With the popularity of electric vehicles, accurate estimation of their transient energy consumption is important for optimizing energy management and improving driving efficiency. To this end, this paper proposes a LightGBM-based transient energy consumption estimation model for highway battery electric vehicles (BEVs). Firstly, the driving trajectory data and energy consumption data of BEVs on highways are collected through real-vehicle experiments, and preprocessing such as data cleaning and feature extraction is carried out. Second, based on the above data, k-means++ clustering was performed on driving behavior according to speed and acceleration, and six key features were selected to construct a transient energy consumption estimation model based on the LightGBM algorithm. Then, it was compared with models such as RF, SVR and MLP to evaluate the model performance. Finally, SHAP is used for feature importance analysis. The results show that the LightGBM-based model has good goodness-of-fit with an R2 of 0.9133, which can accurately estimate the transient energy consumption of BEVs. Among all the features, acceleration and road slope play a greater role in energy consumption estimation.