Study On Multiobjective Optimization Of Train Operation Process And Its Control Strategies

Train running process costs much more energy than those which other parts of rail transit. For the purpose of improving train running efficiency, experts from home and abroad have been studying on the problem extensively. Although much research outcome issues, most of them bases upon the fact that train running process is served as a single objective optimization problem, which minimize the energy consumption with the constraint of safety and time. Essentially speaking, train running process is a multiple objective optimization problem and must ensure it being safe, economic, punctual, comfortable and stopping at adequate position. Such optimization problems don't exist optimal solution, but there exists more one even infinite noninferior Pareto optimal solutions. However,the existent optimization methods can generate only one solution during running optimization code and can only utilize either experience information or datum information provided by control system. So,the following contents are covered in this thesis.Multi-objective optimization model is formulated based on multiple objective optimization methods and theory, which serve energy consumption, punctuality and stopping at adequate position as the goal with the safety constraint.A Hybrid particle swarm algorithm is used to solve the multiobjective optimization problem. Some measures are taken to improve the algorithm performance. Firstly, preference of decision maker is embedded into the algorithm to update solutions, accordingly, all the particles are lead to the expecting zone and this can accelerate convergence of the algorithm. At the same time, train running process are divided into 6 modes according to train operation curves and each solution generated from the algorithm is inevitable to relate to one mode, and regulation strategies are proposed for the each mode. These strategies are used to regulate the new solutions generated from the algorithms, so that accelerating convergence of the algorithm. To guarantee getting satisfied results, human-machine interactive particle swarm algorithm is presented, which shows the decision maker the optimization results after some certain iterations and the decision maker indicates new optimization objectives, Which can be done by modifying the train control strategies in the human-machine interface and content results can be gained after several times regulation.Multiobjective particle swarm algorithm can generate more one Pareto optimal solution simultaneously during its running, which can reveal the real laws among the objectives and is beneficial to make decision more reliable and effective. Aiming at multiobjective optimization problem of train running process, multiobjective particle swarm algorithm based upon preference is suggested. At the same time, several strategies are advanced to modify the algorithm and these strategies improve the computation complexity and convergence of the algorithm evidently.Particle swarm algorithm and multiobjective particle swarm algorithm can not meet the requirement of real time of train control. At the same time, there exists uncertainties during train running, such as, varying load, varying track temporarily and train kinetics varying resulting from different types of vehicles, etc.Because of this, the off-line optimization results is difficult to meet the expecting goal, and even is not a effective strategy anymore. Therefore, adaptive fuzzy logic system which can utilizes experience information and data information is used to adjust the off-line optimization train control strategies. In this way, the uncertainties taking influence to the train running can be avoided.Finally, computer aided multiobjective optimization system is developed with VC++6.0 programming language. The system provides many functions, such as, optimization problem input, algorithm selection, algorithm performance evaluation, optimization results saving and processing, etc, which can be used to optimize train running process, control system and other single and multiple objective optimization problems.