Performance assessment of switched control systems based on tensor space approach

Based on tensor space, a new performance assessment approach is proposed for switched control systems in this paper. Switched control system is widely existing in industrial processes, especially in safety‐important processes where the systems controlled may be switched by protection logic during accident. Clearly, bad performance of control and protection strategies will eventually drive the system to a dangerous condition. To treat this problem, one may set up a performance assessment procedure for the switched control systems using multiple model method, where performance is evaluated through assessment of every individual submodel of the switched system. Obviously, this approach ignores interaction nature between control algorithm and protection strategy. The proposed method can capture the interaction nature of control and protection systems. In the tensor space, the interacting relation of control and protection systems can be synthetically represented by adding the logical switching space. Specifically, the tensor space modeling representation for switched systems has been developed, which can actually model the high coupling interaction of control and protection systems. The data‐driven tensor space algorithm based on higher‐order singular value decomposition has also been developed to assess the performance of switched control systems. By using orthogonal projection in tensor space extended from the matrix space, prediction error approach has been employed to obtain the optimal prediction error variance being as the control performance benchmark for performance assessment. Finally, numerical simulation examples are presented to illustrate the rationality and effectiveness of tensor space approach by comparing with the multiple model approach. Copyright © 2015 John Wiley & Sons, Ltd.