abdulwehab/somalia-energy-ai

AI-powered tool for forecasting hourly electricity demand across Somali regions. Built with Python, Streamlit, and machine learning to support national energy planning and mini-grid optimization. # 🇸🇴 Somalia National Load Forecasting AI Model This project builds a machine learning model to forecast hourly electricity demand across major Somali regions. It supports rural mini-grid optimization and aligns with SESRP and ASCENT energy modernization goals. --- ## 📊 Dataset Synthetic data (Jan–June 2024) for 10 regions: - Mogadishu, Hargeisa, Bosaso, Kismayo, Garowe - Baidoa, Beledweyne, Marka, Galkayo, Dhusamareeb Each record includes: - `timestamp`, `region`, `temperature`, `humidity`, `solar_input`, `population`, `holiday_flag`, `load_kWh` --- ## 🤖 Model - **Algorithm**: Random Forest Regressor - **MAE**: 0.77 kWh - **RMSE**: 1.19 kWh - **R²**: 0.78 Trained on weather, solar, time, and population inputs per region. --- ## 💻 Streamlit Dashboard Predict regional electricity demand interactively. ### Run locally: ```bash streamlit run streamlit_app_national_forecast.py ``` ### Files required: - `national_load_forecast_model.pkl` - `somalia_national_load_dataset.csv` - `requirements.txt` --- ## 🌍 Use Cases - Forecast evening peak loads in rural mini-grids - Optimize solar-diesel hybrid planning - Support ESPs with real-time demand insights --- ## 📝 Concept Note & Pilot A 1-page PDF proposal is available for submission to: - Ministry of Energy & Water Resources - SESRP / ASCENT - Donor organizations (USAID, GIZ, WB) --- ## 📁 Files Included | File | Purpose | |------|---------| | `streamlit_app_national_forecast.py` | Interactive dashboard | | `national_load_forecast_model.pkl` | Trained ML model | | `somalia_national_load_dataset.csv` | Regional synthetic dataset | | `requirements.txt` | Package list | | `README.md` | This file | --- ## 📜 License MIT License – free to use with attribution. --- ## 👨‍💻 Developed by Abdiwahab Khalif Jama Electrical & Electronics Engineer | AI for Energy Access