Wind Turbine Power Prediction

Overview

This project implements three approaches to predict wind turbine power output based on meteorological time series data:

1. Multilayer Perceptron (MLP) neural network
2. Genetic Algorithm (GA) optimization
3. Hybrid MLP-GA approach

The models use a sliding window technique to capture temporal patterns in weather variables for accurate power output forecasting.

Technical Details

• Framework: TensorFlow + Keras
• Loss Function: Mean Absolute Error (MAE)
• Evaluation Metrics:
  • Mean Squared Error (MSE)
  • Coefficient of Determination (R²)
  • Mean Absolute Error (MAE)

Features

• Time series preprocessing with configurable sliding window
• MLP neural network with optimized architecture
• Genetic Algorithm for feature selection and parameter optimization
• Hybrid approach leveraging GA for MLP weight initialization and architecture search
• Comprehensive evaluation metrics for renewable energy forecasting

Requirements

• Python 3.8+
• TensorFlow 2.17.0
• Keras 3.6.0
• NumPy 1.26.4
• Pandas 2.2.3
• scikit-learn 1.5.2
• Matplotlib 3.9.2
• Seaborn 0.13.2
• Jupyter notebooks environment

Additional dependencies are listed in requirements.txt.