nanayawjoshua/streemline-predictive-oms

Predictive Outage Management System — Streemline. AI-powered outage prediction at zone and transformer level for ECG Ghana. WhatsApp-first, human-in-the-loop dispatch. # Streemline Predictive OMS **Outcome:** Predict power outages at zone and transformer level before they happen. Built for the Electricity Company of Ghana (ECG). WhatsApp-first customer interface. Human-in-the-loop dispatch. Continuous learning. ## Architecture ``` ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ WhatsApp │────▶│ Go Bridge │────▶│ Prediction │ │ (whatsmeow) │ │ Connector │ │ Engine │ └──────────────┘ └──────────────┘ │ (FastAPI) │ │ :8001 │ ┌──────────────┐ ┌──────────────┐ └──────┬───────┘ │ ECG Ops │◀────│ Web Admin │ │ │ WhatsApp │ │ Dashboard │ ┌──────▼───────┐ └──────────────┘ └──────────────┘ │ Weather API │ │ (OpenWeather)│ └──────────────┘ ``` ## Quick Start ```bash # All services make dev # Prediction engine only cd prediction-engine pip install -r requirements.txt uvicorn app.main:app --reload --port 8001 # Dashboard only make dev-dashboard ``` ## Prediction Endpoints | Method | Path | Description | |--------|------|-------------| | GET | `/health` | Service health + feature flags | | POST | `/v1/predict/zone` | Zone-level outage risk (6h window) | | POST | `/v1/predict/transformer` | Transformer failure probability (48h) | | POST | `/v1/recommend/dispatch` | HITL dispatch recommendation | ## Deployment ```bash docker compose up --build -d ``` ## ML Roadmap | Phase | Timeline | Model | Inputs | |-------|----------|-------|--------| | 1 | Week 1 | Rules | Weather + outage density | | 2 | Month 3 | XGBoost | + historical outages, load data | | 3 | Month 6 | GNN | + grid topology, cascade effects |