一种基于萤火虫支持向量机的油色谱在线数据校正方法

Method for the Oil Chromatographic On-line Data Reconciliation Based on GSO and SVM

为了解决在线油色谱受外界环境和设备误差影响导致数据失真的问题,笔者提出了一种基于萤火虫支持向量机的油色谱在线数据校正的方法。首先将支持向量机中的一组错误惩罚因子,不敏感参数和核参数作为萤火虫个体,通过萤火虫算法对影响支持向量机性能的重要参数进行优化。然后计算油色谱离线数据间的分段函数,当在线数据超出分段函数误差允许的范围时,认为在线数据异常。利用少数准确的油色谱离线数据对支持向量机回归模型进行训练,当在线数据出现异常时,通过支持向量机回归模型对异常的在线数据进行校正。最后通过某台变压器油色谱的在线和离线数据对文中提出的方法进行验证,结果证明了该方法的可行性和有效性。

To solve the problem of oil chromatographic on-line data distortion caused by outside environment and equipment error,the oil chromatographic on-line data reconciliation based on glowworm swarm optimization（GSO） and support vector machine（SVM）is presented. Firstly,the penalty factor,the insensitive and kernel parameter of SVM is seemed as a glowworm. And the important parameters effecting the performance of SVM are optimized through GSO. Secondly, the piecewise function of oil chromatographic off-line data is computed. The oil chromatographic on-line data is abnormal when it is out of the piecewise function error range. SVM regression model is trained by some precise oil chromatographic off-line data. Then the oil chromatographic on-line data is reconciled by SVM regression model when the on-line data is abnormal. Finally, the feasibility and efficiency of the method proposed in the paper is demonstrated by the oil chromatographic on-line and off-line data of the power transformer.