Nonlinear aerodynamic interference and spatial correlation of wind loads on ultra-large flexible photovoltaic arrays

The passage of Typhoon “Yagi” in 2024 caused severe damage to multiple aquaculture-complementary photovoltaic (PV) power plants, with over 90% of the PV panels in the rear rows of flexible PV arrays suffering wind-induced failure. This indicates that the current design standards significantly underestimate the amplification of pulsating wind loads due to strong aerodynamic interference between arrays, particularly in downstream regions. To address this issue, this study investigates the nonlinear aerodynamic interference mechanisms and spatial correlation characteristics of wind loads in ultra-large flexible PV arrays. Based on a key PV power plant demonstration project of the Three Gorges Corporation in Lianyungang, a nine-row, five-span flexible PV array model (span length: 33.3 m) is established. A wall-adaptive large-eddy simulation method is employed to accurately resolve near-wall flow around the PV panels. The three-dimensional velocity and vorticity fields under 0° and 180° wind directions are analyzed to reveal the flow evolution within the array. Subsequently, an overall wind load reduction factor model is established for both inter-span and inter-row areas. Finally, the maximal information coefficient (MIC) is used to quantify the nonlinear spatial correlation of wind loads, leading to a spatial distribution model of nonlinear correlation. Results show that the numerical model effectively captures the complex flow and pulsating wind load within flexible PV arrays. A low wind speed area with small-scale fragmented vortices forms in the sheltered section at the mid span of the array, while a high wind speed area with large-scale elongated vortices develops in the conical inter-span area. The reduction of wind loads along the row direction is particularly pronounced at the outer side spans and mid-span regions. However, the reduction factor in the inter-span area exceeds the standard values, reaching a maximum of 0.86. Wind loads on adjacent PV panels along the span direction exhibit moderate correlation, with a mean MIC of 0.48. In contrast, the correlation along the row direction is extremely weak. The spatially nonlinear correlation distribution of wind loads further confirms the non-uniformity of wind load reduction across the PV panel surfaces.