【Haichuan Intelligence】Dalian West Pacific Petrochemical Company’s “Data-Driven Real-Time Closed-Loop Optimization Method for Atmospheric and Vacuum Distillation” has been in operation successfully for 100 days

Thread Content

On November 5, the \"data-driven real-time closed-loop optimization method\" developed by Dalian West Pacific Petrochemical Company was in operation for 100 days. It has proven effective in ensuring the stable operation of the plant, optimizing the product mix, and reducing energy and material consumption, marking a formal entry into a new phase of intelligent production driven by data. As the key unit in a refinery, the 10 million tons per year atmospheric and vacuum distillation unit has long faced challenges such as variable crude oil properties, nonlinear relationships among parameters across various systems, and insufficient real-time performance of traditional models; as a result, full-process online optimization has become a common problem across the industry. To this end, the company has assembled a team of experts from various fields such as processes, equipment, automatic control, and information technology. It has established an intelligent optimization approach that covers the entire process of \"sensing–decision-making–execution,\" and created an AI-driven optimization system that spans the whole lifecycle of the facilities. This provides a new pathway for strengthening innovation-driven development and fostering the growth of new types of productive forces in line with local conditions. This project has achieved multiple breakthroughs in technological innovation. Taking into account the characteristics of process big data, specialized processing algorithms for data cleaning and repair, as well as time-series analysis, are developed. By combining techniques such as association rule mining and deep learning, pattern discovery models that incorporate the expertise of industry experts are created, thereby providing a data foundation for the precise optimization of processes. An innovative approach is adopted, utilizing artificial neural network (ANN) models based on the first principle to represent the underlying mechanisms of heat exchanger networks, which are then integrated into digital twin model clusters. Additionally, dynamic models of the equipment, a monitoring system comprising 678 soft instruments with minute-level accuracy, and modules for detecting equipment abnormalities are developed, thereby providing comprehensive technical support throughout the entire process. “The closed-loop system we have developed utilizes AI algorithms to fully exploit the value of production data, enabling precise adaptation to various operating conditions; as a result, efficiency is improved by over 80% compared to traditional manual methods. ”Zhang Jiahua, the person in charge of the scientific research projects at the Second Refining Department, explained. After the system was put into use, the average variance of fluctuations in key process parameters decreased by 40.32%; the yields of jet fuel and diesel increased by 0.21% and 0.86%, respectively. The prediction accuracy for key indicators such as the initial and final boiling points of naphtha, the flash point of straight-run gasoline, and the 95% cut point of straight-run diesel reached industry-leading levels.