Meta-Learning Based Voltage Control for Renewable Energy Integrated Active Distribution Network Against Topology Change

This letter presents a meta-learning based voltage control strategy for renewable energy integrated active distribution network. The multiple interference self-supervised method is first applied to extract features from unlabeled data. Then, an efficient channel attention convolutional neural network is adopted to select targeted information that is most related to topology change from the features and induce knowledge transfer to update the voltage control strategy. This allows the proposed method to learn a novel voltage control strategy when only limited data are available for a new topology. Comparison results based on a 69-bus distribution network demonstrate the advancement of the proposed strategy.

Automated summary

This research presents a meta-learning based voltage control strategy that extracts features from unlabeled data using multiple interference self-supervised method, and then utilizes a convolutional neural network to select targeted information related to topology change for knowledge transfer and updating of voltage control strategy.