Abstract for: Extracting Analytical Surrogates from System Dynamics Models via SINDy and Neural ODEs: An Application to Electric Freight

The transition to electric heavy-duty freight is central to addressing the interconnected crises of climate change, yet it is constrained by a strong interdependence between vehicle adoption and charging infrastructure deployment. System dynamics models are highly effective at simulating these socio-technical feedback loops and accumulations. However, to navigate the uncertainties of this transition and manage systemic instability, there is an opportunity to bridge heuristic simulation with formal mathematical optimization. This paper proposes a solution to that problem through a general hybrid system identification framework that extracts a continuous-time analytical surrogate from complex system dynamics simulations. The approach combines the sparse identification of nonlinear dynamics algorithm with neural ordinary differential equations to form a grey-box model. The first component identifies the underlying stocks and flows structure, while the neural residual captures bounded capacities and conditional logic. The framework is trained using multiple shooting over a 40-year horizon. The resulting model reproduces the training trajectories with a normalized root-mean-square error below 4%. Furthermore, it maintains reliable predictive accuracy and structural integrity when evaluated on unseen initial conditions across alternative socio-economic scenarios. By translating heuristic formulations into closed-form analytical surrogates, this approach enables the direct calculation of system sensitivities and the deployment of formal control methods. Researchers and policymakers can leverage these models to formally identify structural leverage points, optimize interventions, and proactively manage systemic instability in sustainable freight transitions. Rephrasing and errors check