A Nested Design and Analysis of Computer Experiments Approach to Enable Design for Controllability of a System of Electric Vehicle Charging Stations

This paper addresses the optimization of renewable energy–equipped electric vehicle (EV) charging stations in the Dallas–Fort Worth area. We propose a nested design and analysis of computer experiments (DACE) framework that integrates data science and operations research to jointly optimize system design and dynamic operational control, including station locations and capacities. In this framework, Phase 1 performs system design optimization, whereas Phase 2 evaluates each candidate system design through a dynamic stochastic control model. Phase 2 is formulated as an infinite-horizon approximate dynamic programming (ADP) problem that quantifies the controllability and economic performance of a given design, and the resulting expected dynamic operational values are embedded into Phase 1 DACE optimization to efficiently explore the design space. Because embedding dynamic control within system design is computationally challenging, we apply DACE in both phases: A DACE-based ADP model approximates the value function in Phase 2, and a second DACE surrogate model captures the design–performance relationship in Phase 1. This nested structure enables scalable and efficient optimization of complex EV charging systems under uncertainty. Benchmarking against a Kriging-based surrogate within the same nested DACE structure further demonstrates that the proposed multivariate adaptive regression splines–based surrogate provides superior design performance and improved stochastic profitability. Overall, our results show enhanced controllability and profitability compared with deterministic approaches, whereas the interpretable surrogate functions provide actionable insights into how design choices impact operational and financial performance.

History: Kwok-Leung Tsui served as the senior editor for this article.

Funding: This study was partially supported by National Natural Science Foundation of China [Grants 72101066, 72121001, 72571076, and 72131005], National Science Foundation [Grants ECCS-1128871, ECCS-1128826, and ECCS-1938895].

Supplemental Material: Data supporting this research, including supplementary materials, are available in Zenodo at https://zenodo.org/records/19210803.