Experimental investigation of non-Gaussian extreme value distribution for wind loads on photovoltaic arrays

This study investigates the extreme wind pressures acting on a five-row, three-span photovoltaic (PV) array based on wind tunnel experiments and generalized extreme value (GEV) modeling. The wind tunnel experimental results show that wind-induced pressures on the PV array exhibit pronounced non-Gaussian behavior, with spatially varying skewness and kurtosis governed by flow separation and wake interactions. Severe non-Gaussian characteristics are concentrated at windward stagnation regions under the 0° wind direction and at leeward separation zones under the 180° wind direction. To model the extremes, GEV distributions were fitted to block maxima of pressure-coefficient time histories, and non-Gaussian peak factors were derived from the 0.57-quantile of the fitted cumulative distribution function. The resulting GEV-based peak factors consistently exceed the value prescribed in the Chinese load code (g = 2.5), indicating that Gaussian peak-factor assumptions significantly underestimate extreme wind pressures in PV arrays. Furthermore, a normalized shielding coefficient was defined by referencing peak factors and force coefficients to the spanwise leading-edge row. The results demonstrate that, under head-on inflow, downstream rows experience varying degrees of attenuation, whereas oblique and reverse inflows weaken classical shielding and may induce wake-driven amplification of non-Gaussian extremes in interior rows. These findings provide crucial insights for wind-resistant design, extreme load prediction, and future code refinements for large-scale PV support systems.