SMART WIND ENERGY PREDICTION SYSTEM USING ML

Abstract

Wind energy is one of the fastest-growing renewable energy sources globally, playing a
critical role in achieving sustainable energy targets and reducing dependence on fossil fuels.
Accurate prediction of wind turbine power output is vital for grid stability, energy scheduling,
and operational efficiency.
However, wind power is inherently intermittent and difficult to forecast due to the complex
and nonlinear relationships between meteorological parameters such as wind speed, wind
direction, ambient temperature, pressure, humidity, and the actual electrical power generated
by the turbine. This project presents the design, development, training, and deployment of a
Smart Wind Energy Prediction System using Artificial Intelligence. The system employs a
Random Forest Regressor model trained on real turbine SCADA (Supervisory Control and
Data Acquisition) sensor data containing features including wind speed, wind direction,
ambient temperature, atmospheric pressure, humidity, generator speed, generator winding
temperature, reactive power, and active power readings.
The dataset was preprocessed by removing missing values and dropping non-predictive
identifier columns (timestamp and turbine_id). The trained model was evaluated using Mean
Squared Error (MSE) and R-squared (R2) metrics to quantify prediction accuracy. The
trained model was serialized using Joblib for efficient deployment. An interactive Streamlit
web application was developed to enable real-time power output predictions based on user
provided turbine sensor inputs, making the system accessible to wind farm operators and
energy planners. The system demonstrates the viability of machine learning-based approaches
for smart energy forecasting and represents a cost-effective, scalable decision-support tool for
modern wind energy management