Multi-condition calibration method for simulation parameters of electromechanical systems

Abstract

This invention provides a multi-condition calibration method for simulation parameters of electromechanical systems. The method includes: constructing multiple objective functions with the objective of maximizing the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each operating condition; performing multi-objective optimization on the multiple objective functions and updating the empirical knowledge base based on the multi-objective optimization results; performing non-dominated sorting and clustering on each solution in the empirical knowledge base, and sorting each subspace obtained by clustering according to its crowding degree to determine multiple target subspaces; constructing a surrogate model in each target subspace and performing Bayesian optimization, and updating the external archive and empirical knowledge base of the multi-objective optimization based on the Bayesian optimization results; and iterating based on the updated external archive and empirical knowledge base to obtain the multi-condition calibration results of the simulation parameters. This invention can reduce computational costs and improve computational efficiency, while simultaneously improving the versatility, robustness, and prediction accuracy of the electromechanical system simulation model.

Description

Technical Field

[0001] This invention relates to the field of simulation model calibration technology, and in particular to a multi-condition calibration method for simulation parameters of electromechanical systems. Background Technology

[0002] Electromechanical systems are widely used in manufacturing, automotive, aerospace, medical equipment, consumer electronics, and energy systems. They represent a cross-disciplinary integration of mechanical engineering, electronic engineering, control engineering, computer science, and information technology, characterized by high integration. Due to this high integration, functional and performance testing of electromechanical systems often requires the use of simulation models. Therefore, establishing accurate simulation models for electromechanical systems is essential.

[0003] The essence of an electromechanical system simulation model is an approximation of the actual electromechanical system. In order to establish an accurate electromechanical system simulation model, the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system is usually evaluated first. When the consistency is poor, the simulation parameters of the electromechanical system simulation model are adjusted and the simulation data is obtained again. Then, the consistency evaluation is performed again until the simulation data and the measured data reach a high degree of consistency. This process is called the simulation parameter calibration of the electromechanical system simulation model.

[0004] During the calibration of simulation parameters for electromechanical system simulation models, the inventors discovered that the performance of actual electromechanical systems typically varies under different operating conditions. However, traditional simulation parameter calibration methods are usually based on a single operating condition. This leads to similar performance patterns across multiple operating conditions when the calibrated electromechanical system simulation model is used for experimental verification, which is inconsistent with the actual performance of the actual electromechanical system. In other words, traditional simulation parameter calibration methods for electromechanical system simulation models have low accuracy and adaptability under multiple operating conditions. Moreover, because multi-operating-condition scenarios are characterized by high dimensionality, large data volume, and a wide and complex solution space for simulation parameters, calibrating simulation parameters of electromechanical system simulation models based on multiple operating conditions suffers from high computational cost and low efficiency. Summary of the Invention

[0005] This invention provides a multi-condition calibration method for simulation parameters of electromechanical systems, which solves the problem that it is difficult to simultaneously consider the accuracy, adaptability, computational cost and computational efficiency of simulation models when calibrating simulation parameters.

[0006] In a first aspect, embodiments of the present invention provide a multi-condition calibration method for simulation parameters of an electromechanical system, comprising:

[0007] To achieve the highest possible consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each working condition, multiple objective functions are constructed.

[0008] Each simulation parameter of the electromechanical system simulation model is used as a decision vector to perform multi-objective optimization on multiple objective functions, and the experience knowledge base is updated based on the multi-objective optimization results.

[0009] For each solution of the simulation parameters of the corresponding electromechanical system simulation model in the experience knowledge base, perform non-dominated sorting to obtain the Pareto optimal solution set and other solution sets;

[0010] Cluster the Pareto optimal solution set and the other solution sets respectively, and denote each sub-solution set in the clustering result as a subspace;

[0011] The subspaces are sorted according to their crowding level, and multiple target subspaces are determined from each subspace based on the sorting results.

[0012] Within each target subspace, a surrogate model is constructed for each target function. Based on each surrogate model, Bayesian optimization is performed using the desired hypervolume improvement as the acquisition function. The external archive of multi-objective optimization and the empirical knowledge base are updated based on the Bayesian optimization results.

[0013] Based on the updated external archive and the experience knowledge base, the step of "multi-objective optimization of multiple objective functions" and subsequent steps are re-executed until the termination condition is met. Then, the simulation parameters of the electromechanical system simulation model under multiple operating conditions are calibrated based on the global optimal solution of the multi-objective optimization.

[0014] In one possible implementation, each simulation parameter of the electromechanical system simulation model is used as a decision vector, and multi-objective optimization is performed on multiple objective functions. The empirical knowledge base is then updated based on the multi-objective optimization results, including:

[0015] Using each simulation parameter of the electromechanical system simulation model as a decision vector, multi-objective optimization is performed on multiple objective functions to obtain the external archive after multi-objective optimization.

[0016] The experience knowledge base is updated based on the external archives.

[0017] In one possible implementation, the Pareto optimal solution set and the other solution sets are clustered separately, and each sub-solution set in the clustering result is denoted as a subspace, including:

[0018] When the multi-objective optimization result is obtained, the number of iterations corresponding to the multi-objective optimization is obtained;

[0019] Based on the number of iterations and the maximum number of iterations for multi-objective optimization, determine the number of the first cluster corresponding to the Pareto optimal solution set and the number of the second cluster corresponding to the other solution sets;

[0020] The Pareto optimal solution set is clustered according to the first cluster number, and the other solution sets are clustered according to the second cluster number. Each sub-solution set in the clustering result is denoted as a subspace.

[0021] In one possible implementation, determining the number of first clusters corresponding to the Pareto optimal solution set and the number of second clusters corresponding to the other solution sets, based on the number of iterations and the maximum number of iterations in the multi-objective optimization, includes:

[0022] Calculate the ratio of the number of iterations to the maximum number of iterations for multi-objective optimization;

[0023] The ratio is compared with the first threshold and the second threshold, respectively;

[0024] Based on the comparison results, determine the number of the first cluster corresponding to the Pareto optimal solution set and the number of the second cluster corresponding to the other solution sets.

[0025] In one possible implementation, after determining the number of the first clusters corresponding to the Pareto optimal solution set and the number of the second clusters corresponding to the other solution sets, the method further includes:

[0026] Determine whether the number of the first clusters is less than the number of solutions in the Pareto optimal solution set, and determine whether the number of the second clusters is less than the number of solutions in the other solution sets;

[0027] If the number of the first clusters is less than the number of solutions in the Pareto optimal solution set, then the Pareto optimal solution set is clustered according to the number of the first clusters.

[0028] If the number of the first clusters is greater than or equal to the number of solutions in the Pareto optimal solution set, then the Pareto optimal solution set is clustered according to the number of solutions in the Pareto optimal solution set.

[0029] If the number of the second clusters is less than the number of solutions in the other solution sets, then the other solution sets are clustered according to the number of the second clusters.

[0030] If the number of the second clusters is greater than or equal to the number of solutions in the other solution sets, then the other solution sets are clustered according to the number of solutions in the other solution sets.

[0031] In one possible implementation, the subspaces are sorted according to their crowding level, and multiple target subspaces are determined from the subspaces based on the sorting results, including:

[0032] The subspaces are sorted according to their crowding level to obtain a sorting result;

[0033] The number of iterations corresponding to the multi-objective optimization when the multi-objective optimization result is obtained is obtained, and the dynamic congestion threshold is determined based on the number of iterations and the maximum number of iterations of the multi-objective optimization.

[0034] From the sorting results, subspaces with a crowding degree greater than the dynamic crowding degree threshold are selected to obtain multiple target subspaces.

[0035] In one possible implementation, subspaces with a crowding degree greater than the dynamic crowding degree threshold are filtered from the sorting results to obtain multiple target subspaces, including:

[0036] From the sorting results, select subspaces with a crowding degree greater than the dynamic crowding degree threshold as candidate subspaces;

[0037] Based on the crowding degree of the candidate subspaces, candidate subspaces with similar crowding degrees are merged to obtain multiple target subspaces.

[0038] Secondly, embodiments of the present invention provide a multi-condition calibration device for simulation parameters of electromechanical systems, comprising:

[0039] The objective function construction module is used to construct multiple objective functions with the goal of maximizing the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each working condition.

[0040] The first optimization module is used to perform multi-objective optimization on multiple objective functions by taking each simulation parameter of the electromechanical system simulation model as a decision vector, and update the experience knowledge base based on the multi-objective optimization results.

[0041] The non-dominated sorting module is used to perform non-dominated sorting on each solution of the simulation parameters of the corresponding electromechanical system simulation model in the experience knowledge base to obtain the Pareto optimal solution set and other solution sets.

[0042] The first subspace partitioning module is used to cluster the Pareto optimal solution set and the other solution sets respectively, and to record each sub-solution set in the clustering result as a subspace.

[0043] The second subspace partitioning module is used to sort each subspace according to the crowding degree of each subspace, and determine multiple target subspaces from each subspace according to the sorting result;

[0044] The second optimization module is used to construct a surrogate model for each objective function in each objective subspace, perform Bayesian optimization based on each surrogate model with the desired hypervolume improvement as the acquisition function, and update the external archive of multi-objective optimization and the empirical knowledge base based on the Bayesian optimization results.

[0045] The cyclic calibration module is used to re-execute the step of "multi-objective optimization of multiple objective functions" and subsequent steps based on the updated external archive and the experience knowledge base, until the termination condition is met, and obtain the multi-condition calibration result of the simulation parameters of the electromechanical system simulation model based on the global optimal solution of the multi-objective optimization.

[0046] Thirdly, embodiments of the present invention provide a terminal, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory to perform the steps of the method as described in the first aspect or any possible implementation thereof.

[0047] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method as described in the first aspect or any possible implementation thereof.

[0048] This invention provides a multi-condition calibration method for electromechanical system simulation parameters. The method first aims to maximize the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each condition. Multiple objective functions are then constructed, and each simulation parameter of the electromechanical system simulation model is used as a decision vector to perform multi-objective optimization on these objective functions. The empirical knowledge base is updated based on the multi-objective optimization results, allowing population optimization information to be added to the empirical knowledge base. Based on this, by performing non-dominated sorting, clustering, and crowding sorting on each solution in the empirical knowledge base, the empirical knowledge base can be divided into multiple objective subspaces when the optimal solution cannot be determined under multiple operating conditions. Then, within each objective subspace, a surrogate model is constructed for each objective function and Bayesian optimization is performed. The external archive of multi-objective optimization is updated based on the Bayesian optimization results, thereby guiding multi-objective optimization based on each surrogate model. Through the synergistic optimization of multi-objective optimization and Bayesian optimization, the computational cost of simulation parameter calibration under multiple operating conditions is reduced and the computational efficiency is improved. Furthermore, the optimized and calibrated electromechanical system simulation model can output more accurate simulation data under various operating conditions, improving the versatility, robustness, and prediction accuracy of the electromechanical system simulation model. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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.

[0050] Figure 1This is a flowchart illustrating the implementation of the multi-condition calibration method for electromechanical system simulation parameters provided in this embodiment of the invention.

[0051] Figure 2 This is a block diagram illustrating the implementation of the multi-condition calibration method for electromechanical system simulation parameters provided in this embodiment of the invention.

[0052] Figure 3 This is a schematic diagram of the structure of the multi-condition calibration device for electromechanical system simulation parameters provided in this embodiment of the invention;

[0053] Figure 4 This is a schematic diagram of the terminal provided in an embodiment of the present invention. Detailed Implementation

[0054] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0055] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments will be described below in conjunction with the accompanying drawings.

[0056] See Figure 1 The document illustrates a flowchart of the implementation of the multi-condition calibration method for electromechanical system simulation parameters provided in this embodiment of the invention, which is described in detail below:

[0057] In step 101, multiple objective functions are constructed with the goal of maximizing the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each working condition.

[0058] For example, assuming there are K working conditions, we can construct K objective functions, each of which can be: Where k = 1, 2, ..., K, ω k,i Let s be the weighting coefficient of the i-th indicator under the k-th working condition. k,i Let be the consistency evaluation score of the i-th indicator under the k-th working condition.

[0059] In step 102, each simulation parameter of the electromechanical system simulation model is used as a decision vector to perform multi-objective optimization on multiple objective functions, and the experience knowledge base is updated based on the multi-objective optimization results.

[0060] For example, each simulation parameter of the electromechanical system simulation model can be used as a decision vector to perform multi-objective optimization on multiple objective functions, resulting in an external archive after multi-objective optimization; the experience knowledge base can then be updated based on the external archive.

[0061] Combination Figure 2 As shown, the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm can be used for multi-objective optimization.

[0062] When performing multi-objective optimization based on MOPSO, initialization is performed first, including initializing the population size, velocity and position of each particle, the particle's historical best solution, external archive, global best solution, and experience knowledge base.

[0063] The initial global optimal solution can be randomly selected from the initial external archive or selected according to specific rules.

[0064] Then, based on the initial particle velocity and position, combined with the velocity update formula and the position update formula, the particle velocity and position are updated using the initial particle historical optimal solution and the initial global optimal solution, and the corresponding fitness value of the particle after the update is calculated, that is, the values ​​of multiple objective functions corresponding to the particle after the update.

[0065] Based on the fitness value corresponding to the updated particle, the historical optimal solution, external archive, and global optimal solution can be updated.

[0066] The experience knowledge base can be updated based on the updated external archive, thereby adding the optimization history information obtained from each multi-objective optimization to the experience knowledge base, so that subsequent exploration of each subspace can be carried out more deeply based on the experience knowledge base, ensuring the diversity of solutions and global coverage.

[0067] In step 103, each solution of the simulation parameters of the corresponding electromechanical system simulation model in the experience knowledge base is non-dominated and sorted to obtain the Pareto optimal solution set and other solution sets.

[0068] Non-dominated sorting is a sorting method in the field of multi-objective optimization. In multi-objective optimization problems, there is usually no single solution that can simultaneously optimize all objective functions. Therefore, it is necessary to find a set of solutions that are better than other solutions on one objective and at least no worse on others. Such solutions are called Pareto optimal solutions or non-dominated solutions.

[0069] For example, if solution A is no worse than solution B on all objectives and is better than solution B on at least one objective, then solution A is said to be a non-dominated solution to B. Based on this, each solution to the simulation parameters of the corresponding electromechanical system simulation model in the empirical knowledge base can be non-dominated to obtain the Pareto optimal solution set and other solution sets.

[0070] In step 104, the Pareto optimal solution set and other solution sets are clustered separately, and each sub-solution set in the clustering result is denoted as a subspace.

[0071] In this embodiment, based on multi-objective optimization, considering that the objective function of the electromechanical system simulation model is usually not differentiable, a surrogate model is constructed to assist in multi-objective optimization. Simultaneously, considering that as the experience knowledge base is updated, the number of samples in the experience knowledge base gradually increases, leading to a sharp increase in the computational cost and complexity of constructing the surrogate module, a method is considered to reduce the computational cost and complexity of constructing the surrogate model by dividing the experience knowledge base into Pareto optimal solution sets and other solution sets, and then further dividing the Pareto optimal solution sets and other solution sets into subspaces.

[0072] In this embodiment, due to the multi-objective optimization, it is difficult to determine the optimal solution that optimizes multiple objectives in both the Pareto optimal solution set and other solution sets. Therefore, it is difficult to partition the Pareto optimal solution set and other solution sets layer by layer based on the optimal solution. Thus, clustering is considered to partition the Pareto optimal solution set and other solution sets according to their similarity to form subspaces. For example, as shown... Figure 2 As shown, K-Means clustering can be used to cluster the Pareto optimal solution set and other solution sets respectively, and each sub-solution set in the clustering result is denoted as a subspace.

[0073] This embodiment uses clustering to divide the Pareto optimal solution set and other solution sets into multiple subspaces, which helps to distinguish different optimization regions and provide focused search opportunities for each region.

[0074] In step 105, the subspaces are sorted according to their crowding level, and multiple target subspaces are determined from each subspace based on the sorting results.

[0075] In this embodiment, after obtaining multiple subspaces, in order to effectively identify key regions in the solution space, the key regions are prioritized for in-depth optimization. Combining the characteristic that the solutions in the Pareto optimal solution set are usually more sparse than the solutions in other solution sets, these subspaces can be sorted according to their crowding (i.e., sparsity) to ensure that the subspaces of the Pareto optimal solution set are given priority in the subsequent optimization process.

[0076] By identifying multiple target subspaces in this embodiment, it is possible to effectively distinguish and prioritize the processing of subspaces of Pareto optimal solution sets, ensuring that high-quality solutions are explored first. At the same time, it reduces the training time of subsequent surrogate models and improves parallel processing capabilities. This not only optimizes the exploration efficiency of the solution space, but also improves the overall solution efficiency through localization, significantly enhancing the performance of multi-condition optimization.

[0077] In step 106, within each objective subspace, a surrogate model is constructed for each objective function. Based on each surrogate model, Bayesian optimization is performed using the desired hypervolume improvement as the acquisition function. The external archive and empirical knowledge base of multi-objective optimization are updated based on the Bayesian optimization results.

[0078] Combination Figure 2 After dividing the subspace and determining the target subspace, the expected hypervolume improvement (EHI) can be used as the acquisition function in each target subspace to guide the search process and optimize the surrogate model by maximizing the increment of the Pareto hypervolume.

[0079] EHI measures the average increase in Pareto front hypervolume that may result from selecting a new sampling point. The hypervolume index is a performance metric for multi-objective optimization, comprehensively reflecting the quality and distribution of the Pareto front. The hypervolume index refers to the volume of the region between the Pareto front and a reference point; a larger hypervolume index indicates that the Pareto front is closer to the true optimal solution.

[0080] In this embodiment, a surrogate model is constructed for each objective function within each objective subspace. Based on each surrogate model, Bayesian optimization is performed using the expected hypervolume improvement as the sampling function. The expected hypervolume improvement is then evaluated based on the surrogate model, further optimizing the Pareto front. By using EHI based on a dynamically adjusted sampling strategy, it can be ensured that new sampling points can effectively expand the Pareto optimal solution set, while improving the overall optimization efficiency.

[0081] In step 107, based on the updated external archive and experience knowledge base, the step of "multi-objective optimization of multiple objective functions" and subsequent steps are re-executed until the termination condition is met. Then, the simulation parameters of the electromechanical system simulation model are calibrated under multiple operating conditions based on the global optimal solution of the multi-objective optimization.

[0082] Combination Figure 2 After updating the external archive and experience knowledge base of multi-objective optimization based on the Bayesian optimization results, it can continuously introduce potential new points into the Pareto optimal solution set through iterative sampling and updating, thereby effectively promoting the search for the global optimal solution, improving the calibration efficiency and effect of the electromechanical system simulation model under multiple operating conditions, overcoming the limitations of traditional simulation parameter calibration methods in multi-objective optimization scenarios, and significantly improving the global search capability and solution quality of the optimization process.

[0083] This invention first aims to maximize the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each operating condition. Multiple objective functions are constructed, and each simulation parameter of the electromechanical system simulation model is used as a decision vector. Multi-objective optimization is performed on these objective functions, and the empirical knowledge base is updated based on the optimization results, incorporating population optimization information. Furthermore, by performing non-dominated sorting, clustering, and crowding sorting on each solution in the empirical knowledge base, the base can be divided into multiple objective subspaces when the optimal solution cannot be determined under multiple operating conditions. Within each subspace, a surrogate model is constructed for each objective function and Bayesian optimization is performed. The external archive of the multi-objective optimization is updated based on the Bayesian optimization results. This guides multi-objective optimization based on the surrogate models. Through the synergistic optimization of multi-objective optimization and Bayesian optimization, the computational cost of simulation parameter calibration under multiple operating conditions is reduced, and computational efficiency is improved. Moreover, the optimized and calibrated electromechanical system simulation model can output more accurate simulation data under various operating conditions, improving the model's versatility, robustness, and prediction accuracy.

[0084] In one embodiment, clustering is performed on the Pareto optimal solution set and other solution sets respectively, and each sub-solution set in the clustering result is denoted as a subspace, which may include:

[0085] When the multi-objective optimization result is obtained, the number of iterations corresponding to the multi-objective optimization is obtained; based on the number of iterations and the maximum number of iterations of the multi-objective optimization, the number of the first cluster corresponding to the Pareto optimal solution set and the number of the second cluster corresponding to other solution sets are determined; the Pareto optimal solution set is clustered according to the number of the first cluster, and other solution sets are clustered according to the number of the second cluster, and each sub-solution set in the clustering result is recorded as a subspace.

[0086] In this embodiment, considering that the number of samples in the experience knowledge base gradually increases with the iterative process of multi-objective optimization, to avoid the overall multi-objective optimization process getting stuck in local optima and to accelerate the convergence speed, the current optimization stage can be evaluated by the current iteration number and the maximum iteration number of multi-objective optimization. Different numbers of clusters are determined for the Pareto optimal solution set and other solution sets at different optimization stages. For example, in the early stages of optimization, a relatively small number of clusters is determined for the Pareto optimal solution set, and a relatively large number of clusters is determined for other solution sets, thereby obtaining more subspaces on other solution sets to expand the search scope in the early stages of optimization. In the later stages of optimization, a relatively large number of clusters can be determined for the Pareto optimal solution set, and a relatively small number of clusters can be determined for other solution sets, thereby reducing the overall number of subspaces and obtaining more subspaces on the Pareto optimal solution set, thus strengthening the search for the Pareto optimal solution set and focusing on subspaces with higher confidence.

[0087] For example, the ratio of the number of iterations to the maximum number of iterations in multi-objective optimization can be calculated; the ratio can be compared with a first threshold and a second threshold respectively; and the number of first clusters corresponding to the Pareto optimal solution set and the number of second clusters corresponding to other solution sets can be determined based on the comparison results.

[0088] In this embodiment, the ratio of the number of iterations to the maximum number of iterations can be calculated and compared with a first threshold and a second threshold to determine whether the current multi-objective optimization is in the early stage, middle stage, or late stage of the optimization phase. Then, different numbers of clusters can be determined based on the different optimization phases of the current multi-objective optimization.

[0089] The first and second thresholds can be selected according to the actual situation, and this embodiment does not limit them. For example, the first threshold can be 1 / 2 and the second threshold can be 2 / 3. Or, the first threshold can be 2 / 3 and the second threshold can be 4 / 5, and so on.

[0090] For example, after determining the number of the first clusters corresponding to the Pareto optimal solution set and the number of the second clusters corresponding to other solution sets, the method may further include:

[0091] Determine whether the number of clusters in the first cluster is less than the number of solutions in the Pareto optimal solution set, and determine whether the number of clusters in the second cluster is less than the number of solutions in the other solution sets.

[0092] If the number of clusters in the first cluster is less than the number of solutions in the Pareto optimal solution set, then the Pareto optimal solution set is clustered according to the number of clusters in the first cluster.

[0093] If the number of the first cluster is greater than or equal to the number of solutions in the Pareto optimal solution set, then the Pareto optimal solution set is clustered according to the number of solutions in the Pareto optimal solution set.

[0094] If the number of solutions in the second cluster is less than the number of solutions in other solution sets, then the other solution sets are clustered based on the number of solutions in the second cluster.

[0095] If the number of solutions in the second cluster is greater than or equal to the number of solutions in other solution sets, then the other solution sets are clustered according to the number of solutions in the other solution sets.

[0096] In this embodiment, the case where the number of samples in the experience knowledge base is small is further considered. In this case, the number of solutions in the corresponding Pareto optimal solution set and the number of solutions in other solution sets are also relatively small. In order to fully explore the Pareto optimal solution set and other solution sets, when performing clustering in the initial stage, each solution in the Pareto optimal solution set can be regarded as a class, or each solution in other solution sets can be regarded as a class.

[0097] In one embodiment, sorting the subspaces according to their crowding level and determining multiple target subspaces from them based on the sorting results may include:

[0098] The subspaces are sorted according to their crowding levels to obtain the sorting results; the number of iterations corresponding to the multi-objective optimization when the multi-objective optimization results are obtained is obtained, and the dynamic crowding threshold is determined based on the number of iterations and the maximum number of iterations of the multi-objective optimization; the subspaces with crowding levels greater than the dynamic crowding threshold are selected from the sorting results to obtain multiple target subspaces.

[0099] In this embodiment, in order to further determine the appropriate target subspace at different optimization stages, in addition to determining different numbers of clusters according to different optimization stages before clustering, different dynamic crowding thresholds are also determined according to different optimization stages. For example, a relatively small threshold can be determined in the early stage of optimization so that each surrogate model can fully explore as many target subspaces as possible. In the later stage of optimization, a relatively large threshold can be determined for targeted exploration.

[0100] Optionally, subspaces with crowding levels greater than a dynamic crowding threshold can be filtered from the sorting results to obtain multiple target subspaces, which may include:

[0101] From the sorting results, select subspaces with a crowding degree greater than the dynamic crowding degree threshold as candidate subspaces; based on the crowding degree of the candidate subspaces, merge candidate subspaces with similar crowding degrees to obtain multiple target subspaces.

[0102] In this embodiment, considering that after clustering the Pareto optimal solution set and other solution sets respectively, there may be sub-solution sets with similar sparsity in the Pareto optimal solution set and other solution sets, that is, there are subspaces with similar crowding, after screening out the candidate subspaces through the dynamic crowding threshold, the candidate subspaces with similar crowding can be merged to reduce the number of proxy models that need to be constructed.

[0103] In order to avoid excessively increasing the computational cost or affecting the computational efficiency when merging candidate subspaces, only candidate subspaces with a crowding degree greater than the maximum crowding degree threshold can be merged, so as to avoid the number of samples in the merged candidate subspace being too large and affecting the computational efficiency.

[0104] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0105] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0106] Figure 3 A schematic diagram of the multi-condition calibration device for electromechanical system simulation parameters provided in an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below:

[0107] like Figure 3 As shown, the multi-condition calibration device for electromechanical system simulation parameters includes: objective function construction module 31, first optimization module 32, non-dominated sorting module 33, first subspace partitioning module 34, second subspace partitioning module 35, second optimization module 36, and cyclic calibration module 37.

[0108] The objective function construction module 31 is used to construct multiple objective functions with the goal of maximizing the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each working condition.

[0109] The first optimization module 32 is used to perform multi-objective optimization on multiple objective functions by taking each simulation parameter of the electromechanical system simulation model as a decision vector, and update the experience knowledge base according to the multi-objective optimization results.

[0110] The non-dominated sorting module 33 is used to perform non-dominated sorting on each solution of the simulation parameters of the corresponding electromechanical system simulation model in the experience knowledge base to obtain the Pareto optimal solution set and other solution sets.

[0111] The first subspace partitioning module 34 is used to cluster the Pareto optimal solution set and the other solution sets respectively, and to record each sub-solution set in the clustering result as a subspace.

[0112] The second subspace partitioning module 35 is used to sort each subspace according to the crowding degree of each subspace, and determine multiple target subspaces from each subspace according to the sorting result;

[0113] The second optimization module 36 is used to construct a proxy model for each objective function in each objective subspace, perform Bayesian optimization based on each proxy model with the desired hypervolume improvement as the acquisition function, and update the external archive of multi-objective optimization and the empirical knowledge base based on the Bayesian optimization results.

[0114] The cyclic calibration module 37 is used to re-execute the step of "multi-objective optimization of multiple objective functions" and subsequent steps according to the updated external archive and the experience knowledge base, until the termination condition is met, and obtain the multi-condition calibration result of the simulation parameters of the electromechanical system simulation model according to the global optimal solution of multi-objective optimization.

[0115] This invention first aims to maximize the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each operating condition. Multiple objective functions are constructed, and each simulation parameter of the electromechanical system simulation model is used as a decision vector. Multi-objective optimization is performed on these objective functions, and the empirical knowledge base is updated based on the optimization results, incorporating population optimization information. Furthermore, by performing non-dominated sorting, clustering, and crowding sorting on each solution in the empirical knowledge base, the base can be divided into multiple objective subspaces when the optimal solution cannot be determined under multiple operating conditions. Within each subspace, a surrogate model is constructed for each objective function and Bayesian optimization is performed. The external archive of the multi-objective optimization is updated based on the Bayesian optimization results. This guides multi-objective optimization based on the surrogate models. Through the synergistic optimization of multi-objective optimization and Bayesian optimization, the computational cost of simulation parameter calibration under multiple operating conditions is reduced, and computational efficiency is improved. Moreover, the optimized and calibrated electromechanical system simulation model can output more accurate simulation data under various operating conditions, improving the model's versatility, robustness, and prediction accuracy.

[0116] In one possible implementation, the first optimization module 32 can be used to perform multi-objective optimization on multiple objective functions using each simulation parameter of the electromechanical system simulation model as a decision vector to obtain a multi-objective optimized external archive; and update the experience knowledge base based on the external archive.

[0117] In one possible implementation, the first subspace partitioning module 34 can be used to obtain the number of iterations corresponding to the multi-objective optimization when the multi-objective optimization result is obtained; determine the number of first clusters corresponding to the Pareto optimal solution set and the number of second clusters corresponding to the other solution sets based on the number of iterations and the maximum number of iterations of the multi-objective optimization; cluster the Pareto optimal solution set according to the number of first clusters, cluster the other solution sets according to the number of second clusters, and record each sub-solution set in the clustering result as a subspace.

[0118] In one possible implementation, the first subspace partitioning module 34 can be used to calculate the ratio of the number of iterations to the maximum number of iterations in multi-objective optimization; compare the ratio with a first threshold and a second threshold respectively; and determine the number of first clusters corresponding to the Pareto optimal solution set and the number of second clusters corresponding to the other solution sets based on the comparison results.

[0119] In one possible implementation, the first subspace partitioning module 34 can also be used to determine whether the number of the first clusters is less than the number of solutions in the Pareto optimal solution set, and to determine whether the number of the second clusters is less than the number of solutions in the other solution sets; if the number of the first clusters is less than the number of solutions in the Pareto optimal solution set, then the Pareto optimal solution set is clustered according to the number of the first clusters; if the number of the first clusters is greater than or equal to the number of solutions in the Pareto optimal solution set, then the Pareto optimal solution set is clustered according to the number of solutions in the Pareto optimal solution set; if the number of the second clusters is less than the number of solutions in the other solution sets, then the other solution sets are clustered according to the number of the second clusters; if the number of the second clusters is greater than or equal to the number of solutions in the other solution sets, then the other solution sets are clustered according to the number of solutions in the other solution sets.

[0120] In one possible implementation, the second subspace partitioning module 35 can be used to sort each subspace according to the crowding degree of each subspace to obtain a sorting result; obtain the number of iterations corresponding to the multi-objective optimization when the multi-objective optimization result is obtained, and determine the dynamic crowding degree threshold according to the number of iterations and the maximum number of iterations of the multi-objective optimization; and filter subspaces with crowding degree greater than the dynamic crowding degree threshold from the sorting result to obtain multiple target subspaces.

[0121] In one possible implementation, the second subspace partitioning module 35 can be used to filter subspaces with a crowding degree greater than the dynamic crowding degree threshold from the sorting results as candidate subspaces; and to merge candidate subspaces with similar crowding degrees among the candidate subspaces according to their crowding degree to obtain multiple target subspaces.

[0122] Figure 4 This is a schematic diagram of a terminal provided in an embodiment of the present invention. Figure 4 As shown, the terminal 4 in this embodiment includes a processor 40, a memory 41, and a computer program 42 stored in the memory 41 and executable on the processor 40. When the processor 40 executes the computer program 42, it implements the steps in the above embodiments of the multi-condition calibration method for electromechanical system simulation parameters, for example... Figure 1 Steps 101 to 107 are shown. Alternatively, when processor 40 executes computer program 42, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 3 The functions of modules / units 31 to 37 shown.

[0123] For example, computer program 42 can be divided into one or more modules / units, one or more of which are stored in memory 41 and executed by processor 40 to complete the present invention. One or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 42 in terminal 4. For example, computer program 42 can be divided into... Figure 3 Modules / units 31 to 37 are shown.

[0124] Terminal 4 can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. Terminal 4 may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will understand that... Figure 4 This is merely an example of terminal 4 and does not constitute a limitation on terminal 4. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal may also include input / output devices, network access devices, buses, etc.

[0125] The processor 40 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0126] The memory 41 can be an internal storage unit of the terminal 4, such as a hard disk or RAM of the terminal 4. The memory 41 can also be an external storage device of the terminal 4, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal 4. Furthermore, the memory 41 can include both internal and external storage units of the terminal 4. The memory 41 is used to store computer programs and other programs and data required by the terminal. The memory 41 can also be used to temporarily store data that has been output or will be output.

[0127] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0128] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0129] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0130] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0131] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units.

[0132] If integrated modules / units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above embodiments of the multi-condition calibration method for simulation parameters of various electromechanical systems. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0133] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A multi-condition calibration method for simulation parameters of an electromechanical system, characterized in that, include: To achieve the highest possible consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each working condition, multiple objective functions are constructed. Each simulation parameter of the electromechanical system simulation model is used as a decision vector to perform multi-objective optimization on multiple objective functions, and the experience knowledge base is updated based on the multi-objective optimization results. For each solution of the simulation parameters of the corresponding electromechanical system simulation model in the experience knowledge base, perform non-dominated sorting to obtain the Pareto optimal solution set and other solution sets; Cluster the Pareto optimal solution set and the other solution sets respectively, and denote each sub-solution set in the clustering result as a subspace; The subspaces are sorted according to their crowding level, and multiple target subspaces are determined from each subspace based on the sorting results. Within each target subspace, a surrogate model is constructed for each target function. Based on each surrogate model, Bayesian optimization is performed using the desired hypervolume improvement as the acquisition function. The external archive of multi-objective optimization and the empirical knowledge base are updated based on the Bayesian optimization results. Based on the updated external archive and the experience knowledge base, the step of "multi-objective optimization of multiple objective functions" and subsequent steps are re-executed until the termination condition is met. Then, the simulation parameters of the electromechanical system simulation model under multiple operating conditions are calibrated based on the global optimal solution of the multi-objective optimization.

2. The multi-condition calibration method for electromechanical system simulation parameters according to claim 1, characterized in that, Using each simulation parameter of the electromechanical system simulation model as a decision vector, multi-objective optimization is performed on multiple objective functions. The empirical knowledge base is updated based on the multi-objective optimization results, including: Using each simulation parameter of the electromechanical system simulation model as a decision vector, multi-objective optimization is performed on multiple objective functions to obtain the external archive after multi-objective optimization. The experience knowledge base is updated based on the external archives.

3. The multi-condition calibration method for electromechanical system simulation parameters according to claim 1, characterized in that, Clustering is performed on the Pareto optimal solution set and the other solution sets respectively, and each sub-solution set in the clustering result is denoted as a subspace, including: When the multi-objective optimization result is obtained, the number of iterations corresponding to the multi-objective optimization is obtained; Based on the number of iterations and the maximum number of iterations for multi-objective optimization, determine the number of the first cluster corresponding to the Pareto optimal solution set and the number of the second cluster corresponding to the other solution sets; The Pareto optimal solution set is clustered according to the first cluster number, and the other solution sets are clustered according to the second cluster number. Each sub-solution set in the clustering result is recorded as a subspace.

4. The multi-condition calibration method for electromechanical system simulation parameters according to claim 3, characterized in that, Based on the number of iterations and the maximum number of iterations in multi-objective optimization, determine the number of the first cluster corresponding to the Pareto optimal solution set and the number of the second cluster corresponding to the other solution sets, including: Calculate the ratio of the number of iterations to the maximum number of iterations for multi-objective optimization; The ratio is compared with the first threshold and the second threshold, respectively; Based on the comparison results, determine the number of the first cluster corresponding to the Pareto optimal solution set and the number of the second cluster corresponding to the other solution sets.

5. The multi-condition calibration method for electromechanical system simulation parameters according to claim 3, characterized in that, After determining the number of the first clusters corresponding to the Pareto optimal solution set and the number of the second clusters corresponding to the other solution sets, the method further includes: Determine whether the number of the first clusters is less than the number of solutions in the Pareto optimal solution set, and determine whether the number of the second clusters is less than the number of solutions in the other solution sets; If the number of the first clusters is less than the number of solutions in the Pareto optimal solution set, then the Pareto optimal solution set is clustered according to the number of the first clusters. If the number of the first clusters is greater than or equal to the number of solutions in the Pareto optimal solution set, then the Pareto optimal solution set is clustered according to the number of solutions in the Pareto optimal solution set. If the number of the second clusters is less than the number of solutions in the other solution sets, then the other solution sets are clustered according to the number of the second clusters. If the number of the second clusters is greater than or equal to the number of solutions in the other solution sets, then the other solution sets are clustered according to the number of solutions in the other solution sets.

6. The multi-condition calibration method for electromechanical system simulation parameters according to claim 1, characterized in that, The subspaces are sorted according to their crowding level, and multiple target subspaces are determined from each subspace based on the sorting results, including: The subspaces are sorted according to their crowding level to obtain a sorting result; The number of iterations corresponding to the multi-objective optimization when the multi-objective optimization result is obtained is obtained, and the dynamic congestion threshold is determined based on the number of iterations and the maximum number of iterations of the multi-objective optimization. From the sorting results, subspaces with a crowding degree greater than the dynamic crowding degree threshold are selected to obtain multiple target subspaces.

7. The multi-condition calibration method for electromechanical system simulation parameters according to claim 6, characterized in that, From the sorting results, subspaces with a crowding degree greater than the dynamic crowding degree threshold are selected to obtain multiple target subspaces, including: From the sorting results, select subspaces with a crowding degree greater than the dynamic crowding degree threshold as candidate subspaces; Based on the crowding degree of the candidate subspaces, candidate subspaces with similar crowding degrees are merged to obtain multiple target subspaces.

8. A multi-condition calibration device for simulation parameters of an electromechanical system, characterized in that, include: The objective function construction module is used to construct multiple objective functions with the goal of maximizing the consistency between the simulation data output by the electromechanical system simulation model and the measured data output by the actual electromechanical system under each working condition. The first optimization module is used to perform multi-objective optimization on multiple objective functions by taking each simulation parameter of the electromechanical system simulation model as a decision vector, and update the experience knowledge base based on the multi-objective optimization results. The non-dominated sorting module is used to perform non-dominated sorting on each solution of the simulation parameters of the corresponding electromechanical system simulation model in the experience knowledge base to obtain the Pareto optimal solution set and other solution sets. The first subspace partitioning module is used to cluster the Pareto optimal solution set and the other solution sets respectively, and to record each sub-solution set in the clustering result as a subspace. The second subspace partitioning module is used to sort each subspace according to the crowding degree of each subspace, and determine multiple target subspaces from each subspace according to the sorting result; The second optimization module is used to construct a surrogate model for each objective function in each objective subspace, perform Bayesian optimization based on each surrogate model with the desired hypervolume improvement as the acquisition function, and update the external archive of multi-objective optimization and the empirical knowledge base based on the Bayesian optimization results. The cyclic calibration module is used to re-execute the step of "multi-objective optimization of multiple objective functions" and subsequent steps based on the updated external archive and the experience knowledge base, until the termination condition is met, and obtain the multi-condition calibration result of the simulation parameters of the electromechanical system simulation model based on the global optimal solution of the multi-objective optimization.

9. A terminal, characterized in that, It includes a memory and a processor, the memory being used to store a computer program, and the processor being used to call and run the computer program stored in the memory to perform the method as described in any one of claims 1 to 7.

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

Images