Grid Outage Prediction & Restoration Time Estimator

A full-stack AI-powered utility dashboard that transforms how utility companies respond to extreme weather events. By combining real-time weather data with physics-based machine learning models, this system predicts grid failures before they happen and optimizes crew dispatch.

---

📸 Screenshots

| Live Risk Map |

🏗️ Architecture

The system follows a modern microservices architecture containerized with Docker.

graph TD
    WeatherAPI[Open-Meteo API] -->|Polls Data| Ingestion[Ingestion Service]
    Ingestion -->|Stores History| DB[(PostgreSQL)]
    DB -->|Historical Data| ML[ML Service]
    ML -->|Training & Inference| ML
    User[User] -->|Interacts| Frontend[React Frontend]
    Frontend -->|Requests Predictions| API[FastAPI Backend]
    API -->|Fetches Live Risk| ML
    API -->|Reads/Writes| DB

Loading

---

🧠 The Brain: Predictive ML Engine

At the core is a Random Forest Classifier trained on physics-based ground truth data.

1. Data Generation: Synthetic training data simulates failure physics (Wind Load, Thermal Derating, Asset Aging).
2. Training: The model learns non-linear interactions (e.g., Old Asset + High Wind = Critical Risk).
3. Real-Time Inference:
   • Polls live weather for the service area.
   • Feeds data into the model to generate a live Outage Probability (0-100%).

---

� Key Features

• � Interactive Risk Map: Visualizes 200+ distribution feeders, color-coded by real-time risk severity.
• 🌪️ Live Weather Integration: Reacts instantly to changing weather conditions in the Bay Area via Open-Meteo.
• 🧪 Simulation Mode: Inject "What-If" scenarios (e.g., "Simulate a Category 3 Hurricane") to test grid resilience.
• ⏱️ Automated ETR: Instantly calculates Estimated Time of Restoration based on failure severity.
• 📈 Historical Tracking: Records predictions for long-term trend analysis.

---

🛠️ Tech Stack

Frontend

• React + Vite: Blazing fast UI development.
• Leaflet: Interactive mapping for grid visualization.
• TailwindCSS: Modern, responsive styling.
• Chart.js: Data visualization for risk trends.

Backend

• FastAPI: High-performance, async Python API.
• Scikit-learn: Random Forest implementation for prediction.
• Pandas/NumPy: Data manipulation and processing.

Infrastructure

• Docker Compose: Orchestrates API, Database, and Background Workers.
• PostgreSQL: Robust, relational storage for time-series data.

---

Start

Option A: Cloud Mode (Recommended)

Run the full stack (Frontend, Backend, DB, Workers) with one command:

docker-compose up --build

Access the dashboard at http://localhost:5173

Option B: Local Development

1. Start the Backend

cd backend
pip install -r requirements.txt
python -m uvicorn app.main:app --reload --host 0.0.0.0 --port 8081

2. Start the Frontend

cd frontend
npm install
npm run dev or cmd /c "npm run dev'

---

📂 Project Structure

PCD Grid project/
├── backend/                # FastAPI Application
│   ├── app/
│   │   ├── api/            # API Routes
│   │   ├── services/       # ML & Weather Logic
│   │   └── models/         # Pydantic & DB Models
│   ├── Dockerfile
│   └── requirements.txt
├── frontend/               # React Application
│   ├── src/
│   │   ├── components/     # UI Components (Map, Widgets)
│   │   └── services/       # API Client
│   └── Dockerfile
├── docker-compose.yml      # Orchestration
└── README.md               # You are here

---