Power grid fault handling plan matching method based on a hybrid neural network

A matching method based on a hybrid neural network is proposed to improve the accuracy of online matching for a power grid fault handling plan. First, the ERNIE 3.0 encoding and double-pointer decoding module are used to replace the generative model in the universal information extraction (UIE) framework, and the mapping relationship between entities and entity labels of the fault handling plan is trained by adjusting the hyperparameters of the UIE framework. Then, the semantic distance between the fault equipment, fault type, fault phenomenon, and the entity of the fault handling plan is calculated based on the residual vector-embedding vector-encoded vector (RE2). The hybrid neural network model for power grid fault handling plan matching is established. Finally, by verifying the fault-related data of a regional power grid, the proposed fault handling plan matching method shows higher matching accuracy and stronger generalization ability than other algorithms. The average precision rate, recall rate, and F 1 value of the built fault handling plan matching model are 97.61%, 98.24%, and 97.91%, respectively, which can support auxiliary decisions for timely and rapid treatment of power grid faults.