BillalHassan1144/short-term-wind-power-forecast

Wind Power Forecasting FastAPI App

This project is a machine learning-based application for forecasting wind power generation. It serves a predictive model through a FastAPI backend and provides a simple web interface for interaction.

Features

• —Multiple Models: Utilizes an ensemble of models including XGBoost, LightGBM, CatBoost, and LSTM for robust predictions.
• —FastAPI Backend: High-performance, easy-to-use API framework.
• —Web Interface: Simple HTML/JS frontend to input features and view predictions.
• —Real-time Prediction: Returns power output estimates in Kilowatts (KW).

Prerequisites

• —Python 3.8+
• —pip

Installation

1. 1.Clone the repository:

   git clone https://github.com/billalhassan793/short-term-wind-power-forecasting.git
   cd short-term-wind-power-forecasting

1. 1.Create a virtual environment (recommended):

   python -m venv venv
   # Windows
   .\venv\Scripts\activate
   # macOS/Linux
   source venv/bin/activate

1. 1.Install dependencies:

   pip install -r requirements.txt

Usage

1. 1.Start the server:

   uvicorn app:app --reload --port 8000

1. 1.Access the application: Open your browser and navigate to http://127.0.0.1:8000.

1. 1.API Documentation: The interactive API docs are available at http://127.0.0.1:8000/docs.

Project Structure

• —app.py: Main FastAPI application entry point.
• —models/: Directory containing trained model files (.pkl, .cbm, .keras) and scalers.
• —templates/: Contains index.html for the frontend.
• —requirements.txt: List of Python dependencies.

Inputs

The model accepts the following features for prediction:

• —Wind Speed (m/s)
• —Theoretical Power Curve (KWh)
• —Wind Direction (degrees)
• —Time features (Hour, Day of Week)
• —Lag features (Power, Wind Speed)
• —Rolling Mean features

Prediction Determinism

By default, /predict_smart now runs in stateless mode so the same input returns the same output.

• —Default: STATEFUL_FEATURE_HISTORY=0 (deterministic requests)
• —Optional: set STATEFUL_FEATURE_HISTORY=1 to enable cross-request lag history behavior