Knowledge graph question and answer reasoning method and system based on electric power and fusion field

The invention provides a knowledge graph question and answer reasoning method and system based on the electric power and fusion field, and the method comprises the steps: recognizing a subject entity in a question and generating a dependency syntax tree of the question based on the proposed question, and obtaining a sub-graph and an answer candidate set based on a knowledge graph and the subject entity; the dependency syntax tree is converted into a hierarchical structure chart, dependency information is fused in question embedding q through a graph neural network model, and the hierarchical weight of word segmentation of the question is obtained through the hierarchical structure chart; based on question embedding q and hierarchical weight, acquiring a reasoning instruction vector of the reasoning step, scoring based on relation embedding of the reasoning instruction vector and the sub-graph, and after the reasoning step is repeated for multiple times, acquiring a candidate entity in the answer candidate set