TCN-LSTM-based low-voltage transformer area topology identification method and system

Abstract

The invention belongs to the technical field of topology identification, and particularly relates to a TCN-LSTM-based low-voltage transformer area topology identification method and system. The method comprises the following steps: acquiring and preprocessing voltage data of a low-voltage transformer area; inputting the preprocessed voltage data into a power grid topology identification model, extracting periodic features of the voltage data by using a TCN model, and extracting long-time dependence features of the voltage data by using an attention mechanism-based LSTM model; and splicing the extracted features, inputting the spliced features into a KANs network for feature fusion, and obtaining a low-voltage transformer area topology identification result after the fused features pass through an output layer. Considering that voltage time sequence data generally have long-term trend and periodic characteristics, the method captures long-term dependence information and periodic characteristics by constructing parallel characteristic extraction branches, and provides rich characteristic representation for subsequent transformer area identification.

Images