基于去极化电流和扩展徳拜模型的绝缘材料等值电路参数高精度辨识方法

High Precision Parameter Identification Method of Insulating Material Equivalent Circuit Based on Depolarization Current and Extended Debye Model

为了提高绝缘材料等值电路参数辨识的准确性,编程实现基于线性拟合和去极化电流的扩展徳拜模型参数辨识方法,发现线性拟合可能存在较大误差。为了解决该问题,对其进行改进,在考虑噪声情况下给出待拟合区域的适应性选择策略,同时将线性拟合改为自适应选择线性或指数拟合的方式,从而提出具有较高精度的绝缘材料扩展徳拜模型参数辨识方法。采用原始算法和所提改进算法进行参数辨识。结果表明:①针对理想仿真信号,原始算法和改进算法都能很好逼近去极化电流波形,2种算法拟合相对误差的平均值和最大值仅分别为0.03%和0.30%;参数辨识准确性均很高,2种算法辨识得到电阻和电容相对误差的平均值和最大值分别为1.46%和2.76%。②针对实测信号,原始算法得到的去极化电流与目标差距较大,拟合相对误差的平均值和最大值分别为8.00%和37.00%;而改进算法的逼近能力明显优于原始算法,拟合相对误差的平均值和最大值分别为6.08%和17.75%,参数辨识准确性有保障。

In order to improve the accuracy of insulating material equivalent circuit parameter identification,a parameter identification method of extended Debye model based on linear fitting and depolarization current is programmed,and it is found that the linear fitting error may be large.In order to fix the problem,the original algorithm is improved,and the selection strategy of the region to be fitted is presented considering the noise.At the same time,the linear fitting method is changed to the adaptive selection of linear or exponential fitting method,and a high precision parameter identification method of extended Debye model for insulating material is proposed.The original algorithm and the improved algorithm are used to identify the parameters.The results reveal that for ideal simulation signal,both the original and improved algorithms can approximate the depolarization current well.The mean and maximum relative fitting errors of the two algorithms are 0.03%and 0.30%,respectively.The accuracy of parameter identification by the two algorithms is high.The mean and maximum relative errors of resistance and capacitance identified by two algorithms are 1.46%and 2.76%,respectively.For the measured signal,the depolarization current obtained by the original algorithm has a large difference with the expectation.The mean and maximum fitting relative errors are 8.00%and 37.00%,respectively.The current approximation ability of the improved algorithm is obviously better than that of the original algorithm.The mean and maximum fitting relative errors are 6.08%and 17.75%,respectively.Therefore,the accuracy of parameter identification is guaranteed.