Abstract:

A selective recursive ridge extreme learning machine is proposed for online identification of nonlinear systems. First, a recursive algorithm of ridge extreme learning machine with nodes growing is formulated, which can update the online model in an efficient manner. Additionally, by incorporating the relative predictive error of the training set, a strategy by selectively increasing nodes is proposed to restrict the complexity of the identification model. Consequently, a more simple identification model with the recursive update manner can be obtained. Multi-fold simulations on a benchmark nonlinear chemical process have been investigated. And the comparison results verify the simplicity and superiority of the proposed approach, which is more suitable for online identification of nonlinear systems.

Key words: nonlinear systems, dynamic modeling, neural networks, recursive algorithm, extreme learning machine, systems engineering

摘要：

针对非线性系统的在线辨识, 提出了一种选择性递推岭参数极限学习机方法。首先, 推导了岭参数极限学习机模型节点增加的递推算法, 以有效地更新在线模型。其次, 结合训练模型的相对误差, 提出模型节点递推增加的选择性策略, 以限制模型的复杂度, 获得更简单的递推辨识模型。通过一个典型非线性化工过程的在线辨识, 从多方面比较验证了所提出方法的简单有效, 更适合非线性过程的在线辨识。

关键词: 非线性系统, 动态建模, 神经网络, 递推算法, 极限学习机, 系统工程