Resource optimization method for digital twin model of power grid equipment based on model reduction technology

This paper focuses on an optimization method for resource consumption of digital twin models of power grid equipment based on model order reduction technology. The modeling process of digital twin models for power grid equipment is elaborated in detail, including data collection, model construction, and model validation, with necessary mathematical formulas provided. An empirical study taking the electromagnetic-thermal coupling modeling and order reduction of oil-immersed transformers as an example is conducted, demonstrating that model order reduction technology is beneficial for optimizing model resource consumption. By reducing the model order from 15642 to 50 dimensions, the computational efficiency is improved by 18.5 times, memory usage is reduced by 98.1%, and the maximum temperature prediction error is controlled within 2.7 K (relative error <3%), quantitatively verifying the balance between order reduction (99.7% dimension reduction) and accuracy loss.