Load Transfer Strategy Recommendation Method for Distribution Networks Augmented by Improved Graph Retrieval

To address the issue of load shedding in the event of N-1, N-2, and other faults in the distribution network, this paper proposes a load transfer strategy recommendation method for distribution networks augmented by improved graph retrieval, based on an existing dispatch plan electronic rule library. First, the pre-established dispatch plans by experts are transformed into a knowledge graph through relation extraction and entity construction, thus forming an electronic rule library for dispatching. Then, by fine-tuning a large model using LangChain, the graph retrieval method is improved through text2cypher transformation. When a fault occurs in the distribution network, the dispatcher interacts with the large model to generate cypher queries, enabling the rapid recommendation of load transfer strategies for the distribution network.