非均匀采样系统的支持向量回归建模与控制

Modeling and Control of Non-Uniformly Sampled Systems Using Support Vector Regression

针对非均匀采样系统提出了一种基于支持向量回归的预测控制算法.应用提升技术将非均匀采样系统分解为多个并列的子系统,建立了各子系统的支持向量回归模型,将该模型作为预测模型提出了新的优化目标函数,可将控制量更新周期内的系统输出作为优化目标.在多通道电液力伺服同步加载系统中的应用结果表明,该预测模型可以达到比较高的预测精度,且各子系统的模型有一致的预测误差水平.对加载系统控制的仿真试验结果表明,该算法有很好的控制性能,而且通过设计合适的优化目标函数,可以避免数字控制系统中的计算延迟和抑制期望轨迹突变时的超调.

A model predictive control algorithm based on support vector regression（SVR） for the non-uniformly sampled system is presented.The lifted models of the non-uniformly sampled system are derived in the state-space domain,and the system is divided into several parallel subsystems according to the lifted models.SVR is utilized to establish the models of all subsystems,and these models are applied to the algorithm as the predict models.The results of multichannel electrohydraulic force servo synchronous loading system demonstrate that the prediction precision of the predict models gets high and the prediction error level of each submodel remains consistent.The simulation for the loading system control shows the satisfactory performance.Furthermore,the algorithm enables to avoid the computational delay of the digital system and to damp the overshoot caused by the mutation of the desired trajectory by designing optimization object function.