應用船舶大數據預測油耗之研究

STUDY ON PREDICTION OF MARINE FUEL CONSUMPTION WITH BIG DATA

燃油是海運業主要的營運成本，透過改善油耗可提升航程的獲利。本研究針對六艘貨櫃船，設置大數據蒐集設備與管理系統，自動化收集船舶航行資訊，包含船速、轉速、馬力、耗油量及經緯度等數據，同時由船員維護航程的吃水值與俯仰差於系統中，再傳送至岸端資料庫，進行數據前處理。經分析近四年的航行資料庫，篩選適合的航程，建立油耗預測趨勢線，再探討影響油耗變化的因素後，發現本研究的能效指標（power index）與燃油成正相關，能夠反映船舶的航行效率。最後探討油耗預測誤差的修正方式，包含不同的吃水、航況與船速變化等因素。經上述修正後，油耗預測誤差約為5%，可透過持續累積數據改善預測效果。

An improvement of the marine's fuel consumption will enhance energy management efficiency and profitability in ship companies since fuel cost is one of the biggest operating costs. This study collects real-time navigation data from six container ships by using big data equipment and systems on board. The data was collected by equipment automatically such as, ship speed, main engine speed, horsepower, fuel consumption, longitude, latitude, etc. Draft and trim are recorded by crew manually. The data is sent to the onshore database for pre-processing. Filtering suitable voyages from nearly four years of historical navigation data and illustrating scatter diagram support to obtain fuel consumption prediction trend line. Under the influence of various external factors on fuel consumption, the performance indicator Power Index can reflect the ship's navigation performance. The results show that the indicator is proportional to the fuel consumption. After the fuel consumption error caused by different draft, sailing conditions and the ratio of ship speed is corrected, there is only 5% of the error value remained, which indicating that the continuous accumulation of data can improve the predicted fuel consumption.