基于主元分析的RBF神经网络多模型切换控制

RBF neural network multiple model switching control based on principal component analysis

针对传统多模型自适应控制中子模型数量过多、学习和自适应能力的局限性等问题,将神经网络的学习能力和非线性逼近能力与多模型切换的思想相结合,提出基于主元分析的径向基(RBF)神经网络多模型切换控制算法。首先,基于主元分析法进行工况区域识别。其次,在不同的工况区域内采用RBF神经网络建立多个子模型并设计相应的控制器。最后,根据性能指标函数选择相应的控制器以得到最佳的控制效果。仿真结果表明,该算法大大减少了子模型数量,并改善了系统的动态性能。

A multiple model switching control method based on principal component analysis is presented, which aims at the limitation of traditional multiple model adaptive control such as large number of sub-models, learning and adaptation. First, the operating mode area is identified based on the principal component analysis. Secondly, multiple sub-models are established using Radial Basis Function （RBF） neural networks in different operating modes and the corresponding controllers are designed. Finally, the ＂best＂ controller is chosen by performance indices in order to get the best control effect. The simulation results show that the proposed control scheme greatly reduces the number of sub-models and improves the system＇s dynamic performance.