Multi-Dimensional Coupled Evaluation and Prediction Of Solar Energy Utilization Indicators on Building Surfaces

The evaluation of solar energy utilization potential of urban building surfaces currently faces the dilemma of high complexity of large-scale-high-precision-multidimensional coupled computation. This study introduces a more comprehensive method for urban building clusters splitting and type identification, and uses geometric morphology to extract multi-dimensional feature indicators of building clusters. A multi-dimensional sky module technology coupling temporal dimension and radiation type, a dynamic identification method of building surface orientation, a high-performance coupled computational framework and a multi-dimensional metrics parsing module have been developed. Further, a variety of machine learning algorithms were examined, and finally the XGB model, which balances predictive performance (R2>0.95 and MSE<0.10) and prevents overfitting, was selected to predict multidimensional utilization potential indicators for existing urban buildings in non-enriched areas. The study found that: (a) the geographic location of building clusters, geometric morphology and building types can better characterize the variability of urban building clusters and be used to build high-precision prediction models. (b) The shading of typical urban buildings varies across orientations, with roofs distributed between 3.45% and 6.98%, and façades between 34.70 and 50.71%. (c)The variability of solar radiation is more significant both in different directions and in different time dimensions, e.g., winter accounts for about 38% of summer in Chengdu and only 30% of summer in Chongqing. In this study, we further captured the nonlinear relationship between utilization thresholds and effective utilization potentials under different orientations and constructed high-precision dynamic prediction models with bi-directional gains for explaining the science and advancing the applications.