隨著車載排放量測設備技術的演進，以及歐盟輕型車輛實車道路測試程序即將規劃上路，利用實車上路瞬時油耗與行車資料發展微觀油耗模式，將是未來幾年的趨勢。本研究利用工業研究院蒐集之實車上路資料，發展並比較常見之輕型車微觀油耗模式，同時驗證其模式預測力。除普通最小平方迴歸，本研究亦發展分量迴歸，以了解重要變數在不同油耗分量下對油耗的影響。研究結果發現，國內外常用之速度加速度油耗模式，其型態雖然簡單且具一定配適程度，但模式交互作用項多，易產生共線問題，導致模式校估值較為敏感，其配適結果亦容易受資料取樣頻率影響。若以車輛比功率替換速度與加速度之高次與交互作用項，不但可保有類似的模式配適度，且能避免共線問題，亦可同時考量坡度與空氣動力的影響，應用範圍較有彈性。油門開度與引擎轉速是車載量測設備蒐集訊號中對油耗影響最為顯著的變數，為有效降低油耗，當車輛具一定行駛速度時，駕駛人不宜大幅改變油門開度，而在交通壅塞時，則應特別注意控制引擎轉速。

The introduction of portable emission measurement system (PEMS) and the proposal of European Union real driving emission regulations (i.e., (EU) 2016/427 and (EU) 2016/646) have made micro fuel consumption investigations based on real driving data an oncoming issue. This study develops and compares three popular micro fuel consumption models based on real driving data collected by the Industrial Technology Research Institute using both ordinary least squares and quantile regression. The analysis results showed that although the most commonly adopted speed and acceleration models were simple and provided certain model fitting performance, the results were easily affected by sampling frequency. The models also suffered multicollinearity issues due to its complicated interactive terms. When high-order and interactive terms of speed and acceleration variables were replaced with vehicle specific power, the aforementioned issues could be resolved. Additionally, throttle and rpm (revolutions per minute) were found to be the most influential variables among the collected PEMS variables. To effectively reduce fuel consumption, drivers are recommended to avoid changing their throttles abruptly in non-congested traffic, and to carefully control their rpm in congested traffic.