THakgvLO/satnac-openserve-project

This is a project for the SATNAC Industry Solutions Challenge 2025. The topic is on building a system that will allow Openserve to determine viable EV-charging sites around South Africa using Artificial Intelligence. # SATNAC - Openserve EV Charging Site Decision Support (Finalist) Finalist for the SATNAC Industry Solutions Challenge 2025. ## Project summary A strategic decision‑support tool to guide Openserve’s transition to Electric Vehicles (EVs) in South Africa. The system evaluates potential charging locations (1,500+ candidates) using a hybrid geospatial + AI approach and produces a prioritized rollout segmentation (Green / Amber / Red) to help planners focus investments. Key highlights: - Evaluated 1,500+ candidate sites with a 7‑factor weighted scoring system. - Factors: loadshedding risk, security, transformer capacity, operational cost, grid resilience, site footprint, connectivity. - Used K‑Means clustering on weighted scores to generate Green/Amber/Red rollout categories. - Delivered as a Python ML pipeline with an interactive web dashboard (Leaflet.js for maps, Chart.js for visualizations) for geospatial exploration and explainability. - Reached the SATNAC 2025 finale - demonstrated strong modelling and explainability despite methodological limitations. ## What the project does - Ingests geospatial site candidates and auxiliary datasets (grid assets, security layers, connectivity, transformer specs, cost estimates). - Computes normalized scores per site across 7 decision factors. - Applies weighted aggregation and K‑Means clustering to group sites into rollout categories. - Exposes results via an interactive map dashboard supporting exploration, filtering, and basic explainability charts. ## Technical stack - Python (data processing, feature engineering, ML pipeline) - Common Python libs: pandas, geopandas, scikit‑learn, rasterio (or similar) - Web dashboard: Leaflet.js + Chart.js (frontend), lightweight Python server (Flask/FastAPI/Streamlit - adapt to repo) - GIS data formats: GeoJSON / Shapefiles, CSVs for attributes - Tested on Windows (development environment notes below) ## Why combine AI and GIS - Geospatial context is essential for EV planning: proximi …