The flow past a cylinder in proximity to a plane wall is investigated numerically for small gap ratios. Three vortex dynamic processes associated with different hairpin vortex generation mechanisms are identified for the first time, and the wake-induced turbulent transition is analysed. The vortex shedding is suppressed at $G/D = 0.1$ , while the spanwise vortex is generated via a Kelvin–Helmholtz instability and evolves into hairpin vortices. For $G/D= 0.3$ , the upper and lower rollers alternatively shedding from the cylinder, interact with the secondary vortex. The split secondary vortex merges with the upper roller and results in a new vortex downstream, which develops into hairpin vortices. When $G/D = 0.9$ , the secondary vortex interacts with the lower roller and then evolves into hairpin vortices. A tertiary vortex induced by the secondary vortex is observed, rotating in the opposite direction to the secondary vortex the wake-induced transitions share the same route. The velocity fluctuations deviate from the optimal growth theory in the pre-transitional region. In the transitional region low-frequency disturbances penetrate the sheltering edge to generate streaks where the disturbance energy declines. In the turbulent region the logarithmic layer is formed, indicating that the turbulent equilibrium is established.

The experimental investigation focuses on the effects of a short splitter plate on the flow physics of a circular cylinder in proximity to a wall by particle image velocimetry. The Reynolds number is Re = 3900, and the near-wall cylinder is immersed in turbulent boundary layer flow. Three gap ratios (i.e. $G/D$ = 0.25, 0.5 and 1) are considered, and the splitter plate length is $L/D=0$ , 0.25, 0.5, 0.75 and 1. For $G/D$ = 0.5 and 1, as $L/D$ increases from 0 to 1, the splitter plate facilitates the cylinder shear layers to elongate downstream, and the vortex formation length is increased, which leads to the increase of the range of the recirculation region. For $G/D$ = 0.25, the wall suppression on the wake vortex formation is enhanced, and the variations of the vortex formation length and the range of the recirculation region with $L/D$ are small. The Strouhal number St presents a decrease with increasing $L/D$ for the three gap ratios. The effects of $L/D$ on the vortex evolution are revealed. For $G/D$ = 0.5 and 1, as $L/D$ increases, the induction of the lower wake vortex on the wall secondary vortex becomes weaker due to the reduction in strength of the wake vortex and the increase of the vortex formation length. Additionally, the wake fluctuation intensity is decreased with the increase of $L/D$ due to the splitter plate suppression. For $G/D$ = 0.25, theL/D influences on evolution of the wake vortices and wall secondary vortex are small, which result in weaker variation of the wake fluctuation intensity with $L/D$ .

Accurate wind speed prediction is a premise that guarantees the reliable operation of the power grid. This study presents a combined prediction model that integrates data preprocessing, cascade optimization, and deep learning prediction to improve prediction performance. In data preprocessing, the wavelet soft threshold denoising (WSTD) is employed to filter the blurring noise of the original data. Then, the robust empirical mode decomposition (REMD) and adaptive variational mode decomposition (AVMD) are adopted to carry out a two-stage adaptive decomposition. Spearman correlation is used to quantify the mode that need to be decomposed for the second time. In the cascade optimization, the hybrid grey wolf algorithm (HGWO) is employed to optimize the parameters of the VMD and the gated recurrent unit (GRU), which overcomes the problem of empirical parameter adjustment. The HGWO is also adopted in the prediction strategy to optimize the GRU model to predict the grouped intrinsic mode functions (IMFs). Lastly, the final wind speed prediction result is obtained by superimposing the values of all the predicted models. The proposed model was validated with the measured wind speed data of the four quarters in the Bay area of China and was compared with 20 models of the classic method to further evaluate the effectiveness of the model. The results show that the whole process of the proposed model is adaptive, the final multi-step prediction performance is good, and high prediction accuracy can be attained.