The automatic generation method and wind speed prediction technology of high-rise residential layout were studied, and the method of automatic generation, wind environment simulation and comparative optimization for high-rise residential area was developed. The fitness function and optimization goal of genetic algorithm were proposed, and the automatic generation algorithm of building layout was established, according to the data extraction of residential area and building regulations in Yangtze River Delta region of China. A computational fluid dynamics (CFD) proxy model based on fully convolutional neural networks (FCN) was constructed to obtain the wind speed distribution characteristics of optimal layout, based on the image learning characteristics of FCN. The optimization effect and wind environment prediction effect were analyzed, and results showed that the the error of the proposed method was within the effective range. In addition, compared with Phoenics, the proposed method can significantly reduce the wind environment simulation time, and effectively avoid the limitations of architects’ personal experience. This method can automatically learn scheme arrangement, quickly get the optimal solution of high-rise residential layout under specific plot ratio and plot conditions, and provide guidance for the performance of living environment in today’s fast-paced architectural design.