为快速、准确预测接触网复合绝缘子临界污闪电压,减少人工污秽试验工作量,提出了一种复合绝缘子污秽闪络电压预测模型。首先,利用黄金正弦(golden sine algorithm,GSA)与分段线性混沌映射(piecewise linear chaotic map,PWLCM)改进的减...为快速、准确预测接触网复合绝缘子临界污闪电压,减少人工污秽试验工作量,提出了一种复合绝缘子污秽闪络电压预测模型。首先,利用黄金正弦(golden sine algorithm,GSA)与分段线性混沌映射(piecewise linear chaotic map,PWLCM)改进的减法平均优化器(subtraction average based optimizer,SABO)算法增强反向传播(back propagation,BP)神经网络的性能;其次,利用人工污秽试验获取10种不同复合绝缘子的闪络电压,收集相关试验参数;再次,依据Obenaus模型对复合绝缘子污秽闪络表现进行分析,利用斯皮尔曼(Spearman)相关系数法筛选出与复合绝缘子临界污闪电压密切相关的4个参数作为预测模型的输入特征量,以预测复合绝缘子临界污闪电压;最后,利用五折交叉验证法对预测模型进行综合评估,并与常用智能优化算法预测模型的预测结果进行比较。结果表明:GSABO-BP模型预测复合绝缘子污闪电压平均绝对误差为1.244 kV,平均绝对百分比误差为2.25%,模型可决系数稳定在0.98以上;与改进前的SABO-BP模型相比,预测值平均误差下降67.80%。GSABO-BP模型在复合绝缘子污闪电压预测上具有较高的预测精准度,对电气化铁路供电系统的防污保护工作具有重要意义。展开更多
Four rapid spectral comparison methods are introduced in mobile labs.They are conformity test method,general correlation coefficient method,reverse correlation coefficient method and correlation coefficient method usi...Four rapid spectral comparison methods are introduced in mobile labs.They are conformity test method,general correlation coefficient method,reverse correlation coefficient method and correlation coefficient method using characteristic spectral ranges.The first method is used for tracking the movements of drugs in the distribution channels;the second is used for quickly identify new counterfeit drugs;the last two are used to screen drugs illegally added in Traditional Chinese Medicine （TCM）.The applicability of the four methods is evaluated with counterfeit and authentic drugs.Our results show that these methods can be quickly constructed and used to identify counterfeit drugs accurately.展开更多
文摘为快速、准确预测接触网复合绝缘子临界污闪电压,减少人工污秽试验工作量,提出了一种复合绝缘子污秽闪络电压预测模型。首先,利用黄金正弦(golden sine algorithm,GSA)与分段线性混沌映射(piecewise linear chaotic map,PWLCM)改进的减法平均优化器(subtraction average based optimizer,SABO)算法增强反向传播(back propagation,BP)神经网络的性能;其次,利用人工污秽试验获取10种不同复合绝缘子的闪络电压,收集相关试验参数;再次,依据Obenaus模型对复合绝缘子污秽闪络表现进行分析,利用斯皮尔曼(Spearman)相关系数法筛选出与复合绝缘子临界污闪电压密切相关的4个参数作为预测模型的输入特征量,以预测复合绝缘子临界污闪电压;最后,利用五折交叉验证法对预测模型进行综合评估,并与常用智能优化算法预测模型的预测结果进行比较。结果表明:GSABO-BP模型预测复合绝缘子污闪电压平均绝对误差为1.244 kV,平均绝对百分比误差为2.25%,模型可决系数稳定在0.98以上;与改进前的SABO-BP模型相比,预测值平均误差下降67.80%。GSABO-BP模型在复合绝缘子污闪电压预测上具有较高的预测精准度,对电气化铁路供电系统的防污保护工作具有重要意义。
基金National Key Technologies R&D Program Foundation of China(Grant No.2008BAI55B06).
文摘Four rapid spectral comparison methods are introduced in mobile labs.They are conformity test method,general correlation coefficient method,reverse correlation coefficient method and correlation coefficient method using characteristic spectral ranges.The first method is used for tracking the movements of drugs in the distribution channels;the second is used for quickly identify new counterfeit drugs;the last two are used to screen drugs illegally added in Traditional Chinese Medicine （TCM）.The applicability of the four methods is evaluated with counterfeit and authentic drugs.Our results show that these methods can be quickly constructed and used to identify counterfeit drugs accurately.
