基于减法聚类的多模型在线辨识算法

考虑到实际工业过程中复杂系统的工况变化往往具有不确定性的特点, 离线辨识的多模型系统难以自适应反映系统的非线性, 因此本文提出一种新的基于减法聚类的多模型在线辨识算法. 首先采用在线聚类算法辨识多模型系统中的局部模型个数与工况参数, 然后充分考虑聚类发生变化对局部模型参数辨识的影响, 给出相应的局部模型参数在线辨识算法. 最后以某电厂300MW锅炉--汽轮机的协调控制系统为对象, 采用上述辨识方法进行仿真研究, 结果验证了本文算法的有效性.

An Online Multi-model Identification Algorithm Based on Subtractive Clustering

In real industrial processes, the operating regime of complex system is always changed into uncertain or unknown regimes, so multi-model system identified off-line is difficult to adaptively describe nonlinearity of real-life process. A new online multi-model identification algorithm based on subtractive clustering is proposed in order to overcome this difficulty. Firstly, the number and operating parameters of local models in multi-model system is updated on-line by subtractive clustering algorithm, then after fully studying the effect on identification of the parameters of local models if the clustering result is changed, the corresponding online identification algorithm is presented to identify the parameters of local models. The presented online identification algorithm is demonstrated with an MIMO simulated 300MW boiler-turbine coordinately controlled process.