Research on Prediction Method for Vehicle-to-Grid Interactive Feature Parameters Based on Gradient Boosting Decision Trees

Vehicle-to-Grid (V2G) interaction represents the most direct and effective means to mitigate the impact of large-scale integration of charging facilities on power grid operations. With the proliferation and implementation of various V2G initiatives, there is a growing need for more accurate prediction of the flexible response characteristics of charging loads. This paper, taking public charging stations participating in time-of-use electricity pricing as an example, proposes a set of characteristic and effective parameters to evaluate charging load features. Based on preprocessing the charging load data, the paper develops strategy functions targeting key parameters. Subsequently, a machine learning model based on Gradient Boosting Decision Trees (GBDT) and boosting of weak classifiers is established to simulate and predict the characteristic parameters of charging loads. Finally, the paper conducts data analysis on actual samples to verify the feasibility and accuracy of the model.