Power intelligent fusion terminal of distribution transformer

Abstract

The application discloses a power intelligent fusion terminal of distribution transformer, and relates to the technical field of power system intelligent operation and maintenance, comprising a prediction module, a standardized time series data sequence is input into a PINN prediction model, and the residual life probability distribution of the distribution transformer is output; an execution module, the terminal executes a voltage reactive power control action set, adjusts the output voltage and reactive power balance of the distribution transformer, and generates an action execution feature abstract; a federal learning module, using a global reinforcement learning strategy model and the feature abstract of all terminals, generating personalized model update parameter packages of different terminals and distributing them to corresponding terminals; by inputting the standardized time series data into the PINN prediction model and fusing the physical constraint loss term, the robustness and physical interpretability of the life evaluation under data scarcity and noise interference are significantly improved; the safety, economy and adaptive capacity of the distribution transformer operation are effectively improved.

Images