Causal Inference Based Predictive Framework for Electric Vehicle Energy Consumption Forecasting

Abstract

The consumption forecasting of electric vehicles is needed to enhance the batteries and driving efficiency, as well as to support the intelligent transportation systems. Nevertheless, traditional predictive models are highly based on correlation-based learning that in most cases does not provide the actual causality between vehicle operating parameters and energy consumption. This paper suggests a causal inference predictive model of the electric vehicle energy consumption forecasting using the Kaggle Electric Vehicle Energy Consumption dataset. The framework combines learning causal structure, causal feature selection and predictive modeling to establish the actual cause-effect relationships that affect energy consumption. The system preprocesses the vehicle telemetry information, creates a causal graph and picks causally important features to train the predictive model. The suggested framework has a prediction accuracy of 98.91%, a Mean Absolute Error of 0.041, and a root of mean square error of 0.052, which are better than the traditional machine learning and deep learning models. The integration of the causal inference enhances the predictive reliability, minimizes error and increases interpretability. The suggested model gives precise and consistent forecasts of energy consumption under different conditions of driving. The framework will help advance energy-efficient and intelligent electric transportation systems by supporting intelligent battery management, route optimization and operation of the electric vehicles efficiently.