Decentralized optimal control of multicomponent reactor systems in petrochemical and oil-and-gas production based on situational decomposition method

Abstract

In petrochemical and oil-and-gas production, multicomponent reactor systems with continuous production process are widely used. Therefore, to ensure their effective and high-efficiency operation should be considered as an important and primary objective. Automated optimal control for such systems, based on innovative approaches and their realization means, is a key aspect of the named aim. A distinctive feature of the industrial controllable objects of the class under investigation is their characteristic complexity caused by the necessity to allow for a considerable number of parameters having various and numerous interconnections and a complex nature of the interaction. For this reason, the creation of automated control systems based on the traditional centralized structural principle and complete representation of the controllable object, is often highly costly in such objects due to the significant computational resources. The proposed approach incorporates a structurized representation of the reactor system with subsequent optimization problem structuring. In such a case the problem solution is assumed to lie in a decentralized control system incorporating a set of local control systems operating simultaneously and a coordinator guiding their functioning. The required optimum of reactor system as a whole is achieved as a result of informational exchange among the local control systems and the coordinator. During this exchange the local control systems solve optimization problems for individual components of the reactor system, considering the effects of the coordinator. Whereas the coordinator solves a global optimization problem of the reactor system as a whole due to proper correlation of local problems solutions. Positive effect is achieved by decentralization because the emerging optimization problems and global coordination problem are much simpler than the initial control problem in the reactor system. Therefore, their solution is more efficient in terms of required computational resources. To implement the decentralized approach, it is proposed to use the situational decomposition method. This method provides a wide range of possibilities to vary the control while imposing no special requirements on the initial optimization problem structure in contrast to the classical decomposition methods.