Efficient Sensitivity-Based Cooperation Concept for Hierarchical Multilayer Process Automation of Steam-Powered Plants

Hierarchical optimization architectures are typically employed to manage industrial steam processes efficiently. The key challenge today is to find a near global optimum despite these subdivided automation structures. This paper proposes a novel cooperation concept between hierarchical layers. In the upper layer, optimal static setpoints are computed for economic operation schedules of the entire plant. In the lower layer, a model predictive controller realizes these schedules, regulates the dynamic plant parts, and treats occurring disturbances optimally in a stochastic manner. To adequately overcome limitations due to mismatches between the optimization layers (caused by model mismatches or disturbances), it is essential to establish an efficient cooperation concept. Therefore, the presented concept exchanges specific sensitivity information between the optimization layers, where it is skillfully exploited to the benefit of global automation objectives, resulting in optimal expected operating costs. The novel concept is demonstrated via simulation studies calibrated with industrial measurements of a chipboard manufacturer. A performance analysis shows that the proposed cooperation concept outperforms alternative approaches, leading to the best possible trade-off between additional fuel and steam demand violation costs.