Data-driven characteristics and representation of vertical wind profiles in a mountainous region

Against the backdrop of energy transition and sustainable development, wind energy, as a vital clean energy resource, has drawn significant attention. An accurate wind speed profile holds significant engineering importance for assessment of wind energy resources, wind turbine site selection and structural wind-resistant design for both bridges and wind turbines. Power-law function, widely used in plains, neglects meteorological factors and influence of complex topographic conditions, thus restricting applicability in mountainous terrain. In this study, field measurements reveal there are two distinct types of strong wind in mountainous gorge, namely periodic thermally-driven winds and sudden intense winds, each with unique wind characteristics and wind speed profile shapes. To improve the characterization of these profiles, we propose a data-driven method combining Proper Orthogonal Decomposition with a joint probability density model. The analysis reveals the probability density distributions of modal coefficients exhibit significant differences while the modal vectors of wind profiles are similar. By establishing data-driven model linking wind directions and modal coefficients, we can accurately describe wind profile features across all directions. Validation using field data demonstrates that, compared with power-law profile, the model effectively captures the morphological characteristics of complex mountainous wind profiles under various climates and wind directions. Relative to climatic influences, the mountainous terrain has a more significant impact on wind profiles.