Implementation of multi agents based system for process supervision in large-scale chemical plants

• Modern-day chemical plants exhibit scale-, structure- and dynamic-complexity.
• Process supervision systems for such processes should be flexible, cooperative, collaborative, extensible, and scalable.
• In this paper, we propose an agent-based architecture for developing and deploying such process supervision systems.
• The key insight is that the process descriptors used for developing the supervision models are also dynamic.
• The implementation of the agent-based system is described and its benefits demonstrated using an oil & gas case study.

Modern chemical plants have evolved into extremely large and complex operations. Operators rely on the plant automation system, particularly the DCS, for managing the plant operations which themselves have become open, and involve multiple third-party technologies, instruments, and software. The structural-, scale- and dynamic-complexity makes it challenging for operators to infer the conditions in the plant quickly and make timely decisions, especially during abnormal situations. A process supervision system that assists the operators by providing holistic decision support is therefore essential. Here, we propose an multi agents based architecture for supervision of large-scale chemical plants. The key insight in the proposed architecture is that the process descriptors used for developing the supervision models themselves are not static and change routinely. The proposed architecture uses an ontology to represent all the process descriptors formally, so that any changes can be captured and their effects propagated seamlessly. This architecture has been implemented as a multi agent system called ENCORE. The detailed implementation of ENCORE is presented and its benefits are illustrated through an offshore oil and gas production case study.