Ensuring the reliable and economical production and delivery of electric power is a critical national objective with broad economic and national security implications. Power system modeling plays a key role in enabling these objectives by allowing operators and regulators to accurately simulate how a power grid behaves under different conditions, to design and test control strategies for various grid components, and to predict the impact of changes to the power grid structure, among other tasks. Within this context, data assimilation (DA) and model calibration (MC) tasks, which involve the combination of observations and numerical models, are critical to ensure that predictions from power system models are accurate and useful. This project will have a direct impact on a wide range of fields where “computer models” are used, including atmospheric sciences, oceanography, ecology, astronomy and engineering, among many others. This project aims to develop machine learning tools for data assimilation and model calibration aimed to situations in which the behavior of the underlying system can be described through physical laws encoded in (systems of) ordinary differential equations (ODE), differential algebraic equations (DAE) or partial differential equations (PDE). The techniques are based on Bayesian non-parametric regression methods, where the structure of prior is derived from the system of differential equations describing the underlying system. The main expected outcome