"background": "Chronic power outages and low reliability indices in sub-Saharan African distribution networks necessitate robust, data-driven evaluation frameworks. Existing reliability assessments often lack rigorous causal inference designs suitable for isolating the impact of specific infrastructure interventions from confounding factors. ", "purpose and objectives": "This working paper develops and applies a quasi-experimental methodology to evaluate the causal impact of equipment upgrades on the reliability of Rwanda's power-distribution system. It aims to establish diagnostic protocols for network optimisation and to model the relationship between specific interventions and key performance indicators. ", "methodology": "A difference-in-differences framework is employed, comparing treated and control feeder groups before and after the roll-out of automated reclosers and fault indicators. The core statistical model is Y{ft = \0 + \1 (Treatf \ Postt) + \ + \ +, where Y₅ₓ is the SAIDI for feeder f in period t. Inference is based on cluster-robust standard errors at the substation level. ", "findings": "Preliminary diagnostic analysis indicates a marked spatial heterogeneity in failure rates, with one region exhibiting fault densities approximately 40% higher than the national average. The full quasi-experimental results are pending final data collection and model validation; this section will be updated in the final paper. ", "conclusion": "The proposed quasi-experimental design provides a viable, rigorous framework for attributing reliability improvements to specific capital projects within a complex grid network, moving beyond descriptive correlation. ", "recommendations": "Utilities should adopt quasi-experimental designs for post-implementation review of major investments. Priority for network hardening should be given to feeders identified with persistently high fault densities in diagnostic screening. ", "key words": "power distribution, reliability engineering, quasi-experimental design, difference-in-differences, network diagnostics, causal inference, Rwanda", "contribution statement": "This paper introduces a novel application of causal inference methods to power-system
Mukamurenzi et al. (Sun,) studied this question.