Evaluating the impact of renewable resource data imputation on technoeconomic design of coastal multivector nearly zero energy buildings

Gaps in high-resolution solar and wind records are usually filled before a building energy system is sized, yet the effect of the imputation method on the design is rarely tested. Here, eleven statistical, time-series and learning-based imputation techniques were applied to minutely irradiance and wind-speed series for a hydrogen-integrated coastal nearly zero energy building in Dalian, China, under four missingness levels (5–80%) and two gap patterns. A dual validation combined the root mean square error (pointwise accuracy) with the Kolmogorov–Smirnov statistic (distributional fidelity), and the reconstructed series were propagated into off-grid and grid-connected capacity optimization. The two metrics ranked the methods differently, and neither predicted design stability alone. At 80% missingness, annualized-cost deviations ranged from −36% to +95% off-grid, with battery oversizing near 400% for smoothing-dominated methods, while four methods stayed within ±5%. Grid connection capped the worst-case cost penalty at 38% but did not reduce typical deviations; instead, it concealed battery undersizing exceeding 20% for eight methods. A repeated-realization benchmark showed that gap placement, not optimizer noise, dominated variability and that cheaper design from low-error reconstructions violated reliability limits on the true weather. Imputation for design-critical studies should be validated jointly by statistical, capacity and reliability criteria.‬‬‬.