速度差對高速公路效率及安全之影響

Abstract

本研究以計量模型探討速度差對高速公路效率及安全之影響，其中本研究以五種指標來定義速度差，分別是各時間及空間單位內速度分佈的標準差、變異係數和三種不同的百分位間之差距。以最小平方法以及截取迴歸模型來衡量這些速度差指標對流量之影響，並以負二項迴歸模型和卜松迴歸模型來衡量速度差與事故數之間的關係。 本研究使用國道一號及國道三號的電子收費系統資料來進行分析，以路段(電子收費門架間)為基準進行分析，並嘗試同時考量不同類別的因素(包含道路幾何、天氣、時間等)來進行迴歸，希望可以藉此改善過去研究僅能以車輛偵測器資料蒐集限點速度資料和控制變數不足而導致遺漏變項偏誤這兩部份的疑慮。 在效率方面以小時作為時間單位，進行截取迴歸時以各路段流量的第90、第95以及第99百分位作為上限值進行比較。發現到效率的模型當中，以「速度分佈之標準差」作為速度差指標有最好的解釋能力。「速度分佈之標準差」增加1%大約會使流量減少約45輛車/小時。「平均速度」、「車道數」和「日照時數」若增加也會使流量上升，在其他因素不變的情況下，當「車道數」增加1車道時，流量值會增加大約666輛車/小時；全無日照跟全日照相比流量約增加477輛車/小時。「大型車比例」、「大雨」則會對流量造成負向影響，其他因素時，若「大型車比例」上升1%，流量約減少30~60%(視車種而定)；「大雨的降雨時數」若增加1%，會降低流量約732%。 在事故方面則以月為時間單位，本研究發現和卜松迴歸模行相比，負二項迴歸模型擁有較佳的配適能力，且除了變異係數之外的速度差指標在安全的迴歸模型當中都有差不多的分析能力。在迴歸結果中，「重車比例」、「百萬延車公里」和「每月降雨的時間比例」皆呈現正向之影響。皆與事故數呈現正向關係。「速度分布的第99百分位和第1百分位之差距」若增加1%，會使事故數增加約0.027%。若「重車比例」增加1%，可能導致事故數上升1.02%；若該月內的降雨時間比例多1%，可能會對事故數造成0.53%的增加。

In this study, we discuss the impacts of speed variation on efficiency and safety of freeway, and this paper uses five indexes to define the speed variation, which is the standard deviation, the coefficient of variation and the gap between three percentage sets of the speed distribution in each time and space unit. The impacts of these speed variance indexes on the flow rate is measured by the OLS and the censored regression model; the relationship between the speed variance and safety is measured by negative binomial regression model and poisson regression model. Using the data from the ETC system to analyze, each segment between two gates is the spatial unit of this study. Considering different types of factors (including road geometry, weather, time, etc.) to regress and analysis, hopefully, it can improve the doubts of researches in the past, including data can only be collected by vehicle detector which can only provide spot speed and controlling insufficient variables that might cause the omitted variable bias. In terms of efficiency, this study takes hours as time units. The 90th, the 95th, and the 99th percentile of the flow in each segment were defined as the upper limit while doing censored regression. It was found that in the efficiency model, the "standard deviation of the speed distribution" as the speed variation indicators have the best interpretation. An increase of 1% in the "standard deviation of the speed distribution" would result in a reduction of flow about 45 vehicles / hour. The increase of "average speed", "lane number" and "sunshine hours" will also result in the ascension of traffic flow. In the case of considering other factors, the traffic value will increase by about 666 vehicles per hour when the number of lanes increases by 1 lane; the increase in traffic flow per hours is about 477 vehicles per hour when “sunshine hours” is increased by 1 hour. "Large vehicle percentage" and "heavy rain" will have a negative impact on traffic. If other factors are constant, if the proportion of large vehicle increases by 1%, the flow will be reduced by about 30 to 60% (depending on the type of large vehicle). If rainfall hours of heavy rain increase by about 1%, will reduce the flow of about 732%. In the case of safety, we found that the negative binomial regression model had better fitting ability with the indexes of speed variation (except coefficient of variation). In the regression results, the "speed variation indicator", "million vehicle-kilometers" and "precipitation proportion" are positively related to the number of accidents. When "the gap between the 99th percentile and the first percentile of speed distribution" increased by about 1%, increasing the number of accidents by about 0.036%. If the "heavy vehicle ratio" increased by 1%, the number of accidents increased by 1.99%, and if the proportion of rainfall within the month increeased 1%, it may cause 0.55% increase in the number of accidents.