Machine Learning for Renewable Energy Forecasting (2024)

Links

https://energyforecasting.vercel.app/ https://energyforecaster.streamlit.app/

https://github.com/jtwirly/energygenerationforecastdashboard

Machine Learning for Renewable Energy Forecasting

Project Overview

This is the final class project for MIT class 1.125: Architecting and Engineering Software Systems Fall 2024.
The goal of the project is to create an energy forecasting app that is trained on historical weather data and recorded energy production values. Given weather forecasts, the app predicts energy generation and demand for corresponding times.

Energy Generation Forecast Dashboard

The Energy Generation Forecast Dashboard is a comprehensive solution that empowers grid operators to make informed decisions and optimize their renewable energy operations. Powered by advanced AI and machine learning, the dashboard delivers precise, real-time predictions for wind, solar, and demand patterns, enabling operators to balance supply and demand, maximize renewable energy generation, and minimize costly inefficiencies.

This dashboard provides energy generation forecasts using machine learning models trained on historical data.

Features:

-Solar generation prediction

-Wind generation prediction

-Demand forecasting

-Generation mix analysis

-Timezone support (Currently showing: NE (Eastern Time))

Data Sources:

-Historical weather data

-Past generation records

-Demand patterns

Key Features

• Accurate forecasting for wind, solar, and demand using state-of-the-art XGBoost regression models
• Interactive charts and visualizations built with Streamlit and Plotly for a seamless user experience
• Ability to dynamically switch between different prediction views (wind, solar, demand) to gain a holistic understanding of the energy landscape
• Grid optimization insights, including peak generation and demand, to support load balancing and energy storage strategies
• Robust, scalable backend architecture built with Python and SQLite database
• Containerization with Docker for improved portability and scalability
• Branded website built with NextJS

Technology Stack

• Backend: Python, Streamlit, SQLite
• Machine Learning: XGBoost, sklearn
• Visualization: Plotly, Lucide Icons
• Containerization: Docker
• Frontend: NextJS