The downscaling calculation of wind environment has a significant impact on the reliability of urban microclimate simulation. The setting of boundary conditions in numerical simulation affects the accuracy of wind environment simulation. However, urban morphology is complex, and commonly used downscaling simulation methods are time-consuming and computationally intensive. In response to the above issues, this study proposes an ANN wind environment downscaling prediction model based on prototype block models (PBMs). The PBMs created in this study has added urban form indicators that can reflect the characteristics of wind environment. Taking the Nanjing area as an example, identify the prototype blocks within the research area and extract the geometric model of the prototype blocks. Then, through CFD simulation, build a wind environment database of PBMs. Then train the ANN model and validate its wind speed prediction performance with 7 PBM instances. By comparing the predictive performance of ANN under 15 combinations of explanatory variables, it was determined that the combination of FAR + FAI had the best overall predictive performance in terms of MAPE (0.144), RSME (0.237), and R2 (0.953). The ANN model based on PBM can be used to predict downscale wind speeds in cities.