Prediction Model of Corrosion Rate for Oil and Gas Pipelines Based on Knowledge Graph and Neural Network

Aiming to solve the problems of high cost and large error, a prediction model of oil and gas pipeline corrosion rate was established by using a knowledge graph and neural network. The knowledge map of oil and gas pipeline corrosion was constructed by the seven-step method and the fusion of Bidirectional Encoder Representations from Transformers, Bidirectional Long Short-Term Memory, and Conditional Random Fields models (BERT-Bi LSTM-CRF) combined extraction, and the triplet information was stored by the Neo4j graph database. With eight factors, such as carbon dioxide content, as input signals, three kinds of neural networks were used to screen the basic prediction model, and the genetic algorithm (GA) and particle swarm optimization algorithm (PSO) were used to optimize and compare. The results show that the extraction accuracy of BERT-Bi LSTM-CRF is 94.46%, the recall rate is 90.39%, and the F1 value is 92.38%. The optimized PSO-BP model has the best effect, and the coefficient of determination in the training stage and the test stage is as high as 0.9954 and 0.9935, and the total mean square error of the sample is reduced by 92.8% compared with that before optimization. The corrosion rate model proposed in this paper ensures the safe operation of oil and gas pipelines and offers a theoretical foundation for corrosion prediction.