TFE Energy: Village Data Analytics
More than 2 billion people live without reliable electricity supply. Over 210,000 mini-grids are needed to achieve the 7th Sustainable Development Goal (SDG): Ensure access to affordable, reliable, sustainable and modern energy for all.
VIDA is an AI-based, customized software for Earth observation. It enables data-driven investment, business, and policy decisions in rural villages in Africa and Asia.
Background of the project
The problem and the need
A central hurdle is identifying remote villages for off-grid electrification and gathering useful information about them. Data collection is slow, expensive, and inaccurate. This results in long project development times, low operating margins, and limited access to financing.
To scale off-grid electrification, we need to find commercially viable sites. We need a reliable, fast, and scalable method to identify commercially viable sites and provide the information to development organizations, governments, donors, and energy companies.
The solution
This software-based service automatically identifies remote villages and determines their suitability for a mini-grid system. VIDA uses machine learning algorithms to predict the socioeconomic health of a village. This data-driven knowledge reduces the risk of projects and saves time and costs for sustainable off-grid planning and comprehensive investments.
Machine learning algorithms in VIDA
appliedAI is developing AI resources with TFE Energy that contribute to automatic analysis and ranking by the software. The algorithms behind VIDA combine state-of-the-art supervised and unsupervised machine learning methods. The technical architecture of VIDA allows the algorithms to be retrained when new data is added to the system over time.
Based on an interesting area in a remote region, VIDA's custom AI algorithms identify villages that are not connected to the power grid. They use satellite images and other publicly available sources.
A second set of machine learning algorithms then predicts features such as size, density, and socio-economic composition based on the identified village boundaries. The algorithms are trained with satellite images and custom VIDA datasets based on proprietary data.
Finally, a ranking of villages is calculated based on village-level features. Users can filter by specific criteria and mark or save villages of interest for further analysis.