Realtime databases are the basis of modern information technology. In the refining/petrochemical industry, information systems are used to implement plant and corporate initiatives such as process improvement, predictive maintenance, APC/MPC and optimization. Other applications such as production planning, maintenance management, expert systems, and supply chain management systems can be integrated with information systems to bridge the gap between business and realtime production environments.
The advantages of the PI System from OSI Software Inc. is its accessible environment and robust set of tools. These enable incorporation of new ideas and integration with existing system for enhancements. In this talk we will present an idea of building a value-added application on an existing PI platform that makes model predictive controllers (MPC) easy to design and deploy in plants. In particular, an adaptive model identification module on the PI platform will be built.
It is well-accepted fact that MPC controllers can bring huge benefits to refineries and petrochemical plants. However, to commission and to maintain MPC controllers are difficult and expensive. To solve this problem, we will introduce a new identification technology. The first part of this technology is a method of a multivariable, automated and (possibly) closed-loop plant testing, which can save 70% test time and will cause fewer disturbances to unit operation. The second part is an accurate and automated model identification and validation algorithm.
Two case studies will be presented to demonstrate the capability of the technology. The first one is a MPC identification of a crude unit at BP. The second is a partial closed-loop identification of a deethanizer at Dow. The combination of this technology with an information system such as PI will result in an adaptive MPC identification and model based monitoring system. This product will make the existing MPC controllers adaptive, which will reduce the work of MPC commissioning and maintenance considerably.
Yucai Zhu graduated from Xi'an Jiaotong University in 1982 with a bachelor's degree in control. He obtained his Ir. (M.Sc.) degree and PhD degree in 1985 and 1990, both from Eindhoven University of Technology. He joined Datex B.V. in 1990 as a research engineer. During 1991 and 1993 he worked at IPCOS B.V. as a control engineer and he was also a cofounder of the company. He worked at Setpoint Inc. and later AspenTech from 1993 to 1996. He co-founded Tai-Ji Control in 1996. Since 1998, he works part time at Eindhoven University of Technology as a researcher. He has developed the so-called ASYM method of identification and Tai-Ji ID software. He has done consulting works for major companies in the field of MPC/APC, including BP, Dow Chemical, Statoil, ExxonMobil and TotalFinaElf.