函數型方法應用於輕型車油耗與一氧化碳排放資料分析

Abstract

隨著人口發展與經濟成長，人民生活水平上升使得車輛使用率逐漸普及，交通工具成為大部分已發展國家溫室氣體排放主要來源之一，減少運輸工具廢氣排放日趨重要。此外，全球石油產能日漸減少造成油價不斷攀升，對於能源使用應尋求更有效率之方式，除發展公共運輸工具外，私有運具節約用油與能源使用效率亦扮演著重要的角色。因此本研究以汽油輕型車進行油耗與一氧化碳排放分析，研究結果可作為未來車輛油耗或一氧化碳排放管理及決策時的參考依據。 為探討車輛行為與油耗或一氧化碳排放之間相互關係並找出其關鍵影響因子，本研究使用工研院2015 年於新竹路段測試的輕型車資料，以函數型資料觀點進行主成分分析及迴歸分析，並參考文獻傳統方法分析後進行比較。函數化處理後的資料，消除了因蒐集過程中外在因素產生的雜訊對分析的影響，也簡化了傳統方法大量的計算過程及時間。研究結果發現在函數型主成分模式中，車輛油耗與一氧化碳表現出較一致的變化趨勢，更容易表達其局部變異方式以及具有更高的解釋能力。而由於研究樣本數不足之限制，函數型線性模式配適程度雖較傳統模式低，但更容易比較不同道路特性下的校估結果，其中引擎轉速與節氣門開度影響油耗最為顯著，尤其行經市郊區等壅塞的低速路段轉往順暢的高快速路段、進出道路交會處及交織路段，駕駛受路型轉換而改變其駕駛行為，是造成油耗劇烈上升之主因，而各解釋變數影響一氧化碳排放方式雖與油耗相似，但變化較為平緩且影響力不如油耗顯著，未來研究可在樣本數充足情況下使用更適當之基底數配適後進行討論。

With the growth of population and economy development, people’s living standards rose thus making the use of vehicles become more and more common. Vehicles became one of the main source to cause greenhouse gas in developed countries, therefore, it’s important to reduce the emission of greenhouse gas. Furthermore, the lack of global oil resource made oil price rise continuously, so it’s never too late to find a more efficient way for energy use. Besides public transportation, conservation of oil of privately-owned vehicles and energy-saving are also crucial. This research analyzed the fuel consumption and CO emission of light duty vehicles. The study result could be the reference for vehicles’ fuel consumption and CO emission management in the future. In order to find key factors of the relationship between vehicle-behavior and fuel consumption or CO emission, this study used Industrial Technology Research Institute’s data of Hsinchu area from 2015. Using functional Principal Component Analysis and Regression Analysis, and also compared with traditional method to analyze. We eliminated the noise from external factors, and simplified the calculating process and time after functionalizing the data. The result shows that fuel consumption and CO emission have the similar tendency in functional PCA pattern, and it’s easier to express and explain the variation in parts. Due to insufficient samples, functional linear model’s goodness of fit is lower than traditional model, but on the other hand, it’s easier to compare calibration results of different roads’ features. Engine-speed and throttle have significant effect on fuel consumption. Especially at those transfer routes from low-speed to high-speed sections, and interchange sections. Changing the driving behavior because of road conversion is the main cause of intense rise of fuel consumption. CO emission has similar performance with fuel consumption in explanatory variables, however, the impact is less significant. Future study could use sufficient samples to fit with appropriate basis number to discuss.