基于量子神经网络的平面变压器分布参数的预测方法

Estimation method of the distributing parameters for planar transformer based on quantum neural network

平面变压器寄生参数的预测值是变压器结构设计以及电源产品环路稳定性分的重要依据。针对平面变压器的寄生参数与其结构、材料和工作状态之间的非线性关系,分析了已有寄生参数预测方法的特点和不足,提出了一种基于量子神经网络的平面变压器寄生参数预测方法,并使用36组平面变压器的结构参数和分布参数对神经网络模型进行了训练和验证。预测参数值与测试参数值吻合良好,平均误差在3%以下,最大误差在8%以下。最后,使用量子神经网络得出的预测模型,与使用传统的BP（误差反向传播）神经网络得出的模型进行对比,论证了量子神经网络在变压器分布参数预测上的优越性。

The prediction of distributing parameters is important for the design of the planar transformer and the power product loop stability analysis.Based on the nonlinear relationship between the parasitic parameters of planar transformer and its structure,material and working state,the characteristics and shortcomings of the existing parasitic parameter prediction methods are analyzed.A distributing parameters prediction model for the planar transformer,which is based on the quantum neural network（QNN）,is proposed.36 groups of planar transformers＇ construction parameters and distributing parameters were used to train and verify the accuracy of the QNN prediction model.The predicted results obtained by the model are in good agreement with the measured ones at the average error less than 3%and maximum error less than 8%.Compared with the model obtained by BP（error back propagation）neural network,the superiority of the quantum neural network model in the prediction of transformer distribution parameters is demonstrated.