Solar Power Generation Forecasting

Machine learning-based solar energy forecasting system using the Anikannal Solar Power Generation Dataset.

Features

• Time-series forecasting with weather data
• Model comparison (Linear Regression vs Random Forest)
• Interactive Streamlit web interface
• Temporal feature engineering (lag, rolling mean)

Dataset

• Source: Anikannal Solar Power Generation Dataset (Plant 1)
• Generation Data: 15-minute intervals
• Weather Data: Hourly measurements
• Merge Strategy: Time-aware merge using merge_asof

Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Train Models

python train_model.py

3. Run Streamlit App

streamlit run app.py

Models

Linear Regression (Baseline)

• MAE: ~16.80
• RMSE: ~32.10

Random Forest (Final Model)

• MAE: ~16.75
• RMSE: ~32.04

Project Structure

new_aiml/
├── app.py                      # Streamlit web app
├── train_model.py              # Model training script
├── requirements.txt            # Dependencies
├── linear_model.pkl            # Trained Linear Regression
├── random_forest_model.pkl     # Trained Random Forest
├── feature_names.pkl           # Feature list
└── README.md                   # Documentation

Features Used

1. Weather Features: Ambient Temperature, Module Temperature, Irradiation
2. Temporal Features: Hour, Month, Day of Week
3. Lag Features: Previous AC Power (lag_1)
4. Rolling Features: 3-period rolling mean

Methodology

1. Data Preprocessing: Time-aware merge, remove night periods (AC_POWER = 0)
2. Feature Engineering: Extract temporal features + lag/rolling features
3. Train-Test Split: 80-20 chronological split (no shuffle)
4. Model Training: Linear Regression + Random Forest
5. Evaluation: MAE & RMSE metrics

Deployment

Deploy on Streamlit Cloud:

1. Push to GitHub
2. Connect repository to Streamlit Cloud
3. Deploy with one click

Author

Built for AI/ML Course Project

License

Educational Project