The fence on the MPC of heavy haul trains

The application of fence methodology on the optimal operation of heavy haul trains through model predictive control (MPC) approach is investigated in this paper. This is mainly motivated by the fact that without fencing, the large number of cars in a long heavy haul train present difficulties in solving the train optimal operation problem under MPC framework. Firstly, the large car number leads to a problem with a large dimension, which slows down the optimization procedure in MPC. Secondly, the signal channels in a train is limited physically so that individual manipulation of each car is physically infeasible. The fixed fence method is employed and its influence on the optimization problem and performances of the train is analyzed. Simulation is carried out to demonstrate the advantage of fence technology. It is shown that with a large fence factor, the complexity of the optimization is largely reduced as well as the computation time and memory demand in the real time implementation. Simulation verifies that it is possible to use fence method to facilitate the implementation of MPC approach for heavy haul trains with little sacrifice of the train's performances. In addition, it is analyzed and demonstrated by simulations that the fence factor, however, should be chosen carefully.