Adaptive surrogate-based algorithm for integrated scheduling and dynamic optimization of sequential batch processes

We propose a novel solution algorithm for the integrated scheduling and dynamic optimization for sequential batch processes in this work. The integrated problem is formulated as a mixed-integer nonlinear programming (MINLP) problem, which could be large scale and challenging to solve. To address this computational challenge, we propose an efficient and adaptive surrogate-based algorithm for solving the integrated MINLP problem. Based on the bilevel structure of the integrated problem, we first decompose the dynamic optimization problems from the scheduling problem and replace them with a set of surrogate models. We then update the surrogate models adaptively, either by adding a new sampling point to the current surrogate model, or by doubling the upper bound of the current surrogate model's total processing time. Our proposed method is demonstrated through a case study involving a multi-product sequential batch process. The results show that the proposed algorithm leads to a 31% higher profit than the conventional method. The full space simultaneous method increases the computational time by more than four orders of magnitude compared with the proposed method but returns an 8.7% lower profit than the proposed method.