Research On Combined Optimization Model And Algorithm Of Highway Network Toll Location And Toll Level

The present condition of highway network toll in some regions, where economy are not flourishing and transportation flows are low, indicates that if we carry out the charges to all roads which can be tolled, it will be placed some toll roads to the deficiency condition maybe. Therefore, when we determine the optimal toll level of the highway network, we should need to choose the optimal toll location in the network at the same time. It is significant to study this problem, especially for western far region of China.The dissertation attempts to determine optimal toll level combine with toll location in highway network based on bi-level programming model. Considering both sides of government, toll road operator and road user three square benefitses, the bi-level model's aim is to reduce the network system resistance, improve the utilities of highway and its service quality, and realize maximum of social welfare. The main research content is as follows.1. Multi-vehicle elastic demand stochastic user equilibrium model are established, the model reflects really elastic variety of the network transportation demand, which influenced by the links charge level of highway network etc. And the model reflects the stochastic of different vehicle users choose travelling path behavior.2. Proposing a bi-level programming model to determine optimal toll level combine with toll location in highway network. The upper-level problem's aim is to maximum consumer surplus or network travel demand, and the lower-level of the model is the multi-vehicle elastic demand stochastic user equilibrium model.3. Bi-level programming problems are generally difficult to solve because of its nonconvex, so this paper adopt a kind of improved genetic algorithm-simulated annealing hybrid optimization algorithm( IGA/SA), which combine the parallel searching structure of genetic algorithms(GA) with the probabilistic jumping property of simulated annealing(SA).Lastly, the result of numerical example indicate that multi-vehicle model is more close to fact than the single-vehicle type equilibrium model. It also indicate that charge to a part of roads which can be toll in highway network is feasible, it is necessary to sdudy optimization toll level combine with toll location. In addition, judging from the results and running time of the four algorithms, IGA/SA optimization strategy is more reliable and efficient in solving bilevel programming problems than the other three algorithms: GA, SA, GA/SA.