A data-driven state identification method for intelligent control of the joint station export system

Area of Science:

• Petroleum Engineering
• Artificial Intelligence
• Control Systems

Background:

• Intelligent control of joint stations requires accurate system state identification and adaptive operational adjustments.
• Current methods for identifying abnormal conditions and optimizing operations often lack precision and adaptability.

Purpose of the Study:

• To develop an accurate pressure drop prediction model for joint station export systems.
• To establish an intelligent method for identifying abnormal working conditions using a dynamic threshold.
• To enhance the adaptive control capabilities of joint station export systems.

Main Methods:

• A hybrid optimization algorithm combining Particle Swarm Optimization (PSO) and Gray Wolf Optimizer (GWO) was used to optimize a Backpropagation Neural Network (BP) model, creating the PSO-GWO-BP model.
• A pressure drop prediction model was established using the PSO-GWO-BP approach.
• A state identification method based on a dynamic threshold was developed, utilizing the PSO-GWO-BP pressure drop prediction model.

Main Results:

• The PSO-GWO-BP model demonstrated superior prediction accuracy compared to traditional hydraulic calculation modified (THCM) models and other machine learning algorithms.
• The proposed dynamic threshold method successfully identified abnormal working conditions in the joint station.
• The effectiveness and accuracy of the developed method were verified using production and operation data.

Conclusions:

• The PSO-GWO-BP model offers significant advantages in pressure drop prediction accuracy for joint station export systems.
• The dynamic threshold-based state identification method enables intelligent recognition of system operation states.
• This approach enhances the ability to detect abnormal conditions and adaptively adjusts operational schemes, improving the overall intelligence of the system.