Using High-fidelity Time-Domain Simulation Data to Construct Multi-fidelity State Derivative Function Surrogate Models for use in Control and Optimization (2308.07419v1)

Abstract: Models that balance accuracy against computational costs are advantageous when designing dynamic systems with optimization studies, as several hundred predictive function evaluations might be necessary to identify the optimal solution. The efficacy and use of derivative function surrogate models (DFSMs), or approximate models of the state derivative function, have been well-established in the literature. However, previous studies have assumed an a priori state dynamic model is available that can be directly evaluated to construct the DFSM. In this article, we propose an approach to extract the state derivative information from system simulations using piecewise polynomial approximations. Once the required information is available, we propose a multi-fidelity DFSM approach as a predictive model for the system's dynamic response. This multi-fidelity model consists of summation between a linear-fit lower-fidelity model and an additional nonlinear error corrective function that compensates for the error between the high-fidelity simulations and low-fidelity models. We validate the model by comparing the simulation results from the DFSM to the high-fidelity tools. The DFSM model is, on average, five times faster than the high-fidelity tools while capturing the key time domain and power spectral density~(PSD) trends. Then, an optimal control study using the DFSM is conducted with outcomes showing that the DFSM approach can be used for complex systems like floating offshore wind turbines~(FOWTs) and help identify control trends and trade-offs.