短期捷運運量預測模式發展

Abstract

運量在運輸系統是一項基本的要素。探索短期運量的時變對於運輸規劃及營運是重要的課題。當運量資料被充分的分析時，不僅能提供給運輸規劃者做為決策正確的參考，而且亦能提升運輸系統的績效。本研究應用希爾伯特-黃轉換(Hilbert-Huang transform, HHT)探索運量時變，該方法包含經驗模態分解法(Empirical Mode Decomposition, EMD)及希爾伯特時頻分析(Hilbert Spectra Analysis, HSA)。本研究藉由經驗模態分解法，將短期運量資料分解出數個運量本質模態函數(intrinsic mode function) ，而這些本質模態函數將可呈現運量的局部特性及具有意義的時變，例如，尖峰時段型態、半營運型態、半日型態及全日型態。另外，藉由比較希爾伯特-黃轉換及傅立葉轉換(Fourier transform)的結果，本研究發現，希爾伯特-黃轉換能夠獲得較窄的頻寬，精準地獲得時間-頻率-能量分布，進而可提升運輸系統的績效。 此外，本研究應用共振經驗模態分解法(Ensemble Empirical Mode Decomposition, EEMD)萃取EEMD分量。結果顯示，被萃取的EEMD分量相較於EMD分量更能呈現出運量的局部特性，且更易於解釋運量的特性。在以EEMD分量為基礎的情況下，藉由比較希爾伯特-黃轉換及傅立葉轉換的結果，本研究發現，希爾伯特-黃轉換能夠獲得較窄的頻寬及精準地獲得時間-頻率-能量分布。 短期運量預測是運輸系統一項很重要的要素。運量預測的結果將可作為運輸系統管理參考，例如，營運規劃、車站人潮擁擠管制計畫等。本研究結合經驗模態分解法及導傳遞類神經網路(back-propagation neural networks, BPN)發展EMD-BPN混和預測模式，俾進行短期運量預測。混和預測模式包含三個階段，第一階段(EMD階段)如同上述EMD方法，第二階段(分量界定階段)係界定EMD分量的意涵，第三階段(BPN階段)則係應用BPN方法以進行短期運量預測。本階段預測模式輸入變數包含運量資料、EMD分量及時間因子(包含：星期、時段及上班日或周末型態)。本研究結果發現，以EMD-BPN進行捷運運量預測，其預測的準確度及穩定性皆佳。此外，為了改善EMD分量混疊情形，本研究結合共振經驗模態分解法與導傳遞類神經網路，並以EMD-BPN模式架構發展出EEMD-BPN短期預測模式，俾提升模式預測能力及系統績效。最後，本研究結果發現EEMD-BPN預測模式較EMD-BPN預測模式能夠獲得較佳的預測準確度與穩定度。

Passenger flow is a fundamental element in a transportation system. It is important to explore the time variants of short-term passenger flow for transportation planning. When the data are sufficiently analyzed, transportation planners not only can make better decisions, but also enhance the performance of transportation systems. This study applies Hilbert-Huang transform (HHT), including Empirical Mode Decomposition (EMD) and Hilbert Spectral Analysis, to explore the time variants of passenger flow. The intrinsic mode functions extracted by EMD can represent the local characteristics of passenger flow and imply its meaningful time variants such as peak period pattern, semi-service period pattern, semi-daily pattern and daily pattern. By comparing the results of HHT with that of fast Fourier transform, it indicates that HHT can obtain the narrower frequency band, capture accurately time-frequency-energy distribution, and help to enhance the performance of transportation systems. Additionally, this study uses Ensemble Empirical Mode Decomposition (EEMD) to extract EEMD components. The results indicate that the extracted meaningful EEMD components reveal a more unique pattern than the extracted EMD components. The patterns of these EEMD components in the metro system are more specific and can be explained more easily for management purposes. Comparing HHT and FFT based on EEMD, the results indicate that HHT based on EEMD can obtain a narrower frequency band, and accurately capture the time-frequency-energy distribution. Short-term passenger flow forecasting is a vital component of transportation systems. The forecasting results can be applied to support transportation system management such as operation planning, station passenger crowd regulation planning, and so on. In this study, a hybrid EMD-BPN forecasting approach which combines EMD and back-propagation neural networks (BPN) is developed to predict the short-term passenger flow in metro systems. There are three stages in the EMD-BPN forecasting approach. The first stage (EMD Stage) refers to the aforementioned EMD method. The second stage (Component Identification Stage) identifies the meaningful IMFs as inputs for BPN. The third stage (BPN Stage) applies BPN to perform the passenger flow forecasting. The historical passenger flow data, the extracted EMD components and temporal factors (i.e., the day of the week, the time period of the day, and weekday or weekend) are taken as inputs in the third stage. The experimental results indicate that the proposed hybrid EMD-BPN approach performs well and stably in forecasting the short-term metro passenger flow. Additionally, in order to deal with the problem of mode mixing of components, this study uses EEMD and BPN to develop another hybrid EEMD-BPN forecasting model which combines EEMD and BPN, and is based on the framework of the hybrid EMD-BPN model to enhance the capability of the forecasting model. The experimental results reveal that the hybrid EEMD-BPN forecasting model outperforms the hybrid EMD-BPN model.