基于熵准则的发酵过程TSK模糊建模

Fermentation process TSK fuzzy modeling based on entropy criteria

提出了一种基于熵准则函数的TSK模糊系统建模方法.不同于传统的基于MSE经验误差最小的准则函数,该准则函数能从训练样本的整体分布结构来进行参数学习,有效地避免了由于过学习而导致泛化能力差的缺点.将其应用于复杂的发酵过程建模,结果表明新方法具有良好的预测精度、泛化能力和鲁棒性.为解决发酵过程建模中试验数据含有噪音,导致模型预测精度下降的问题提供了一条研究思路.

A TSK fuzzy modeling approach based on entropy criteria is presented in this paper. It considers the whole distribution structure of the training data set in the parameter learning process, which is different from the traditional MSE-criteria based parameter learning, and effectively avoids the bad generalization caused by over-learning. Then the proposed method is applied to the complex fermentation process modeling, and the results demonstrate that this method has better prediction accuracy, generalization and robustness. Such that it offers a research viewpoint to circumvent the problem of the prediction accuracy deteriorated by noise existing in the corresponding experimental data.