Nuclear power plant state report severity level screening method based on BERT-BiGRU model

The invention provides a nuclear power plant status report severity level screening method based on a BERT-BiGRU model, and the method comprises the following steps: obtaining status report work order data, and carrying out the data cleaning; the preprocessed data serve as a training corpus to train a BERT-BiGRU model, parameters are adjusted, and an optimal BERT-BiGRU prediction model is obtained through training; receiving status report work order data input by a user, and calling the data preprocessing part to perform data cleaning on the received data; inputting the text data processed by the data input part into a pre-trained BERT model to obtain vectorization representation of the text data; inputting a vector matrix obtained in the text vectorization process into a pre-trained BiGRU neural network to realize classification and discrimination of the severity level of the state report to obtain a level screening result; if the state report input by the user meets the requirement, returning a level screen