Answers are grounded in this paper and cite their sources.
Computational modeling reveals accurate rooftop solar energy prediction across urban buildings, highlighting improved tools for regional renewable energy planning.
With the swift advancement of distributed photovoltaic power generation roof available light capacity has become an important basis for urban renewable energy planning and grid dispatching. In response to the problems of insufficient fusion of multi-source meteorological information and inadequate extraction of spatio-temporal features in existing methods, a Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) hybrid prediction model that fuses multi-source meteorological data with roof geometry was introduced. The findings demonstrated that the CNN-BiLSTM model converged to 96.3% accuracy and 95.2% recall after 80 iterations, and the mean absolute error (MAE) and mean absolute percentage error were reduced to 0.58 kW and 5.3% respectively, both clearly outperforming the control model; Under six different weather conditions, the model MAE remained in the range of 0.43–0.73 kW, especially under conditions of intense radiation fluctuations such as cloudy and overcast days. In actual photovoltaic project validation, the model’s prediction error for the annual power generation of six real scenarios was controlled within ± 3.1%, superior to the general tool PVGIS (with an error of up to ± 12.4%). The results show that the CNN-BiLSTM model outperforms the comparison models in terms of prediction accuracy and convergence speed, and demonstrates good adaptability in the complex meteorological conditions and real scenarios of Taiyuan region. The research results can provide useful technical reference for the assessment of urban distributed photovoltaic resources, the prioritization of roof development, and regional energy planning.
No takes yet. Share an insight, caveat, or question.
Ren et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: