以類神經技術為基礎之柴油車輛油耗預估研究

Research of a Predict System for Diesel Car Fuel Consumption Using Artificial Neural Network

吳建達(Jian-Da Wu)；宋狄恩(Di-En Song)

燃油消耗 ； 人工類神經網路 ； 倒傳遞類神經演算法 ； 徑向基底函數類神經 ； Artificial neural network ； back-propagation neural network ； radial basis function neural network ； fuel consumption

車輛工程學刊

10期（2013 / 06 / 01）

75 - 88

繁體中文

本篇報告提出了一種採用人工神經網絡技術為基礎的柴油車輛燃料消耗預測系統。本系統由三個主要部分組成：油耗數據搜集及分類，油耗的預估模式建立和預測性能的分析。在實際的行駛狀況下，柴油汽車的燃料消耗受到多種因素的影響。然而在本系統中的燃料消耗的影響因素簡化設定為車輛廠牌，車輛型式，車輛的重量，車輛類型，變速箱型式，共軌系統，渦輪增壓系統和傳輸模式。根據當前的燃料消耗標準，八項條件作為輸入，用於神經網絡的訓練和燃料消耗量的預測。在測試的數據中使用人工神經網絡的倒傳遞神經網絡（BP神經網絡）和徑向基底函數神經網絡（RBF神經網絡）。由測試的結果顯示，使用類神經網路技術於柴油車輛的燃料消耗量預測是有一定的準確度。在方法比較上，徑向基底函數神經網絡的效果比較傳統的倒傳遞神經網絡效果較佳。

This report presents a predictive system for the fuel consumption of diesel vehicles using an artificial neural network. The system consists of three main parts: data acquisition, fuel consumption forecasting and performance analysis. In the practical drive procedures, the fuel consumption of a diesel vehicle is effected by many factors. However, in the present system, the factors impacting the fuel consumption are vehicle make, vehicle type, vehicle weight, vehicle type, transmission type, common-rail systems, turbocharging systems and transmission mode. According to the current fuel consumption norms, eight conditions are used as the system inputs for neural network training and fuel consumption prediction. In an artificial neural network, both of the back-propagation neural network (BPNN) and radial basis function neural network (RBFNN) are used and compared in the expert system to predict the fuel consumption of diesel vehicles. The prediction results show that the neural network predictive system is effective for predicting the fuel consumption of diesel vehicles and the RBFNN demonstrated better performance than BPNN in the present study.