The rural electrification planning problemstrategies and solutions

Résumé

Universal Access to Energy is one of the most significant challenges of our time, and energy is an enabling factor that fosters development in several fields such as education and healthcare. The United Nations’ seventh Sustainable Development Goal (SDG7) acknowledges the importance of energy access, and it establishes the target of achieving universal access to modern forms of energy that are affordable, reliable, and sustainable by 2030. Significant efforts are imperative to meet this deadline as there are approximately 840 million people that currently do not have access to electricity. Establishing an electrification agenda is a complex task that depends on many socio-political factors. A suitable electrification plan should rely on solid hypotheses, rigorous analysis, and accurate data. Computer-based models have recently gained momentum in electrification planning, as they can identify the lowest-cost designs that provide desired levels of electricity access in large-scale areas. The automated calculation of the designs can help optimize the allocation of resources devoted to universal electricity access, expediting development. In this thesis, we focus on one electrification planning tool: the Reference Electrification Model (REM). REM determines the least-cost electrification mode for each consumer (i.e., a standalone system, a mini-grid, or an extension of the power grid). REM calculates detailed technical designs at the building level, optimizing the generation of off-grid systems and the networks of mini-grids and grid extensions. REM is the result of ongoing teamwork. The first prototype of REM was presented in the master thesis of Douglas Ellman, which was defended at MIT, Cambridge, Massachusets, USA, in 2015. This first prototype is the starting point of this thesis. The first prototype of REM provided inconsistent results, and substantial efforts were devoted to scrutinizing and improving its algorithms. The first part of this thesis describes several upgrades implemented into the first prototype of REM, which resulted in robust performance after the upgrades. The second part of this thesis focuses on the development of new algorithms in REM. We present a novel method that quickly estimates the network cost of any potential low-voltage mini-grid that could appear in the solution of a large-scale planning case. We also present two clustering algorithms. The first clustering algorithm groups the consumers into mini-grids, and the second one determines which consumers should be electrified with extensions of the power grid. The new algorithms provide more optimal results than the original algorithms of REM or present other advantages.