Mixed-frequency external grey information model for forecasting new energy power generation under the global 2 °C trajectory

Against the backdrop of global climate action, energy transition and the construction of new power systems, accurate forecasting of wind and solar-dominated new energy power generation is crucial. However, the shortcomings of existing methods in modeling time-varying and energy-economy-environment system mixed-frequency data in new energy power generation have led to limitations in prediction accuracy. Therefore, this study proposes a reverse constrained mixed-frequency grey model. The model incorporates an improved lag polynomial of external grey information, optimized via particle swarm optimization. The model demonstrates high accuracy, robust performance, and superior explanatory power in both Monte Carlo simulations and case analysis. Furthermore, the model is applied to predict China’s new energy power generation before 2035 under the current trajectory and below 2 °C scenarios, quantitatively evaluating the consistency between China’s energy transition and global climate goals. The values in February 2035 will reach 7.8 times and 7.9 times the level of 2025, respectively. This provides methodological support for frequency mixing prediction in new power systems and lays a data foundation for evaluating energy transition paths and carbon neutrality processes.