Protocol for the review and meta-analysis of the accuracy of Advanced Methods (AI, Ensemble and Hybrid) versus Classical Methods (Pure Physical and Statistical) for forecasting wind speed and wind power at onshore and offshore sites influenced by coastal or trade wind dynamics or located in complex terrain

Do advanced forecasting architectures (artificial intelligence, ensemble and hybrid models) reduce error more than classical ones (physical and statistical models) in complex wind environments? This systematic review and meta-analysis synthesizes the error metrics reported by studies of wind speed and wind power forecasting at onshore and offshore sites influenced by coastal dynamics or trade winds, or located in complex terrain, published from 2020 to 2026 (databases searched on 22 September 2026). Reported error metrics are standardized as Hedges' g and pooled with random-effects models; meta-regression examines whether the architecture of the proposed model, the design of the comparison (within a family or between families), the prediction target and the metric family explain differences between studies. The review distinguishes the gain of a newly proposed model over its own baseline from the contrast between architectural families, and it reports prediction intervals, sensitivity analyses and small-study diagnostics. Its aim is to state what the published comparative evidence supports, not to rank models. This registration was updated in October 2026 to match the study as conducted; the reason for the update and the post hoc changes are documented in the registration. Data and code: https://gitlab.com/maralcar/wind-forecasting-meta-analysis.