Natural gas valve modeling study based on long and short-term memory neural network

The paper proposed a modeling approach for a natural gas compressor valve using a long and short-term memory neural network. The main objective is to simulate the dynamic response characteristics of the valve and offer support for optimizing the anti-surge control strategy. Using the Damadics Actuator Benchmark Library generated samples to train the model. The results indicate that the model's performance gradually improves on the training set. Despite some overfitting phenomena occurring when the excitation signals change, the trained network demonstrates a good fit with the actual data. This study presents an effective modeling and identification method for compressor anti-surge control, which serves as a valuable reference for ensuring the safe and stable operation of the system.