Evaluating machine learning algorithms for energy consumption prediction in electric vehicles: A comparative study

Area of Science:

• Electrical Engineering
• Computer Science
• Data Science

Background:

• Increasing electric vehicle (EV) adoption necessitates accurate energy consumption predictions for robust power grid management.
• Effective grid management requires precise forecasting of energy demand influenced by EV charging patterns.

Purpose of the Study:

• To evaluate the performance of eleven distinct machine learning models for electric vehicle energy consumption prediction.
• To identify the most effective model for forecasting energy usage based on historical data.

Main Methods:

• Utilized eleven machine learning models: Ridge Regression, Lasso Regression, K-Nearest Neighbors, Gradient Boosting, Support Vector Regression, Multi-Layer Perceptron, XGBoost, CatBoost, LightGBM, Gaussian Processes for Regression (GPR), and Extra Trees Regressor.
• Evaluated models using Mean Absolute Error (MAE), Mean Squared Error (MSE), R², Root Mean Squared Error (RMSE), and Normalized Root Mean Squared Error (NRMSE).
• Performed visual analyses using scatter plots and time series plots for comprehensive model assessment.

Main Results:

• The Extra Trees Regressor demonstrated superior performance, achieving an MAE of 0.5888, MSE of 3.2683, R² of 0.9592, RMSE of 1.8078, and NRMSE of 0.020.
• Gradient Boosting and K-Nearest Neighbors (KNN) also yielded strong results, though with slightly higher variability.
• Non-linear and linear models showed limitations in predicting extreme energy consumption levels, indicating challenges in capturing complex interactions.

Conclusions:

• Extra Trees Regressor is highly effective for electric vehicle energy consumption prediction, offering significant improvements for power grid management.
• Ensemble and advanced non-linear machine learning models show promise in capturing intricate time series patterns for energy demand forecasting.
• Further research into these models can enhance the accuracy of energy projections and support grid stability with rising EV integration.