Energy reliability and feasibility analysis for battery-constrained off-grid solar street lighting: an uncertainty-aware dimming control framework

Off grid solar street lighting serves millions of people in energy-scarce communities where grid access is unavailable, yet battery aging and uncertain nightly discharge endurance make reliable service difficult to guarantee. Strict blackout budgets can be physically unattainable under minimum comfort floor constraints, a feasibility limit that existing control frameworks do not explicitly diagnose. We present a simulation-based benchmarking framework that links cycle-level battery discharge forecasting to phase-specific brightness control, and introduces two contributions to the energy reliability analysis of off grid lighting systems: (i) comfort-floor feasibility bounds that distinguish physically infeasible blackout targets from controller shortcomings, and (ii) blackout severity evaluation via unmet minutes conditional on blackout (UM|BO p90), which exposes tail-risk differences invisible to frequency-only metrics. Forecast uncertainty is handled through a model agnostic residual quantile safety buffer with battery-ID grouped splits to reduce leakage. Across repeated grouped battery splits, the floors-only feasibility estimate was 0.86% [95% CI: 0.00, 3.26] for H = 0.75 h, 6.06% [2.07, 10.04] for H = 1.0 h, and 26.47% [18.03, 34.91] for H = 1.5 h. Results are simulation-based benchmarks on NASA laboratory battery discharge data; field validation remains future work. Empirical coverage validation on held-out grouped test batteries confirmed that the residual-quantile buffer achieved 93.03% pooled empirical coverage (mean 91.61%, 95% CI: [88.69%, 94.53%] across 10 repeated runs) against a nominal 80% target, supporting its conservative empirical robustness properties.