Simulation-based dynamic optimization of discretely controlled continuous processes

Abstract Discretely controlled continuous processes (DCCPs) is a special type of hybrid dynamical systems which is of great practical relevance. In this work, a novel simulation-based approach to dynamic optimization under uncertainty of DCCPs is proposed using multi-modal Gaussian Process Dynamic Programming (mGPDP). A remarkable advantage of the proposed approach is that instead of resorting to a global metamodel, which is very inefficient, mGPDP uses probabilistic models (Gaussian Processes) to simultaneously learn the transition dynamics descriptive of mode execution and to represent the optimal control policy for mode switching. Throughput maximization and smoothness in a typical PVC production line in the face of significant schedule variability due to resource sharing is used as a case study.