Climate-Zone-Stratified Solar PV Investment Viability Across 20 Sites: An Integrated ML Forecasting Pipeline with SURFRAD Reanalysis Validation

Read full text Summarise with VIRAT
Authors: Atick Hassan Nabil
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Solar PV investment viability is strongly climate-dependent, yet practitioners lack a quantitative framework for screening investment decisions across climate zones that simultaneously accounts for forecast uncertainty, reanalysis data bias, and site-specific financial conditions. This study provides that framework through a 20-site integrated ML-to-investment pipeline spanning four Köppen climate zones: Hot Desert (BWh), High-Altitude Semi-Arid (BSk), Humid Subtropical (Cfa), and Oceanic (Cfb), using NASA POWER v8.2 GHI data (2020–2023). The principal findings are: (1) Investment viability is strongly zone-ordered — 40% of BWh and BSk sites exceed the 6% IRR hurdle under base-case assumptions, versus 0% for Cfa and Cfb (excluding tariff outliers); a Kruskal–Wallis test confirms significant inter-zone IRR differences (H = 8.14, p = 0.043); electricity tariff is the dominant financial lever, with San Francisco (IRR = 19.4%) and Dublin (IRR = 7.6%) achieving investment-grade returns driven by tariff structure rather than solar resource quality. (2) ML forecast accuracy is climate-dependent — Hot Desert sites achieve RMSE ≈ 0.77 kWh/m²/day (R² ≈ 0.82) while Humid Subtropical and Oceanic sites reach RMSE ≈ 1.25–1.27 kWh/m²/day; regularized and linear models (Lasso, Ridge, or Linear Regression) dominate at 19 of 20 sites, consistent with the linear GHI–clear-sky relationship at daily resolution; the Diebold–Mariano test identifies significant improvement over the clear-sky baseline at 9 of 20 sites after Bonferroni correction (α = 0.0025 for 20 simultaneous tests). (3) NOAA SURFRAD ground-station validation confirms systematic reanalysis overstatement: R² drops by 0.092 units at Denver (TBL, 30 km offset) and 0.100 units at Phoenix (DRA, 300 km offset, attributable primarily to geographic displacement); applying the Denver R² reduction to the 20-site pipeline's marginal Denver IRR of 6.04% implies a ground-corrected IRR below the 6% hurdle, establishing Denver as a site requiring on-site pyranometer validation before capital commitment. (4) Carbon-price integration can render Cfa sites (Tampa, Houston, Atlanta) investment-grade at $66–$128/tCO₂ — within the range of existing and near-term carbon pricing schemes — while Pacific Northwest Oceanic sites require prohibitive breakeven prices (~$1,260–$1,320/t) owing to their near-zero-carbon hydro grids. All performance claims are supported by Diebold–Mariano tests with Bonferroni correction and bootstrapped confidence intervals.

Fields of Study & Subjects

Solar Radiation and Photovoltaics Photovoltaic Systems and Sustainability Capital Investment and Risk Analysis

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
28 March 2026
Language
EN
Document type
Preprint