<p>The development of large-scale renewable energy sources, such as wind energy, is regarded as an important and effective means to achieve carbon neutrality. Accurate evaluation and projection of wind energy resources are therefore drawing widespread attention due to the rapid growth of the economy and increasing environmental concerns and demands. Thus far, most studies have assessed historical wind energy over China and projected its future changes based on simple multi-model ensemble (MME) from the Coupled Model Intercomparison Project phase 6 (CMIP6) projection. In this study, we use the improved genetic algorithm (GA) to reduce the uncertainties in reproducing the past historical near-surface wind speed (SWS) of China and further projected the future changes in wind power density (WPD) under different shared socioeconomic pathways (SSP; SSP1-2.6, SSP2-4.5 and SSP5-8.5) by CMIP6 outputs. The results indicate that most of the global climate models (GCMs) from the CMIP6 suitably represent the spatial structure and variability of the observational seasonal and annual mean SWS in China and its temporal variability, although the models largely deviate from the observations. The GA-based MME (GMME) better represents the SWS than other MME methods and exhibits the smallest biases in SWS simulation. Furthermore, the WPD of China is projected to decrease sharply throughout the 21st century under all three SSPs based on the GA, with the multi-model medians of WPD in the near-term (2024–2050), mid-term (2051–2080) and long-term (2081–2100) of all three SSPs being 31.50, 31.84, and 31.09 w·m^− 2, 30.71, 30.70, and 30.34 w·m^− 2, and 30.49, 29.99, and 28.81 w·m^− 2, respectively. In addition, in the near-term, the WPD is projected to decrease in most parts of China under all three SSPs, and the WPD decreases in most regions of China in the mid-term and long-term future under the SSP1-2.6 scenario; however, these values are projected to increase across East China under the SSP2-4.5 and SSP5-8.5 scenarios.</p>