Development of a transient fuel consumption model

- A transient fuel consumption model is developed and it consists of steady-state and transient correction section.
- Transient correction was introduced on the basis of steady-state prediction.
- The Mean Absolute Percentage Error of the new developed fuel consumption model decreased approximately 29.44-34.48% when transient corrections were introduced.
- Vehicle Specific Power (VSP) is also introduced to better correct fuel consumption prediction.
- The slopes of regression lines for the new model between the measured and the predicted fuel consumption range from 0.9408 to 0.9608.

Most existing fuel consumption models are based on steady-state fuel mapping. However, these models cannot provide satisfactory predictions for vehicles operating under transient conditions. Consequently, a new transient model that can precisely predict fuel consumption under steady-state and transient conditions was developed on the basis of the steady-state model. This new model is characterised by two sub-modules: the steady-state module whose inputs are engine speed and torque, and the transient correction module whose inputs are vehicle speed and acceleration. Vehicle Specific Power (VSP) was introduced as a filter to help develop different correction factors for positive and negative VSP. The Argonne National Laboratory's measurements of different driving cycles including the steady-state cycle, the UDDS cycle, the Highway cycle and the US06 cycle were used to develop the new model. According to the results of this study, the prediction accuracy of fuel consumption models can be improved significantly by introducing transient corrections. The Mean Absolute Percentage Error (MAPE) between the predicted and measured fuel consumption decreased from approximately 58% for the steady-state model alone to approximately 23.5% when the transient correction module was introduced. Moreover, this new model performs better than the VT-Micro model. The MAPE values of the new transient models were approximately 2.7-4.6% lower than that of the VT-Micro model.