3: Design a genetic algorithm fitness function according to the performance indicators fitted by the Kriging surrogate model;

[0015] Step 4: Embed a gradient penalty function in the genetic algorithm, dynamically adjust the crossover rate and mutation rate according to the uncertainty quantification result of the Kriging surrogate model, achieve directional search in the parameter space, and obtain an improved genetic algorithm;

[0016] Step 5: Iteratively generate new parameters and performance metrics based on the improved genetic algorithm, add them to the training set of the model, and update the Kriging surrogate model;

[0017] Step 6: Utilize the updated Kriging model in combination with the genetic algorithm to implement the closed-loop feedback mechanism of the entire system;

[0018] Step 7: Iteratively execute Step 4 to Step 6 until the termination condition is met and the minimum extreme point is found;

[0019] Step 8: Obtain the controller parameters corresponding to the optimal individual and the fitness function value under the controller parameters;

[0020] Step 9: Apply the controller parameters corresponding to the optimal fitness to the actual buck converter system.

[0021] Step 1 includes: collecting historical grid data through smart meters, sensors, phasor measurement units, and grid management systems, and performing normalization processing; the historical grid data includes the input parameters kp and ki of the buck converter, as well as the output performance metrics, and the output performance metrics include the tracking time st, overshoot ov, and steady-state error sse.

[0022] Step 2 includes:

[0023] Step 2.1: Define the Kriging surrogate model architecture: The preprocessed parameters kp and ki in Step 1 form the input training set S. The input layer of the Kriging surrogate model is used to receive the input training set S, and the output layer is used to predict the tracking time st, overshoot ov, and steady-state error sse. The predicted tracking time st, overshoot ov, and steady-state error sse form the input training set Y. Select the Kriging surrogate model with a Gaussian process regression framework to characterize the nonlinear relationship in the parameter space through the covariance function;

[0024] Step 2.2: Model training: Call the dacefit function of the MATLAB DACE toolbox, input the training set S and the training set Y, set the linear regression function for the metrics ov and sse, set the quadratic polynomial regression function for the metric st, set the Gaussian exponential model for the metrics ov and sse, and set the exponential model for the metric st, and verify the accuracy of the model through the mean square error and root mean square error.

[0025] Step 3 includes: The fitness function is defined as:

[0026] ,

[0027] Among them is the penalty coefficient.

[0028] Step 4 includes:

[0029] Step 4.1, set the initial parameters of the genetic algorithm, including the population size, mutation rate, crossover rate, and number of evolutions, and substitute the performance index fitted by the Kriging surrogate model into the fitness function defined in Step 3;

[0030] Step 4.2, design a gradient penalty function and define the constraint conditions: the ranges of the parameters kp and ki are [0.0001, 0.01] and [100, 200] respectively, and the performance indexes need to satisfy st>0, sse>0, ov>0. Calculate the penalty value to check whether each group of predicted data satisfies the constraint conditions, and calculate the penalty value according to the degree of violation of the constraints of the prediction results. The calculation formula of the penalty value is:

[0031] ,

[0032] Among them is the penalty coefficient, which is determined by sensitivity analysis;

[0033] Step 4.3, update the fitness function value: fitness = fitness + penalty;

[0034] Step 4.4, perform dynamic adjustment using an uncertainty-driven genetic algorithm: the population enters iterative optimization. If the following formula is satisfied, it is determined that the index is in the high-uncertainty region: ;

[0035] Among them is the predicted variance of ov, is the predicted variance of sse, is the predicted variance of st, which is calculated by the Kriging surrogate model. The mutation rate Pm increases from 0.8 to 0.9, and the crossover probability Pc decreases from 0.8 to 0.05. Retain high-potential individuals to enhance the exploration ability;

[0036] If not satisfied, it is determined that the index is in the low-uncertainty region, the mutation rate is 0.01, the crossover rate is 0.2, retain the population diversity, and dynamically update the mutation rate and crossover rate to achieve an adaptive balance between exploration and exploitation;

[0037] Step 4.5, put the initialized population into selection, crossover, and mutation operations for iterative optimization;

[0038] Step 4.6, the population enters the selection operation. Using the roulette wheel method and based on the fitness proportion selection strategy, individuals with higher fitness are selected to form a new population. The selection probability of each individual is inversely proportional to its fitness value.

[0039] Step 4.7, the population enters the crossover operation. Using the real number crossover method, the selected individuals are subjected to the crossover operation to generate new individuals. The crossover points are randomly selected, and the individuals after crossover form a new population.

[0040] Step 5 includes:

[0041] Step 5.1, after the improved genetic algorithm iterates for 2 generations, it regularly calls the real data collected by the buck - type circuit simulation, integrates the new parameters kp, ki and the performance indicators tracking time, overshoot and steady - state error into a new sample buffer to ensure that the model update is based on actual new data.

[0042] Step 5.2, normalize the new parameters kp, ki and performance indicators obtained in Step 5.1, and use the new data to train the Kriging surrogate model to obtain an updated Kriging surrogate model to maintain the timeliness and accuracy of the model. It realizes the dynamic coordination of data generation, model optimization and parameter search.

[0043] Step 6 includes the following steps:

[0044] Step 6.1, fit the performance indicators through the updated Kriging surrogate model and substitute them into the fitness function to obtain the fitness function value. According to the prediction variances of different performance indicators provided by the updated Kriging surrogate model, the high - uncertainty region and low - uncertainty region are divided.

[0045] Step 6.2, combining the high - uncertainty region and low - uncertainty region, call the improved genetic algorithm for directional search, dynamically update the mutation rate and crossover rate, and realize the dynamic coordination mechanism between the Kriging surrogate model and the genetic algorithm.

[0046] Through the above steps, the updated model improves the genetic algorithm, realizes the closed - loop feedback mechanism of the entire system, and forms the dynamic coordination of model optimization, parameter search and data generation.

[0047] Step 7 includes:

[0048] Apply the fitness function obtained in Step 4 to the min function of MATLAB to find the extreme point. The specific implementation code is:

[0049] ,

[0050] where fit is the extreme value and id is the position order of the extreme value.

[0051] The present invention also provides an energy storage converter intelligent control optimization device implemented by using the above method, including: a data acquisition module for randomly collecting historical power grid data; a surrogate model module for fitting the relationship between the parameters of the energy storage converter controller and the performance indicators; a weight coefficient setting module for quantitatively setting the weight coefficients between the performance indicators; a fitness function construction module for constructing the fitness function of the energy storage converter system; and an optimization algorithm module for executing the improved genetic algorithm.

[0052] The present invention also provides an electronic device, including a processor and a memory, where the memory stores program code, and when the program code is executed by the processor, the processor is caused to execute the steps of the above method.

[0053] The present invention has the following technical features:

[0054] 1) Two-way closed-loop regulation of Kriging surrogate model and genetic algorithm:

[0055] According to the characteristics of the surrogate model and the genetic algorithm, different from the traditional method where the surrogate model only statically replaces the simulation, this solution realizes the dynamic interaction of driving each other between the two, forming a closed-loop system for model optimization and parameter search. The genetic algorithm dynamically adjusts the mutation rate and crossover rate of the genetic algorithm according to the prediction results of the surrogate model, and at the same time feeds the newly generated high-value samples back to the Kriging surrogate model for incremental training.

[0056] 2) Dynamic constraint mechanism of gradient penalty function:

[0057] The introduction of the formula optimizes the penalty coefficient , and through sensitivity analysis, and adjusts the degree of constraint violation. Forcing the population to evolve towards the feasible region, while retaining the potential optimal solutions on the boundary, avoiding the overfitting or under-constraint problems caused by fixed penalties.

[0058] 3) Dynamic feedback parameter search:

[0059] The genetic algorithm dynamically divides the exploration and exploitation intervals according to the prediction variance provided by the Kriging surrogate model: in the high-variance interval, increasing the population diversity to generate new samples and improving the local accuracy of the surrogate model; in the low-variance interval, focusing on local search to accelerate convergence.

[0060] In summary, by combining the surrogate model and the genetic algorithm, the present invention provides an effective intelligent control optimization strategy for the buck converter. This strategy not only shortens the optimization time, but also can perform control optimization efficiently and accurately, while reducing the computational burden and avoiding the problem of the optimization process falling into local optima.

[0061] Beneficial effects: The present invention combines the dynamic collaborative advantages of the Kriging surrogate model and the genetic algorithm, and provides an efficient intelligent control optimization method for buck converters. First, based on the high-precision fitting of the Kriging surrogate model to the nonlinear dynamic characteristics of the circuit, the parameter evaluation time required for traditional simulation, which is in the order of minutes, is compressed to seconds. Combining with the global parallel search ability of the genetic algorithm, the design iteration speed is significantly improved. Second, the surrogate model is equivalent to the buck converter, and combined with the genetic algorithm, it ensures that while reducing the consumption of computing resources, the algorithm accurately captures the coupling relationship between the controller parameters and the dynamic performance (overshoot, steady-state error). Through the closed-loop feedback mechanism, the Kriging surrogate model is incrementally updated with the iteration of the genetic algorithm, and the training set is updated after 10 iterations to avoid the problem of accuracy degradation of traditional static surrogate models. On the other hand, the Kriging surrogate model is used to provide uncertain information to guide the genetic algorithm for search, dynamically updating the crossover rate and mutation rate of the genetic algorithm, improving the diversity of the population, and avoiding the omission of potential optimal solutions by traditional algorithms. Finally, a gradient penalty function is added to the genetic algorithm to enhance the generalization ability of the updated surrogate model to input voltage fluctuations, significantly improving the robustness of the system. Description of the Drawings

[0062] Figure 1 is a flowchart of the method of the embodiment of the present invention.

[0063] Figure 2 is an optimization diagram of the control system of the buck converter in the embodiment of the present invention.

[0064] Figure 3 is a prediction comparison diagram of the prediction tracking time of the surrogate model.

[0065] Figure 4 is a prediction comparison diagram of the predicted steady-state error of the surrogate model.

[0066] Figure 5 is a prediction comparison diagram of the predicted overshoot of the surrogate model.

[0067] Figure 6 is a schematic diagram of the comparison of the simulation output voltage tracking between the method of the present invention and the traditional method.

[0068] Figure 7 is a comparison diagram of the running time of the real-time simulation model and the surrogate model after 10 iterations. Detailed Embodiments

[0069] The following further specifically describes the present invention in conjunction with the drawings and specific embodiments, and the above and / or other advantages of the present invention will become clearer.

[0070] As Figure 1As shown in the figure, this embodiment provides an intelligent control method for a buck converter based on a Kriging surrogate model-assisted genetic algorithm, including the following steps:

[0071] Step 1: Use the optimal Latin hypercube method to collect historical power grid data and perform preprocessing, specifically including the following steps:

[0072] Step 1.1: Use the optimal Latin hypercube method to collect historical power grid data and put the collected data into an excel table, including various combinations of fitting the parameters kp and ki of the energy storage converter controller. Among them, kp is the gain coefficient of the proportional control term, which determines the immediate response intensity of the controller to the error. kp mainly affects the dynamic response speed of the system. ki eliminates the steady-state error through the integral action, and the corresponding performance index time (st), overshoot (ov), and steady-state error (sse). The selection of kp and ki is random. The range of kp is [0.0001, 0.01], and the approximate range of ki is [100, 200]. There are a total of 800 groups of data, and the data is normalized.

[0073] Step 2: Build a Kriging surrogate model with two inputs and three outputs, and use the data in Step 1 to fit the relationship between the buck converter parameters and performance indicators, specifically including the following steps:

[0074] Step 2.1: Use MATLAB software to build a Kriging surrogate model in the form of code to fit the relationship between the buck converter parameters and performance indicators, including:

[0075] kp, ki and st;

[0076] kp, ki, and ov;

[0077] kp, ki and sse;

[0078] Step 2.2: In this example, the Kriging surrogate model in the DACE toolbox provided by MATLAB is used. It can be called using the relevant code. The specific formula is as follows:

[0079] ,

[0080] Where S represents the sample data matrix of the input variables, which is the sample matrix composed of kp and ki here, Y represents the sample data matrix of the output variables, which is the matrix composed of performance indicators here, regr represents the handle of the regression polynomial function for fitting the regression relationship between the input variables and the response variables, corr represents the handle of the correlation function for describing the correlation between the input variables, theta represents the parameter vector of the correlation function for adjusting the shape and range of the correlation function, lob represents the lower bound vector of the correlation function parameters, upb represents the upper bound vector of the correlation function parameters, dmodel represents the fitted DACE model that can be used for prediction and interpolation, and perf represents the performance indicators of the fitted model;

[0081] In this embodiment, there are two parameters and three performance indicators, so a Kriging surrogate model with two inputs and three outputs is designed;

[0082] 80% of the total number of the fitting training setting data is used for training and 20% is used for testing. The predicted value of the Kriging surrogate model is:

[0083] ,

[0084] Where (x) represents the mean of the random process, is expressed as a Gaussian process with a mean of zero, x is the input variable, represents the predicted value of the Kriging surrogate model;

[0085] Step 2.3, perform model parameter settings. S is the vector composed of kp and ki, Y is the vector composed of sse, st, and ov, theta = [10 10], lob = [0.0001 100], upb = [0.1 200];

[0086] When performing the fitting training of the controller parameters with ov, sse, and st,

[0087] Set S as the vector composed of kp and ki;

[0088] Y is the vector composed of ov, sse, and st;

[0089] ov and sse are @regpoly1: linear regression;

[0090] st adopts @regpoly2: quadratic polynomial regression;

[0091] ov and sse are @corrgauss: Gaussian model;

[0092] st is @correxp: exponential model;

[0093] Step 2.5: Use 160 sets of independent test set data to verify the prediction accuracy of the surrogate model.

[0094] Step 2.6: Predict the output of the test set, and judge the goodness of fit of the trained Kriging surrogate model through specific evaluation indexes. These indexes include mean absolute error, mean square error, root mean square error, and coefficient of determination, etc. In this example, mean square error and root mean square error are used as evaluation indexes, and their calculation formulas are as follows:

[0095] ,

[0096] ,

[0097] where y1 represents the true value, represents the predicted value, MSE represents the mean square error, RMSE represents the root mean square error. As shown in Table 1 below, it shows the comparison of the root mean square errors of the Kriging surrogate model and the neural network prediction indexes.

[0098] Table 1

[0099] Root Mean Square Error of Kriging Surrogate Model Root Mean Square Error of Neural Network Tracking Time st: 0.000160 Tracking Time st: 0.00004967 Overshoot sse: 0.005370 Overshoot sse: 0.0082589 Steady-State Error ov: 0.025180 Steady-State Error ov: 0.02259

[0100] In this example, the MSE and RMSE of the three prediction models are less than 0.001;

[0101] The three performance indexes in this example are as described in Step 1.1, Figure 3 is the comparison chart of st predicted value and true value, Figure 4 is the comparison chart of sse predicted value and true value, Figure 5 is the comparison chart of ov predicted value and true value. It is found by comparison that the prediction of the surrogate model is close enough to the true simulation result;

[0102] Step 3: Since the existing 800 sets of data cannot fully represent the coupling between the parameters and performance indexes of the buck converter, more data is needed. Therefore, according to the existing surrogate model, fit the relationship between the parameters and performance indexes of the buck converter, quantitatively set the weight coefficients between the performance indexes, and construct a fitness function in the form of:

[0103] ,

[0104] k1 is the weight coefficient of the performance index ov, k2 is the weight coefficient of the performance index sse, k3 is the weight coefficient of the performance index st, where k1 = 50, k = 50, k3 = 50;

[0105] Step 4. Embed a gradient penalty function in the genetic algorithm. Predict the system's performance at different controller parameters and prediction variances through the Kriging surrogate model, and dynamically adjust the crossover rate and mutation rate to achieve directional search in the parameter space. The specific steps are as follows:

[0106] Step 4.1. Set and initialize the parameters of the genetic algorithm. Set the population size to 40, the mutation probability Pm to 0.8, the crossover probability Pc to 0.05, and the number of evolutions to 10. Predict the performance metrics st, ov, and sse of the system under the proportional-integral controller parameter combinations through the trained Kriging surrogate model by fitting.

[0107] Step 4.2. Design a gradient penalty function and define the constraint conditions: the ranges of the parameters kp and ki are 0.0001, 0.01, 100, and 200 respectively, and the performance metrics st > 0, sse > 0, and ov > 0. Analyze the sensitivity of each constraint term through the orthogonal experiment method, and determine the penalty coefficients β = 120 (corresponding to the constraint weight of the overshoot ov), γ = 80 (corresponding to the constraint weight of the steady-state error sse), and δ = 60 (corresponding to the constraint weight of the tracking time st). The calculation formula for the penalty value is:

[0108] ,

[0109] where are the penalty coefficients, determined through sensitivity analysis, which are 120, 80, and 60 respectively here.

[0110] Step 4.3. Update the fitness function value: Fitness = fitness + penalty;

[0111] fitness is the fitness function calculated in Step 3, and penalty is the penalty value calculated in Step 4.2;

[0112] Step 4.4. Use the Kriging surrogate model to fit the prediction variances provided by different metrics, and divide the population into high and low uncertainty regions. If the following formula is satisfied, it is determined that the metric is in the high uncertainty region:

[0113] ,

[0114] is the prediction variance of x, calculated by the Kriging surrogate model. The mutation rate Pm increases from 0.8 to 0.9, and the crossover probability Pc decreases from 0.8 to 0.05. Retain high-potential individuals to enhance the exploration ability; if not satisfied, in the low uncertainty region, the mutation rate is 0.01, the crossover rate is 0.2, retain the population diversity, and dynamically update the mutation rate and crossover rate, and dynamically update the mutation rate and crossover rate;

[0115] Step 4.5: The 40 initialized populations enter the selection, crossover, and mutation operations for iterative optimization;

[0116] Step 4.6: The population enters the selection operation. Using the roulette wheel method and based on the fitness proportion selection strategy, individuals with higher fitness are selected to form a new population, and the selection probability of each individual is inversely proportional to its fitness value;

[0117] Step 4.7: The population enters the crossover operation. Using the real number crossover method, the selected individuals are crossover-operated to generate new individuals. The crossover points are randomly selected, and the individuals after crossover form a new population;

[0118] Step 5: Based on the improved genetic algorithm in Step 4, new parameters and performance indicators are iteratively generated, added to the training set of the model, and the Kriging surrogate model is updated.

[0119] Step 5.1: First, combine the improved genetic algorithm obtained in Step 4. After iterating 2 generations, regularly collect the real data from the buck - type circuit simulation, integrate the new parameters and performance indicators into a new sample buffer to ensure that the model update is based on actual new data.

[0120] Step 5.2: Normalize the new parameters kp, ki and performance indicators (time, overshoot, and steady - state error) obtained in Step 5.1, update the Kriging surrogate model with the new data to obtain an updated surrogate model to maintain the timeliness and accuracy of the model. It realizes the dynamic coordination of data generation, model optimization, and parameter search.

[0121] Step 6: Use the updated surrogate model obtained in Step 5 in combination with the genetic algorithm. Through the closed - loop feedback mechanism, the optimization results are fed back to the training set of the Kriging surrogate model to form a dynamic coordination of model optimization, parameter search, and data generation, which specifically includes the following steps:

[0122] Step 6.1: Substitute the performance indicators fitted by the surrogate model updated in Step 2.3 into the fitness function to obtain the fitness function value. According to the prediction variances of different performance indicators provided by the Kriging surrogate model, high - uncertainty regions and low - uncertainty regions are divided;

[0123] Step 6.2: Combine the high - uncertainty regions and low - uncertainty regions obtained in Step 5.1, call the genetic algorithm for directional search, and dynamically update the mutation rate and crossover rate to realize the dynamic coordination mechanism between the surrogate model and the genetic algorithm.

[0124] Step 7: Use the fitness function obtained in Step 5 and apply the min function in MATLAB to find the extreme point. The specific implementation code is:

[0125] ,

[0126] where fit is the extreme value, id is the position order where the extreme value is located, min is the function name, Fitness is the predicted data set,

[0127] Iteratively execute steps 4 to 7 until the termination condition is met and the smallest extreme point is found, such as reaching the preset number of iterations or the fitness value converges. In each iteration, update the population and record the optimal individual and its fitness value;

[0128] Step 8, result output. Output the controller parameters corresponding to the optimal individual and the fitness function value under this parameter, that is, the extreme point and extreme value of the system performance. In this example, a set of parameters is obtained as kp = 0.03217078, ki = 160.9565. The system voltage output under this set of controller parameters meets the requirements and the next step can be carried out. If not, repeat steps 4 to 7 until it meets the requirements;

[0129] Step 9, result verification and output. Apply the optimal controller parameters to the actual buck converter system. As Figure 6 shown, it represents the schematic diagram of output voltage tracking comparison. As can be seen from Figure 6 it, the performance of this experiment is improved and according to the above steps, the method proposed by the present invention reduces the simulation time. As Figure 7 shown, it represents the time of real simulation and the time of substituting the surrogate model into the genetic algorithm for 40 populations to iterate 10 times. It can be seen that the experimental time of the method proposed by the present invention is less than that of the traditional method.

[0130] As Figure 2 shown, the control system optimization diagram consists of three parts: a buck circuit, a proportional-integral controller, and a genetic algorithm.

[0131] This embodiment also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the intelligent control method for a buck converter based on a Kriging surrogate model-assisted genetic algorithm are implemented.

[0132] This embodiment also provides a computer device, including:

[0133] A memory for storing instructions.

[0134] A processor for executing the instructions, so that the computer device executes the steps of the intelligent control method for a buck converter based on a Kriging surrogate model-assisted genetic algorithm.

[0135] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing device generate means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0136] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0137] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0138] The present invention provides an intelligent control method for a buck converter based on a Kriging surrogate model-assisted genetic algorithm. There are many methods and ways to specifically implement this technical solution. The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art of this technology, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. Each component not clearly defined in this embodiment can be implemented by existing technologies.

Claims

1. A buck converter intelligent control method based on Kriging agent model assisted genetic algorithm, characterized in that: The following steps are involved: Step 1, using the optimal Latin hypercube method to collect historical data of the power grid to form a training set of the model, and preprocessing the historical data of the power grid; Step 2: construct a Kriging proxy model and fit the mapping relationship between controller parameters and performance indicators through Gaussian process regression; Step 3, designing the fitness function of the genetic algorithm according to the performance index of the Kriging proxy model fitting; Step 4: embed the gradient penalty function in the genetic algorithm, dynamically adjust the crossover rate and mutation rate according to the uncertainty quantification result of the Kriging proxy model, realize the directional search of the parameter space, and obtain the improved genetic algorithm; Step 5, iteratively generate new parameters and performance indicators based on the improved genetic algorithm, add them to the training set of the model, and update the Kriging proxy model; Step 6, using the updated Kriging model combined with the genetic algorithm to implement a closed-loop feedback mechanism for the entire system; Step 7, iteratively execute steps 4 to 6 until the termination condition is met and the minimum extreme point is found; Step 8, obtaining controller parameters corresponding to the optimal individual and the fitness function value under the controller parameters; Step 9, applying the controller parameters corresponding to the optimal fitness to the actual buck converter system; Step 1 includes: collecting historical grid data through smart meters, sensors, phasor measurement units and grid management systems, and performing normalization processing; the historical grid data includes input parameters kp and ki of the buck converter, and output performance indicators, and the output performance indicators include tracking time st, overshoot ov and steady-state error sse; Step 2 includes: Step 2.1, define the Kriging proxy model architecture: the parameters kp and ki preprocessed in step 1 constitute the input training set S. The input layer of the Kriging proxy model is used to receive the input training set S. The output layer is used to predict the tracking time st, overshoot ov, and steady-state error sse. The predicted tracking time st, overshoot ov, and steady-state error sse constitute the input training set Y. The Kriging proxy model of the Gaussian process regression framework is selected, and the nonlinear relationship of the parameter space is characterized by the covariance function. Step 2.2, model training: call the dacefit function of the MATLAB DACE toolbox, input the training set S and the training set Y, set the linear regression function for the indicators ov and sse, set the quadratic polynomial regression function for the indicator st, set the Gaussian exponential model for the indicators ov and sse, set the exponential model for the indicator st, and verify the accuracy of the model through the mean square error and root mean square error; Step 3 includes: The fitness function fitness is defined as: fitness=k1*ov+k2*sse+k3*st, Where k1, k2 and k3 are weight coefficients; Step 4 includes: Step 4.1, set the initial parameters of the genetic algorithm, including population size, mutation rate, crossover rate and number of evolutions, and substitute the performance index of the Kriging proxy model into the fitness function defined in step 3; Step 4.2, design the gradient penalty function and define the constraints: the ranges of parameters kp and ki are [0.0001, 0.01], [100, 200] respectively, the performance indicators need to satisfy st>0, sse>0, ov>0, calculate the penalty value to check whether each set of predicted data meets the constraints, and calculate the penalty value according to the degree of violation of the constraints of the prediction results. The calculation formula of the penalty value penalty is: penalty=β*max(0,-ov) 2 +γ*max(0,-sse) 2 +δ*max(0,-st) 2 , Among them, β, γ, δ are penalty coefficients; Step 4.3, update the fitness function value: fitness = fitness + penalty; Step 4.4, use uncertainty-driven genetic algorithm for dynamic adjustment: the population enters iterative optimization, and if the following formula is satisfied, the indicator is determined to be in the high uncertainty region: max(δ 2 (ov),δ 2 (sse),δ 2 (st))>0.1; where δ 2 (ov) is the prediction variance of ov, δ 2 (sse) is the prediction variance of sse, δ 2 (st) is the prediction variance of st, calculated by the Kriging proxy model, the mutation rate Pm increases from 0.8 to 0.9, and the crossover probability Pc decreases from 0.8 to 0.05; If it is not satisfied, the judgment index is in the low uncertainty region, with a mutation rate of 0.01 and a crossover rate of 0.2; Step 4.5, the initialized population is put into selection, crossover and mutation operations to iteratively search for the best result; Step 4.6, the population enters the selection operation, using the roulette method, based on the fitness ratio selection strategy, to select individuals with high fitness to form a new population, and the selection probability of each individual is inversely proportional to the fitness value; Step 4.7, the population enters the crossover operation, and the real number crossover method is used to perform a crossover operation on the selected individuals to generate new individuals. The crossover points are randomly selected, and the individuals after the crossover form a new population.

2. The method according to claim 1, characterized in that Step 5 includes: Step 5.1, after the improved genetic algorithm iterates for 2 generations, it regularly calls the real data collected by the buck circuit simulation, and integrates the new parameters kp, ki and the performance indicators tracking time, overshoot and steady-state error into the new sample buffer; Step 5.2, normalize the new parameters kp, ki and performance indicators obtained in step 5.1, use the new data to train the Kriging proxy model, and obtain the updated Kriging proxy model.

3. The method according to claim 2, characterized in that Step 6 includes the following steps: Step 6.1, fit the performance index through the updated Kriging proxy model, and bring it into the fitness function to obtain the fitness function value, and divide the high uncertainty area and low uncertainty area according to the prediction variance of different performance indicators provided by the updated Kriging proxy model; Step 6.2, combining the high uncertainty area and the low uncertainty area, calling the improved genetic algorithm for directional search, dynamically updating the mutation rate and crossover rate, and realizing the dynamic coordination mechanism of the Kriging agent model and the genetic algorithm.

4. The method according to claim 3, characterized in that Step 7 includes: Use the MATLAB min function to find the extreme point of the fitness function obtained in step 4. The specific implementation code is: [fit,id]=min(Finness), Among them, fit is the extreme value, and id is the position order of the extreme value.

5. An intelligent control optimization device for energy storage converter implemented by the method according to any one of claims 1 to 4, characterized in that: include: Data acquisition module, used to randomly collect historical data of power grid; A proxy model module is used to fit the relationship between the controller parameters and performance indicators of the energy storage converter; The weight coefficient setting module is used to quantitatively set the weight coefficients between various performance indicators; the fitness function construction module is used to construct the fitness function of the energy storage converter system; and the optimization algorithm module is used to execute the improved genetic algorithm.

6. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores program codes, and when the program codes are executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 4.

Images