Hybrid Modeling Approach Integrating First-Principles Models with Subspace Identification

This paper addresses the problem of synergizing first-principles models with data-driven models. This is achieved by building a hybrid model where the subspace model identification algorithm is used to create a model for the residuals (mismatch in the outputs generated by the first-principles model and the plant output) rather than being used to create a dynamic model for the process outputs. A continuous stirred tank reactor (CSTR) setup is used to illustrate the proposed approach on a continuous system. To further evaluate its efficacy, the proposed methodology is applied on a batch poly(methyl methacrylate) (PMMA) polymerization reactor and the predictions are compared with that of first-principles modeling and the data-driven approach alone. The paper demonstrates the improved modeling capability of the hybrid model over either of its components.