Determination of GPS sampling interval for trip energy consumption estimation of electric buses: Analysis of real-world data

Abstract

This paper analyzes the effect of GPS sampling interval on the performance of the longitudinal dynamic model for estimating electric bus trip energy consumption based on real-world operational data. The performance of the estimation model under different sampling intervals is primarily evaluated by the MAE, RMSE, MAPE, and probability distribution functions. It is observed that the relationships between sampling intervals and the MAE, RMSE, and MAPE are all roughly S-shaped. The estimation accuracy is similar when the sampling interval is less than or equal to 13 s. The probability distribution functions of residuals are no longer stably consistent with that of the observed trip energy consumption when the sampling interval is larger than 16 s. In addition, the adaptability of the estimation model under different sampling intervals is analyzed from the perspective of bus operation management. Results indicate that the threshold value of the sampling interval is 16 s at the current battery rated capacity of 162.3 kWh. The threshold value of the sampling interval will become smaller with the decrease in battery rated capacity and increase in daily operation mileage.