⚡ GridGuardian AI

AI-Powered Smart Grid Intelligence Platform

---

🎯 Overview

GridGuardian AI is a production-ready intelligent platform that predicts electricity demand, forecasts solar generation, detects grid anomalies, and generates AI-driven recommendations — all through a cyberpunk-themed, futuristic dashboard.

---

🚀 Features

Module	Description
⚡ Demand Forecast	Predicts next-hour MW demand using RF/XGBoost/LightGBM
☀️ Solar Forecast	Forecasts solar AC output from inverter generation patterns
🚨 Anomaly Detection	Isolation Forest detects spikes and grid instabilities
💡 Recommendations	AI-generated actionable grid management advice
🧪 Simulation Lab	What-if scenario testing with sliders
📊 Analytics	Full model performance comparison and leaderboard
📜 History	Complete audit trail of all predictions and events

---

🗂 Datasets Used

Dataset	Rows	Purpose
AEP_hourly.csv	121,273	Demand forecasting (AEP power company, 2004–2018)
Plant_1_Generation_Data.csv	68,778	Solar energy generation
smart_grid_stability_augmented.csv	60,000	Grid stability classification
DailyDelhiClimateTest.csv	114	Weather context features

---

🛠 Tech Stack

• Frontend: HTML5, CSS3, Bootstrap 5, Chart.js, Font Awesome
• Backend: Flask 3, Python 3.11
• ML: Random Forest, XGBoost, LightGBM, Isolation Forest
• Database: SQLite3
• Visualisation: Matplotlib
• Deployment: Render (free tier)

---

⚙️ Local Setup

1. Clone & Install

git clone https://github.com/yourusername/GridGuardianAI.git
cd GridGuardianAI
pip install -r requirements.txt

2. Add Datasets

Place the following CSV files in the project root:

AEP_hourly.csv
Plant_1_Generation_Data.csv
smart_grid_stability_augmented.csv
DailyDelhiClimateTest.csv

3. Train Models

python train_model.py

This will:

• Load and preprocess all 4 datasets
• Engineer 15+ features per dataset
• Train Random Forest, XGBoost, LightGBM for each task
• Auto-select the best model per task
• Save models to models/
• Generate 8 charts to static/charts/

Expected output:

★ Best demand model: RandomForest (RMSE=21.9)
★ Best solar model: RandomForest (RMSE=1153.6)
Grid stability classifier accuracy: 1.0000
✅ TRAINING COMPLETE

4. Run the App

python app.py

Visit: http://localhost:5000

Register a new account and start using the dashboard.

---

🌐 Render Deployment

Option A: render.yaml (recommended)

1. Push your repo to GitHub (including CSVs and trained .pkl files)
2. Go to render.com → New Web Service
3. Connect GitHub repo → Render auto-detects render.yaml
4. Click Deploy

The build command runs python train_model.py automatically.

Option B: Manual

• Build command: pip install -r requirements.txt && python train_model.py
• Start command: gunicorn app:app --bind 0.0.0.0:$PORT --workers 2 --timeout 120
• Runtime: Python 3.11

Note: On Render's free tier, models are re-trained on every deploy. For faster deploys, commit your models/*.pkl files to Git.

---

📁 Project Structure

GridGuardianAI/
├── static/
│   ├── css/
│   │   ├── style.css          # Core cyberpunk theme
│   │   ├── dashboard.css      # Dashboard-specific styles
│   │   └── animations.css     # All animations
│   ├── js/
│   │   ├── script.js          # Clock, counters, sidebar
│   │   ├── dashboard.js       # Chart.js dashboard charts
│   │   └── simulation.js      # Simulation lab sliders/gauge
│   ├── charts/                # Matplotlib-generated PNGs
│   └── images/
├── templates/
│   ├── base.html              # Master layout + sidebar
│   ├── login.html             # Auth
│   ├── register.html          # Auth
│   ├── dashboard.html         # Main KPI dashboard
│   ├── demand_forecast.html   # Demand prediction
│   ├── renewable_forecast.html# Solar prediction
│   ├── anomaly_detection.html # Anomaly scanner
│   ├── recommendation_center.html
│   ├── simulation_lab.html    # What-if scenarios
│   ├── analytics.html         # Model leaderboards
│   └── history.html           # Audit trail
├── models/                    # Trained .pkl files
├── database/                  # SQLite users.db
├── train_model.py             # Full ML pipeline
├── app.py                     # Flask application
├── requirements.txt
├── Procfile
├── runtime.txt
├── render.yaml
└── README.md

---

🧠 ML Model Performance

Demand Forecasting (120,842 training samples)

Model	MAE (MW)	RMSE (MW)	R²
RandomForest ★	9.4	21.9	0.9999
XGBoost	46.4	64.2	0.9993
LightGBM	47.4	65.8	0.9993

Solar Forecasting (3,134 training samples)

Model	MAE (kW)	RMSE (kW)	R²
RandomForest ★	400.8	1153.6	0.9791
XGBoost	414.7	1180.3	0.9782
LightGBM	463.6	1197.0	0.9775

Grid Stability Classifier

Model	Accuracy
Random Forest	100%

---

🔒 Security

• Passwords are SHA-256 hashed before storage
• Session management via Flask sessions
• All routes protected with @login_required
• Input validation on all forms

---

👤 Author

Built for AI/ML Hackathon — GridGuardian AI
Gowshalya P

---

📄 License

MIT License — free to use, modify, and deploy.