磁性元件的磁芯损耗预测研究
Core Loss Prediction Studies for Magnetic Components

随着社会的发展和电子技术的快速发展，磁性元件在电能变换系统中发挥越来越重要的作用，尤其是在高效率和高功率密度需求日益增长的背景下。研究磁性元件中的损耗，特别是磁芯损耗，已成为高效设计的重要课题之一。本文首先对训练集数据进行特征提取，用SVM分类器先对测试集进行训练，利用训练好的模型对测试集进行测试，达到100%的分类准确率。然后对原斯坦麦茨方程添加了温度修正项及系数，用遗传算法优化进行参数识别，对比后发现，修正后的斯坦麦茨方程有更大的异常数据解释性，表明修正后明参数识别效果更好。最后再对测试集先进行特征提取，主要分为四个相关性较高的因素，温度、磁芯材料、励磁波形、磁芯损耗，再进行逐一相关性分析，两两交互相关性分析，方差分析以及分层分析，得到另外三个因素对磁芯损耗的相关性。再进行方差分析，将分析结果使用决策树优化策略得到最优解，预测出测试集的温度为90℃、励磁波形为正弦波、磁芯材料为材料4时，达到最小的磁芯损耗。
With the development of society and the rapid development of electronic technology, magnetic components play an increasingly important role in power conversion systems, especially in the context of the increasing demand for high efficiency and high-power density. Studying the losses in magnetic components, especially the core losses, has become one of the important topics for efficient design. In this paper, the feature extraction is performed on the training set data, the test set is trained by the SVM classifier, and the test set is tested by the trained model, achieving 100% classification accuracy. Then, the temperature correction term and coefficient are added to the original Steinmetz equation, and the genetic algorithm is used to optimize the parameter identification, and it is found that the modified Steinmetz equation has greater explanatory data of abnormal data, indicating that the modified Ming parameter recognition effect is better. Finally, the test set was first extracted by features, which were mainly divided into four factors with high correlation, such as temperature, core material, excitation waveform, and core loss, and then the correlation analysis of the test set was carried out one by one, the interaction correlation analysis of two pairs, the analysis of variance and the stratification analysis was carried out to obtain the correlation of the other three factors to the core loss. Then, the analysis of variance is carried out, and the analysis results are obtained by using the decision tree optimization strategy to obtain the optimal solution, and the minimum core loss is predicted when the temperature of the test set is 90？C, the excitation waveform is sine wave, and the core material is material 4.