Using machine learning to expound energy poverty in the global south: Understanding and predicting access to cooking with clean energy

Abstract

Efforts towards achieving high access to cooking with clean energy have not been transformative due to a limited understanding of the clean-energy drivers and a lack of evidence-based clean-energy policy recommendations. This study addresses this gap by building a high-performing machine learning model to predict and understand the mechanisms driving energy poverty - specifically access to cooking with clean energy. In a first-of-a-kind, the estimated cost of US$14.5 trillion to enable universal access to cooking with clean energy encompasses all the intermediate inputs required to build self-sufficient ecosystems by creating value-addition sectors. Unlike previous studies, the data-driven clean-cooking transition pathways provide foundations for shaping policy and building energy models that can transform the complex energy and cooking landscape. Developing these pathways is necessary to increase people's financial resilience to tackle energy poverty. The findings also show the absence of a linear relationship between electricity access and clean cooking - evidencing the need for a rapid paradigm shift to address energy poverty. A new fundamental approach that focuses on improving and sustaining the financial capacity of households through a systems approach is required so that they can afford electricity or fuels for cooking.