Under China’s carbon peaking and carbon neutrality goals, building-integrated photovoltaics (BIPV) are a key option for low-carbon urban transition. However, how urban morphology shapes effective PV potential in severe cold cities remains poorly understood. Previous work focuses on single buildings or citywide resource mapping and rarely yields actionable planning controls. Using Harbin as a case, this study integrates GIS with explainable machine learning to relate building-cluster morphology to effective PV generation potential. An XGBoost model is interpreted with SHAP and partial dependence analysis to quantify factor importance and response ranges. Building density (BD) and floor area ratio (FAR) are the dominant predictors, ranking above the other morphological indicators. PV density peaks at moderate BD (≈0.20–0.35) under medium-to-high development intensity, and it increases when building distribution is moderately even (NNI ≈ 1.3–1.5) with moderate height differentiation. These coupled responses define a Morphological Sweet Spot, indicating that higher PV performance depends on coordinated morphological configurations rather than on any single parameter. The framework provides an interpretable, data-driven basis for building-cluster BIPV assessment and for translating model outputs into morphology-based planning guidance for low-carbon renewal in severe cold regions.
Yin et al. (Thu,) studied this question.