⚡ Smart Grid Nexus: Real-Time Energy Telemetry & AI Forecasting

An enterprise-grade, high-performance telemetry dashboard designed to simulate real-time electrical grid monitoring and AI-driven load forecasting. Built with a robust Machine Learning pipeline and a cutting-edge React-based Python frontend.

🌌 Overview

Managing modern energy grids requires lightning-fast data processing and highly accurate predictions. Smart Grid Nexus processes historical hourly energy consumption data from various eastern US regions, trains highly optimized XGBoost models, and serves the results in a simulated real-time, zero-latency streaming dashboard.

The UI features a modern Cyberpunk / Glassmorphism aesthetic, utilizing custom CSS overrides to deliver a dark-mode, neon-accented experience typical of advanced aerospace and industrial control rooms.

✨ Key Features

• 🧠 AI-Powered Forecasting: Utilizes XGBoost regressors optimized with time-series feature engineering (lags, rolling means, datetime properties) to predict energy loads with high precision.
• 📡 Real-Time Telemetry Simulation: Smooth, interval-based data streaming that mimics live sensor feeds.
• 🕹️ Advanced State Management: Includes Play, Pause, and Reset functionality. The stream can be paused for deep visual analysis and resumed seamlessly without losing the index state.
• 🎨 Glassmorphism UI: A custom-styled, dark-theme interface built with dash-bootstrap-components featuring backdrop blurs, neon typography (Roboto Mono), and aggressive CSS tooltip overrides.
• 🌍 Multi-Region Scalability: Dynamically switch between 12 different regional datasets (e.g., PJM, AEP, COMED, DAYTON) with localized model caching.
• ⚡ High-Performance Rendering: Uses @app.callback and dcc.Store for memory-efficient client-side state management, preventing the UI from freezing during continuous updates.

🛠️ Technology Stack

Machine Learning Pipeline

• Pandas & NumPy: For robust data wrangling and feature engineering.
• Scikit-Learn: Model evaluation (MAE, RMSE, R2) and data splitting.
• XGBoost: The core gradient boosting algorithm for time-series forecasting.
• Joblib: Fast serialization of trained models.

Frontend & Telemetry

• Dash by Plotly: React.js powered framework for Python, enabling reactive, component-based UI without writing JavaScript.
• Plotly Graph Objects: For rendering hardware-accelerated, high-framerate interactive charts.
• Dash Bootstrap Components (DBC): Responsive grid layouts utilizing the CYBORG theme.

📂 Project Architecture

Smart_Grid_Energy_Forecasting/
├── app/
│   └── app.py                  # The main Dash telemetry application
├── data/
│   └── *_hourly.csv            # Raw hourly dataset files (AEP, COMED, PJM, etc.)
├── models/
│   └── *_forecast.pkl          # Serialized XGBoost models per region
├── outputs/
│   ├── *_processed.csv         # Feature-engineered datasets
│   └── *_forecast_results.csv  # Model evaluation results
├── notebooks/
│   ├── 01_eda.ipynb            # Exploratory Data Analysis
│   ├── 02_feature_engineering.ipynb # Time-series feature extraction
│   ├── 03_training.ipynb       # XGBoost model training and evaluation
│   └── 04_evaluation.ipynb     # Visualizing metrics and error distributions
├── requirements.txt            # Python dependencies
└── README.md

🚀 Installation & Quick Start

1. Clone the repository:

  git clone [https://github.com/yourusername/Smart-Grid-Nexus.git](https://github.com/yourusername/Smart-Grid-Nexus.git)
  cd Smart-Grid-Nexus

2. Install dependencies:

pip install -r requirements.txt

3. Run the Machine Learning Pipeline: Execute the Jupyter notebooks in the notebooks/ directory sequentially to generate the processed data and train the regional models.

4. Launch the Telemetry Dashboard:

cd app
python app.py

Navigate to http://127.0.0.1:8050 in your web browser.

📊 Dataset Attribution

This project utilizes the Hourly Energy Consumption dataset provided by PJM Interconnection LLC (via Kaggle). PJM operates the largest competitive wholesale electricity market in the US.

📄 License

Distributed under the MIT License. See LICENSE for more information.