How much energy will electric buses consume?
In an article recently published in the open access journal Energiesthe researchers discussed estimating the energy consumption of electric buses using a physical and data-driven fusion model.
Study: Estimation of the energy consumption of electric buses based on a physical and data-driven fusion model. Image Credit: Scharfsinn/Shuterstock.com
The global energy structure is gradually changing as energy constraints and pollution problems increase. Due to their low emissions and excellent fuel efficiency, electric vehicles are gaining popularity. Electric buses possess more advantages in the public transport system than ordinary diesel buses.
However, several issues remain, such as challenges in assessing charging needs, planning vehicle routes, and designing battery energy storage systems. Therefore, the development of an accurate vehicle energy consumption prediction model can solve the aforementioned problems that are critical for the widespread adoption of electric buses.
China’s National New Energy Vehicle Monitoring and Administration Platform (NEVS) has collected vast amounts of data on new energy vehicles. For vehicle energy consumption research and modeling, the platform can provide a large volume of vehicle driving data. Previous studies have used physical modeling approaches or artificial intelligence algorithms to develop vehicle energy consumption models. The results of a single type of model or a single method of estimating energy consumption are less reliable. A more reliable energy consumption estimation model with few input features is required.
Transmission of electrical bus operating data and pre-processing results. (a) Road route, (b) big data cloud platform for electric vehicles and (vs) battery SOC preprocessing results. Image credit: Li, X et al., Energies
About the study
In this study, the authors discussed the development of a physical and data-driven fusion model for estimating the energy consumption of electric buses. A simplified physical model was used to model the basic energy consumption of the electric bus. The model took into account the effects of brake use, rolling drag and air conditioning consumption. A CatBoost decision tree model was built to account for volatility in power consumption caused by many causes. After that, a fusion model was created.
The team tested the performance of the energy consumption model using electric bus data analyzed on a big data platform. A physical and data-driven fusion model has been proposed for vehicle energy consumption. Physical formulas have been used to express some of the energy used by the vehicle while driving, such as the energy used by rolling resistance and air resistance. Model complexity has been reduced by using formula modeling directly.
The researchers statistically analyzed the data from the electric buses. To acquire continuous data in the car’s charging and driving cycle process, the original data was pre-processed and recreated. The energy consumption estimation model was also created. Based on the dynamic performance of the electric bus powertrain, a physical model of vehicle power consumption was constructed.
The least squares method was used to calibrate the model parameters first. In addition, elements that influence changes in vehicle energy consumption, such as driving habits and environmental factors, were summarized and discussed. The impacts were described using the CatBoost decision tree model, and finally, the two models were combined to obtain the final result of the vehicle energy consumption estimate. The energy consumption estimation model was reviewed and validated.
The statistical results of the impact of various energy consumption fluctuation factors on vehicle energy consumption. (a–h) The relationships between the speed variation, the average speed, the number of presses on the accelerator pedal, the number of presses on the deceleration pedal, the departure time, the departure date, the ambient temperature , battery internal resistance and power consumption, respectively. Image credit: Li, X et al., Energies
The time taken to process the data was six seconds, while the time spent training the model was only 0.9 seconds. The results revealed that for many vehicles, the errors in estimating energy consumption were less than 8.1%. The results of the estimation of vehicle energy consumption had an average inaccuracy of 7.5%. On the Bus 1 dataset, the relative error of the fusion model was 4.8%.
The proposed model had a high level of accuracy, with an average relative error of 6.1%. The fusion model has been a useful tool for vehicle planning, optimizing the energy consumption of electric buses and the rational organization of charging stations. Other driving factors, such as vehicle departure time, ambient temperature, etc., impacted the vehicle’s driving energy consumption. The correlation coefficient was 0.79. With both indicators, the fusion model outperformed conventional energy consumption modeling methods. The proposed merger model was used as the basis for vehicle planning, optimization of the energy consumption of electric buses and the layout of charging stations. The majority of points with substantial inaccuracies were concentrated in severe weather.
The basic estimate of the energy consumption results from this. (a,b) The fitting curves of two electric buses. (vs,D) The power consumption adjustment results for each room. (e,F) Assembly errors. Image credit: Li, X et al., Energies
In conclusion, this study estimated the energy consumption of electric buses using a data-based and physical fusion model. A rudimentary model of energy consumption was built in terms of physical modeling. The effects of kinetic energy consumption, rolling drag and air conditioning have all been considered. The main factors determining the variability of vehicle energy consumption were investigated using data-driven modelling. The model input features have been simplified so that the model input can be built before the car is driven. The variable power consumption assessment model was built using the CatBoost decision tree modeling approach. The concept of built-in learning was used to optimize the model in a hierarchical iteration during the model training process.
The average relative imprecision of the vehicle energy consumption estimation result was 6.1%. Also, during the driving process, the mass of the vehicle was treated as a constant value, which contributed to the model error. The authors mentioned that meteorological parameters could be added to the model in the future to further improve its accuracy.
Li, X., Wang, T., Li, J., et al. Estimation of the energy consumption of electric buses based on a physical and data-driven fusion model. Energies 15(11) 4160 (2022). https://www.mdpi.com/1996-1073/15/11/4160