Abstract: Decision-making of residential planning schemes has an important influence on building loads. A building load prediction model suitable for planning can quickly calculate differences of total building loads of different schemes, and effectively promote the optimization for energy-saving in residential quarters was set up. In the study, the samples were selected in Xi'an's built-up areas by an Universal Hashing method, analyzing the influence of the spacing relationship on the radiation value of building south external wall and the correlation between the load change rate and radiation reduction coefficient with the south building blocking. The results showed that the load change rate was linear correlation with the radiation reduction coefficient. On this basis, the quantitative relationship between the total building loads and the spatial form parameters of residential quarters was set up, and rapid prediction models of building loads for residential quarter planning were structured and perfected by conducting a multivariate regression analysis of spatial form parameters for residential quarters and total building loads of residential quarters, which would provide the technical reference to optimization of residential quarter planning schemes.