A web-based machine learning application to forecast wind energy generation using time-series feature engineering and a pre-trained XGBoost regression model. Built with Flask and visualized through Matplotlib plots.
Wind energy is a crucial source of renewable energy in today's sustainable world.
This project predicts future energy generation from wind turbines based on timestamp-driven features like hour, day, and month — enabling better energy planning and optimization.
Users can dynamically input a date range through a web interface and receive a real-time prediction plot for the specified period.
- Interactive Web App: Users input a date range to forecast energy generation.
- Machine Learning Model: Pre-trained XGBoost Regressor (
model2.pkl) for high-accuracy predictions. - Automatic Feature Engineering: Generates time-based features from timestamps (hour, day, month, etc.).
- Visualization: Real-time prediction graphs created with Matplotlib and displayed via Flask.
- Lightweight Backend: Simple Flask server handles user requests, model inference, and visualization generation.
- Backend: Python, Flask
- Machine Learning: XGBoost, Pickle
- Data Processing: Pandas, NumPy
- Visualization: Matplotlib, Seaborn
- Frontend: HTML (via Flask Jinja2 Templates)
| Component | Description |
|---|---|
model.py |
Loads pre-trained model, handles feature engineering, prepares future timestamps |
app.py |
Flask app managing routes, user inputs, prediction logic, and plot generation |
/templates/index.html |
Frontend web form for date input |
/static/output.png |
Saves generated prediction plot |
-
User Input:
- User selects a From Date and To Date via the web form.
-
Feature Generation:
- The app creates a date range with 10-minute intervals between the two dates.
- Features such as
hour,minute,day,month,year, etc., are automatically generated.
-
Prediction:
- The feature set is fed into the pre-trained XGBoost model to predict future energy output.
-
Visualization:
- Predictions are plotted as a time-series graph and saved dynamically.
-
Result Display:
- The generated graph is displayed back to the user on the same page.
User selects:
From Date: 2025-04-20
To Date: 2025-04-22
-> System generates 10-min interval timestamps
-> Features extracted from timestamps
-> Model predicts energy output at each interval
-> Graph created showing energy trends between April 20–22
-> Graph displayed on the web page
- Deploy on cloud platform (AWS/GCP) for wider access.
- Add user authentication to save prediction history.
- Improve model accuracy by integrating external weather features (wind speed, humidity, temperature).
- Make plots interactive with Plotly or Dash.