Abstract In [Sass et al., Eur. J. Oper. Res., 316 (1): 36 – 45, 2024], we proposed a branch-and-bound (B&B) algorithm with growing datasets for the deterministic global optimization of parameter estimation problems based on large datasets. Therein, we start the B&B algorithm with a reduced dataset and augment it until reaching the full dataset upon convergence. However, convergence may be slowed down by a gap between the lower bounds of the reduced and the original problem, in particular for noisy measurement data. Thus, we propose the use of out-of-sample estimation for improving the lower bounds calculated with reduced datasets. Based on this, we extend the deterministic approach and propose two heuristic approaches. The computational performance of all approaches is compared with the standard B&B algorithm as a benchmark based on real-world estimation problems from process systems engineering, biochemistry, and machine learning covering datasets with and without measurement noise. Our results indicate that the heuristic approaches can improve the final lower bounds on the optimal objective value without cutting off the global solution. Aside from this, we prove that resampling can decrease the variance of the lower bounds calculated based on random initial datasets. In our case study, resampling hardly affects the performance of the approaches which indicates that the B&B algorithm with growing datasets does not suffer from large variances.

Abstract Hammerstein–Wiener models constitute a significant class of block-structured dynamic models, as they approximate process nonlinearities on the basis of input–output data without requiring identification of a full nonlinear process model. Optimization problems with Hammerstein–Wiener models embedded are nonconvex, and thus local optimization methods may obtain suboptimal solutions. In this work, we develop a deterministic global optimization strategy that exploits the specific structure of Hammerstein–Wiener models to extend existing theory on global optimization of systems with linear dynamics. At first, we discuss alternative formulations of the dynamic optimization problem with Hammerstein–Wiener models embedded, demonstrating that careful selection of the optimization variables of the problem can offer significant numerical advantages to the solution approach. Then, we develop convex relaxations for the proposed optimization problem and discuss implementation aspects to obtain the global solution focusing on a control parametrization technique. Finally, we apply our optimization strategy to case studies comprising both offline and online dynamic optimization problems. The results confirm an improved computational performance of the proposed solution approach over alternative options not exploiting the linear dynamics for all considered examples. They also underline the tractability of deterministic global dynamic optimization when using few control intervals in online applications like nonlinear model predictive control.