Modeling And Control Of Transport Network Systems

Nowadays, various transportation network systems, including traffic network, pipeline network and Internet etc., are playing more and more important roles in the human produc-tion and life. Generally, these systems can always be represented as the combination of nodes and edges in terms of graph theory. The mass or energy flow transmitted through the edges should follow the basic physical law such as Ohms theorem, thermodynamics and so on. Meanwhile, control devices are installed at the nodes to operate the networks and regu-late the flow with certain objectives. However, since these transportation networks always distribute over large area and consist of various functional components with strong cou-pling existing between each other, it is complicated to build unified mathematical models to depict the How dynamics, resulting in the difficulty in controller design. For the above reasons, this dissertation mainly proposes several novel modeling, analysis and control ap-proaches on the transportation networks, integrated with the knowledge of physics, graph theory and statistical methods. The main contents are outlined as follows:1. We adopt the Coarse Graining method proposed by H.K.Lee et.al. to develop a macroscopic model from a microscopic traffic model-GOVM. The proposed model inherits the parameter p which considers the influence of next-nearest car introduced in the GOVM model. The simulation results show that the new model is exactly con-sistent with the former microscopic model. The establishment of macroscopic model is significant to solve the control problem over the whole networks analytically.2. We employ LOGIT model, which is widely used in statistical field, to describe the individual decision-making process instead of shortest path rule in transportation net-work models. Combine it with mass conservation equation and congestion effects, we can derive a stationary flow distribution solution analytically under any given net-work structure which is also solved by a numeric iterative approach approximately. Through this approach, the influence of topologies on the traffic load distribution has been analyzed quantitatively which provides powerful support to network configura-tion optimization and traffic scheduling.3. On basis of the linear dynamic network (LDN) model of traffic network provided by E.Camponogara et.al., the cooperative predictive control method is introduced to regulate the split of traffic signal lights in urban traffic networks. The whole network is decomposed to several subsystems and the objective of each subsystem contains the information of neighborhood which is affected by the nonlocal decision made on the subsystem itself. During the decision-making process, we carry on the communi-cations between neighboring subsystems and do the optimization of each subsystem alternatively, a near-optimal control decision is finally derived which is close to the one obtained by centralized method. Furthermore, the novel distributed control sys-tem is flexible and shows high reliability such that the risk is reduced when fault occurs.4. Based on the deterministic state space model of traffic network, we employ Gauss process to depict the uncertainty of traffic flow in the real life, and the chance con-straints MPC approach is applied to control the traffic volume in the arterial roads which afford the heaviest traffic load over the network. This method defines the un-certainty quantitatively and sacrifices the performance of the minor roads to guaran-tee the security of the arteries so that the traffic congestion is prevented.5. We adopt an electric analogue method provided by H.C.Ti et.al. to build a unified model on natural gas network which is easy to compute with tolerant accuracy. On basis of it, we successfully develop a dynamic optimal problem with state constraints which is formulated with minimizing energy costs subject to the customer demand in gas flow and pressure. Meanwhile, an iterative dynamic programming(IDP) al-gorithm is proposed to cope with the nonlinear optimization problem. The novel approach breaks through the limitation of conventional method which can only solve the static optimization problem of pressure dispatching, and highlights a new way to realize real-time control on pipeline network.The conclusions and perspectives are presented in the end of the dissertation.