Methods, devices, equipment, storage media, and program products for fine-tuning large-scale power system models

Abstract

This application relates to a method, apparatus, device, storage medium, and program product for fine-tuning a large-scale power system model, relating to the field of power systems. It includes: acquiring multimodal time-series data of the power system, inputting it into a pre-trained multimodal large-scale model, and obtaining the predicted sequences of each signal output in the time dimension; scoring the importance of each signal at each time point based on the volatility index of the predicted sequences; filtering signal time-series correlation points whose importance scores exceed a preset threshold; for each signal time-series correlation point, calculating the deviation between its predicted value and the historical steady-state mean, fusing the deviation correlation features between different signals at the same time point, and constructing a state vector; inputting the state vector into a reinforcement learning policy network to generate fine-tuning actions; and updating the parameters of the corresponding model substructures in the multimodal large-scale model according to the fine-tuning actions. This method can improve fine-tuning efficiency and enhance the response efficiency and reliability of the multimodal large-scale model in complex power scenarios.

Images