Loss Optimal Design Of Traction Transformer Based On Stochastic Spectral Embedding Model

With the gradual improvement of China’s economic strength,the country has formulated a strategic plan to strengthen its position as a leader in both science and technology,and a strong country in transportation in order to coordinate and promote the development of transportation science and technology innovation.Additionally,in response to the major challenge of "carbon peak carbon neutral",China has put forth measures to optimize its transportation structure and improve the energy efficiency of its transport equipment.These measures include promoting the use of low-carbon transport equipment and strengthening research and development efforts in energy-saving technologies.Of particular concern is the largest number of electric locomotives in China’s transportation system,which have seen a steady increase in carbon emissions as transport demand continues to rise.The loss characteristics of the traction transformer,a core component of electric locomotives,directly impact the efficiency of the entire power supply system.However,optimizing the loss characteristics of traction transformer is a complex task that involves multiple structural variables,a large amount of simulation calculations,a long optimization period,and poor optimization effects with low accuracy of target parameters.Therefore,this thesis aims to address the optimization design problem of energy efficiency performance of traction transformers by carrying out an optimization of the loss characteristics based on the parameters of structural components.The specific work undertaken in this thesis includes:First,we have conducted a study on the traction transformer that is installed on board of electric locomotives.The transformer has a rated capacity and voltage of 10160 k V·A/25 k V.Using the transformer eddy current field theory and finite element method,we have created a model of the traction transformer with the initial structure parameters.The performance of the traction transformer is analyzed by FEM simulation software,including the analysis of current waveform and core flux density distribution under no-load condition.Additionally,the distribution of magnetic field leakage and eddy current density distribution of the traction transformer tank,upper and lower clamps under load condition with and without a shield are also analyzed.Through this analysis,the loss characteristics of the traction transformer under the initial parameters and the importance of the shield to reduce the eddy current loss of the equipment are clarified.Then,this thesis addressed the challenges of large computational volume and low accuracy of data fitting models in the process of optimizing structural component parameters of the traction transformer.To overcome these challenges,an optimal design scheme is proposed that utilizes a high precision,low-computational-cost surrogate model technique combined with a multi-objective optimization algorithm.Through research and analysis,the stochastic spectral embedding surrogate model is suitable to deal with practical engineering problems.This model uses a disjoint recursive division of the input domain,allowing input samples on different regions to replace the overall complexity of the model and refine the localization of the final domain.After constructing the corresponding objective function,the challenge in this thesis is to address the problem of minimizing both eddy current loss and stray loss.To achieve this,the thesis employs an improved type II non-dominated ranking genetic optimization algorithm that meets the requirements for both solution speed and convergence accuracy.Subsequently,the practicality and feasibility of the optimization scheme for optimizing practical engineering problems are verified by three standard test functions.Finally,to achieve the optimal design of minimizing eddy current loss and stray loss characteristics,we conduct a sensitivity analysis of different structural parts to determine the optimization variables related to the magnetic shield sheet and the size of the clamped parts.After obtaining the sample data by parametric analysis,the Pareto optimal frontier solution set is imported into the optimization scheme.The optimized solution delivers eddy current loss and stray loss values of 6.01 k W and 8.74 k W,respectively.The corresponding dimensions of the magnetic shield thickness,magnetic shield distance from the tank,and the clip without magnet slot length are 16.60 mm,4.32 mm,and 159.85 mm,respectively.The loss values obtained from the simulation validation model based on the optimized data are 6.01 k W and 8.71 k W,which differ from the optimized results by 0.13% and 0.25%,respectively.Compared with the loss of the initial model,these values were reduced by 21.27% and 6.38%,respectively.These results demonstrate the effectiveness of the optimization scheme proposed in this thesis for addressing the traction transformer loss characteristics optimization problem.