Comparative Analysis of Machine Learning Regression Models for Wind Turbine Power Prediction

Document Type

Conference Proceeding

Publication Date

12-8-2025

Department

Department of Mechanical and Aerospace Engineering

Abstract

In renewable energy systems, precise wind turbine power production forecasting is essential for efficient grid integration and operational planning. In this study, six Machine Learning (ML) regression models: Decision Tree (DT), Extra Tree (ET), Gradient Boosting (GB), Linear Regression (LR), Random Forest (RF), and Ridge Regression (RR), were compared for simulating low-voltage (LV) active power output using SCADA data. To increase prediction accuracy, several engineering features were used, including wind direction, speed (m/s), and their transformations. Determination coefficient (R2), mean absolute error (MAE), mean squared error (MSE), and root mean square error (RMSE) were used to assess performance. With the greatest test R2 of 0.9220 and the lowest RMSE of 0.1012, GB outperformed the other models. The robustness and interpretability of the model were further confirmed by visual diagnostics such as residual plots, confidence intervals, and feature importance analysis. These results supported the use of ensemble-based non-linear models in real-time energy management and highlighted their effectiveness in wind power forecasting.

Publication Title

2025 IEEE 19th International Conference on Application of Information and Communication Technologies (AICT)

Share

COinS