改进的粒度支持向量机在甲醇合成中的应用

Application of the Improved Granular Support Vector Machine in Methanol Synthesis

针对甲醇合成过程中的复杂性和非线性等问题,利用共享最近邻(SNN)相似度将训练样本划分成若干个信息粒,然后分别进行支持向量提取,最后将提取出的支持向量融合,建立最终粗甲醇转化率预测模型。试验结果表明,改进的粒度支持向量机(GSVM)可以将"冗余数据"进行删减,获得更"稀疏"的回归模型,精度也高于传统支持向量机的粗甲醇转化率模型,从而能更好地指导甲醇生产。

To counter the problems of complexity and nonlinearity in methanol synthesis process, by using shared nearest neighbor （ SNN ）similarity, the training samples are divided into several information granules, then support vector extraction is conducted respectively, finally theprediction model of crude methanol conversion rate is built from these extracted support vectors. The experimental results show that the improvedgranular support vector machine can delete Fredundant data＂ and to get ＂sparse＂ regression model, and offer higher accuracy than traditionalsupport vector machine crude methanol conversion rate model, thus the methanol production can be guided better.