Page 304 - Distributed model predictive control for plant-wide systems

278 Distributed Model Predictive Control for Plant-Wide Systems

The traction of the third coach is shown in Figure 12.13.
Figures 12.10 and 12.11 show track status of the first and the second coach, which demon-
strates that the global optimization has the best performance, the decentralized optimization
has the biggest steady-state error and fierce oscillation, and the neighborhood optimization
is much better than the decentralized optimization and similar to the global. Figures 12.12
and 12.13 show the traction status of the powered coach, which demonstrates that using the
neighborhood optimization can get a smaller traction.

12.5 Conclusion
In this chapter, DMPC based on the neighborhood optimization is designed for the high-speed
train with distributed traction. The spring–mass model of the longitudinal dynamics systems
is divided into local subsystems. Based on the neighbor definition, the local subsystems can be
transferred to the neighbor subsystems for the distributed model controller. Simulation results
show that the N-DMPC with information constraint performs better than the method based on
the local optimization with more smooth and less steady-state error.