Application Research On Urban Rail Transit Passenger Flow Prediction Based On Optimized Wavelet Neural Network

Urban rail transit has many points of advantage: high punctuality rate and speed, large quantity and long distance of transportation, high comport, low source consumption and little outside affected by outside. It not only meets the new travel requirement of the residents, but also can effectively alleviate the urban public traffic congestion phenomenon. Moreover, it also makes contribution to energy saving, noise reduction and the improvement of the atmospheric environment of the city. Above all, the urban rail transit gradually becomes the main choice of public transport to traveler. With the gradual development of urban rail transit system and expansion of road network, it attracts more travelers to choose the urban rail transit. The station appears the passengers excessive saturation with the increase of passengers. Lots of passengers focus on the path having shortest distance or time without detailed information of passenger distribution and passenger flow-induced on the basis of accurate passenger prediction. It also has many problems in traffic capacity deployment, passenger organization, early warning and evacuation plan. Therefore, predicting the short-term passenger flow of urban rail transit has the very vital significance. According to the short-term passenger flow predicion and the real passenger flow data of Zhengzhou rail transit Line 1, this thesis mainly carried on the following research work:Firstly, the characteristics of temporal and spatial distribution of urban rail transit passenger flow are summarized to lay the foundation for the next short-term passenger flow forecasting.Secondly, the methods of short-term passenger flow prediction are summarized, and the characteristics of different methods are analyzed. Compared with traditional forecasting methods, as a new non-linear function approximation tools, the wavelet neural network(WNN) has distinct advantages, which is determined to predict the short-term passenger flow. And the theory, structure, algorithm steps of WNN are elaborated.Thirdly, the original passenger flow data and related factors are preprocessed by cluster analysis and Spearman correlation coefficient, which not only ensure rationality and reliability of input and output, but also improve the training ability of neural network.Finally, aiming at the instability and lower prediction accuracy in wavelet neural network, the Shuffled Frog Leaping and Bat Algorithm with Gauss Mutation(SFLBAWGM) is proposed. It is used to do the prophase optimization of the initial parameter combinations in WNN, which makes up for the defects of sensitive to initial value in WNN. The simulation results show that, the SFLBAWGM has the distinct advantages compared to the Bat Algorithm(BA), Shuffled Frog-Leaping Algorithm(SFLA) and the improved algorithm in the references. What’s more, the predict result of the examples of short-term passenger flow shows that, the model which this thesis builds is the closest one to the measured data.