张闹闹, 杨帅, 刘关平, 王航, 肖集雄. 电子概率密度对气体绝缘强度预测的影响[J]. 高电压技术, 2022, 48(11): 4323-4329. DOI: 10.13336/j.1003-6520.hve.20211634
引用本文: 张闹闹, 杨帅, 刘关平, 王航, 肖集雄. 电子概率密度对气体绝缘强度预测的影响[J]. 高电压技术, 2022, 48(11): 4323-4329. DOI: 10.13336/j.1003-6520.hve.20211634
ZHANG Naonao, YANG Shuai, LIU Guanping, WANG Hang, XIAO Jixiong. Influence of Electron Probability Density on Prediction for Insulation Strength[J]. High Voltage Engineering, 2022, 48(11): 4323-4329. DOI: 10.13336/j.1003-6520.hve.20211634
Citation: ZHANG Naonao, YANG Shuai, LIU Guanping, WANG Hang, XIAO Jixiong. Influence of Electron Probability Density on Prediction for Insulation Strength[J]. High Voltage Engineering, 2022, 48(11): 4323-4329. DOI: 10.13336/j.1003-6520.hve.20211634

电子概率密度对气体绝缘强度预测的影响

Influence of Electron Probability Density on Prediction for Insulation Strength

  • 摘要: 为提高SF6替代气体研发效率,需分析气体分子微观参数与宏观绝缘强度的关系。本文基于密度泛函理论,采用M06−2X方法与6-311++G(d,p)基组,计算了72种气体分子的结构参数,以及不同电子概率密度等值面的相互作用性质函数(general interaction properties function,简称GIPF)算符,研究了各参数与气体相对绝缘强度的相关性,并提出了绝缘强度预测模型。结构参数中分子量、极化率,以及GIPF参数中分子体积、分子表面积、分子正静电势表面积、分子非极性表面积、分子极性表面积,与相对绝缘强度相关性较强。结构参数不受电子概率密度取值影响,GIPF参数在不同电子概率密度等值面差异较大,多数GIPF参数在电子概率密度3.374~33.742 nm–3的范围内与绝缘强度的相关性最强。仅使用GIPF参数的预测模型,在电子概率密度10.123 nm–3时其决定系数R2达到最大值0.762,均方误差为0.119;加入极化率后,在电子概率密度16.871 nm–3R2可达最大值0.801,均方误差为0.099。考虑不同电子概率密度等值面的GIPF参数,联合结构参数,可有效提高绝缘强度预测精度。

     

    Abstract: In order to improve the research efficiency of substitutes for SF6, structure-activity relationship between microscopic descriptors and insulation strength of gaseous molecules is necessary to be analyzed. Based on the density functional theory, the M06–2X method with 6-311++G(d, p) basic set is used to calculate the structural parameters and the GIPF parameters at multiple electron probability density isosurfaces for 72 kinds of molecules. Correlation coefficients of different parameters with the insulation strength are analyzed, and the model for predicting insulation strength is presented. The molecular weight, the polarizability in the structural parameters and the molecular volume, the total surface area, the positive potential surface, the non-polar surface area, and the polar surface area in the GIPF parameters all are strongly correlated with relative insulation strength. The structural parameters are not affected by the value of electron probability density. However, the correlation of GIPF parameters varies widely at different electron probability density isosurfaces. Most of the GIPF parameters have the strongest correlation with insulation strength at 3.374 nm–3~33.742 nm–3 electron probability density. The prediction model is built based on GIPF parameters at 10.123 nm–3 electron probability density isosurface, and the determination coefficient is 0.762 with a mean squared error(MSE) of 0.119. After taking the polarizability α into consideration, the model determination coefficient is 0.801 with a MSE of 0.099 at 16.871 nm–3 electron probability density isosurface. The prediction accuracy of insulation strength can be effectively improved by stratifying the electron probability density, and combining the structural parameters with the GIPF parameters.

     

/

返回文章
返回