Abstract

Advances in optimization theory, software, and modeling environments promote widespread application of optimization-based decision making. However, the gap between the optimization model and the real-world counterpart impedes optimality of the solution and leads to loss of performance in practice. To overcome this barrier, this paper proposes to reconcile the model and the real plant with iterative parameter estimation and real-time optimization based on a trust-region framework. Three algorithms are developed with different means of constructing/maintaining the local model, managing the trust-region radius, and assessing the trial point. Analysis shows that all these algorithms converge to the correct first-order KKT point despite structural model mismatch. Case studies of reduced model-based optimization demonstrate effectiveness and performance of the proposed methods.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call