大功率高频变压器三维温升计算及优化设计方法

Three-Dimensional Temperature Calculation and Optimization Design Method for High Power High-Frequency Transformer

三维温升计算是大功率高频变压器(HFT)建模中的难点问题,准确、快速的热模型对于HFT的优化设计和稳定运行有着重要意义。目前,普遍采用降维方式建立集总参数热网络模型,其计算精度易受结构参数影响,在宽参数范围寻优中难以实现准确的热点预测。该文基于有限差分方法(FDM)构建了HFT三维热模型,并对离散化误差及其控制方法进行分析。借助有限元仿真对三维热模型的参数适应性进行详细校验,包括结构尺寸、损耗密度及散热条件,校验结果表明,该模型在宽参数范围内的最大误差小于10%。以10 kHz、150 kW变压器为例进行优化设计,单次三维温升计算平均耗时2.8 ms,最优设计边界上的最高温升计算误差低于5.5%。根据优化设计结果加工150 kW变压器原型样机并进行额定功率测试,最高温升计算误差为3%。

Temperature rise calculation is a complex problem in modeling high-power high-frequency transformers(HFT).An accurate and fast thermal model is significant for the optimal design and stable operation of HFT.At present,lumped parameter thermal network model is usually built through dimensionality reduction,and its calculation precision is easily affected by structural parameters.It is difficult to achieve accurate hot spot prediction in optimizing a wide parameter range.A three-dimensional thermal model of HFT considering anisotropic thermal conductivity is constructed based on the finite difference method(FDM).The discretization error of the three-dimensional thermal model is quantitatively analyzed,which is mainly affected by the loss density,thermal conductivity,and size of the finite difference element.In order to minimize the number of finite differential elements while ensuring the calculation accuracy of the thermal model,three-dimensional sizes of the differential element are adjusted actively according to the discretization error expression.For HTF with nanocrystalline core and litz wire winding,the element sizes in the direction along the width,height of winding and the thickness of core are the key parameters affecting the accuracy of temperature rise calculation.The parameter adaptability of the proposed thermal model is verified in detail with finite element simulation,including structural parameters,loss density,and heat dissipation conditions.The results show that the maximum error of the model is less than 10%in a wide range of parameters.However,the error of the traditional lumped parameter thermal network model is significantly affected by the structural parameters,and the error varies in the range of 10%~80%within the same parameter variation range.Based on the proposed thermal model,the efficiency-power density optimal design of a 10 kHz 150 kW transformer was carried out with the parameter scanning method.The upper limit of temperature rise was set to 100 K,and the differential element sizes were adjusted according to the error of less than 5 K.The optimization design program calculated 500000 design points within 371 s with parallel computing,and the average time of three-dimensional thermal model calculation for a single design point was 2.8 ms.The temperature rise calculation results of the design points on the optimal design boundary were verified,and the error was less than 5.5%.Compared with finite element simulation and the lumped parameter thermal network model,the proposed three-dimensional thermal model based on FDM balances calculation accuracy and calculation speed.On the optimal design boundary,with the increase of power density,the transformer efficiency gradually decreases,and the temperature rise gradually increases.The maximum power density to meet the 100 K temperature rise requirement was about 9 kW/L,and the transformer efficiency was about 99.82%.A 150 kW transformer prototype was processed according to the optimized design point with maximum power density,and the maximum temperature rise was 103 K under rated working conditions.