In our previous feature on Lamwo, we highlighted how data-driven planning contributed to the district’s reliable electricity supply. Behind that success is a workflow that quietly powers every decision, from identifying settlements to placing mini-grid sites. This delves deeper into that process, showing how AI transforms raw satellite imagery into actionable planning insights that reach the communities who need power most.
Seeing Growth from Space.
The starting point for planning was visibility. Instead of relying on population estimates that can quickly become outdated, the team used Google’s Open Buildings dataset to identify individual building footprints visible in satellite imagery. By applying machine learning to a timeseries dataset of satellite images of building counts over time, the system could estimate the rate of expansion in villages and compare growth patterns across hundreds of locations.
This provided planners with a consistent, evidence-based view of settlement growth and helped determine which areas were most likely to benefit from electrification. In Lamwo, this step turned years of satellite images into measurable community-level growth trends.
Turning Data into Priorities
Once we could see where growth was happening, the next question was where to begin electrification. To answer this, Sunbird AI, working with the Ministry of Energy and GIZ, developed a machine learning ranking model based on the Random Forest algorithm.
The model integrates multiple datasets, including building density, road access, schools, health centers, land use, and renewable energy resources such as solar and wind. The Random Forest machine learning algorithm was selected for its ability to weigh many variables at once and produce a transparent, data-driven ranking of villages. The result is a fair and reproducible way to identify communities most likely to benefit from new mini-grids.
These ranked outputs, available on the tool developed by Sunbird AI, the Lamwo Electrification Platform, help planners visualize priorities and make investment decisions guided by evidence rather than assumptions.
Refining Site Placement
Even after identifying the right villages, placement within each settlement matters. Early candidate sites were sometimes positioned far from where most homes were clustered. To address this, the team applied DBSCAN(Density-Based Spatial Clustering of Applications with Noise), a clustering method that identifies the densest groups of buildings.
By repositioning mini-grid sites toward these clusters, the workflow ensured that electricity would reach the greatest number of households and community institutions. Field validation showed that repositioned sites align closely with household clusters, improving accessibility and supporting more inclusive electrification planning.
From Analysis to Interaction
The workflow also includes an LLM-powered geospatial assistant, a large language model integrated directly into the planning interface. On the Lamwo Electrification Platform, users can draw any area on the map and ask natural language questions such as “What is the agricultural potential here?” or “How developed is the infrastructure in this zone?”
The assistant retrieves relevant datasets, analyzes them, and responds in clear, conversational language. This makes complex geospatial analysis intuitive, even for planners without technical backgrounds. It bridges the gap between AI systems and human decision-making, turning advanced computation into an accessible planning tool.
Validation and Scale.
To verify that the model reflected reality, its predictions were compared with field survey data from Lamwo. The close match between AI outputs and on-the-ground findings confirmed that the workflow reflected community growth patterns and site suitability.
Because it was built with open data, code, and a modular architecture, the same system can be scaled to other regions in Uganda or adapted for new energy access projects.
From Data to Decisions.
Every component, from satellite imagery and regression models to clustering algorithms and LLM assistants, contributes to a single goal: turning data into better planning decisions.
Our workflow demonstrates how AI can strengthen national electrification efforts by making energy planning more transparent, data-driven, and inclusive. What began in Lamwo offers a practical blueprint for scaling across Uganda, transforming pixels from satellites into brighter nights for communities, schools, and businesses.
Explore more about the project.
We welcome collaborations with institutions working to strengthen evidence-based planning and data-driven decision-making across Africa. To learn more or explore partnership opportunities, contact us at info@sunbird.ai