Efficient evaluation of wind energy and carbon mitigation potential under land resource constraints via deep learning

To achieve China's carbon neutrality by 2060, wind energy development requires balanced consideration of land-use constraints and regional disparities. Yet, a precise evaluation of future wind energy potential in a nation marked by pronounced land heterogeneity remains a major challenge. This study develops a novel Delta-Residual Gated Recurrent Unit-Corrected Long Short-Term Memory (DGCLSTM)-based framework to enhance the accuracy of long-term wind energy forecasting for barren land. Using 370 urban regions across China as case studies, DGCLSTM demonstrates significant improvements in predictive accuracy over 11 mainstream models. After systematically assessing Wind Power Density (WPD), Wind-Based Carbon Reduction Efficiency Density (WED), and the spatial configuration of barren land resources over the period 2025–2060, the results revealed that despite decadal declines of 0.997 W/m2 in WPD and 0.165 kg CO2/m2 in WED, China's remaining undeveloped barren lands retain the capacity to generate an annual average of 11,107.92 TWh in electricity and reduce 6737.33 Mt of CO2 emissions, corresponding to an estimated economic benefit of 613.16 billion USD and a theoretical mitigation potential of 0.04 °C in warming per year. We further identify high-potential regions that simultaneously exhibit advantages in power generation, emission reduction, and economic return through a three-dimensional feature analysis. This work offers accurate long-term assessments of wind energy and carbon mitigation potential under China's typical lands, supporting regional planning for energy transitions.