A Data-Driven Surrogate Modeling Optimization Framework for the Economic Dispatch of Microturbines Using Hydrogen Blends

The buildings and industrial sectors require reliable energy supply, while simultaneously mitigating greenhouse gas (GHG) emissions. This poses interesting opportunities for the use of microturbine technology for on-site energy generation using natural gas blended with electrolytic hydrogen (H2) from renewable electricity. Previous work has identified a commercial gas micro-turbine that can adopt a range of H2-Natural gas (NG) blends. In this paper, we present an economic dispatch framework for a Capstone C65 microturbine, that can deliver up to 65 kW of power using various H2-NG fuel blends to optimize for system operating costs of meeting demand for electricity in a given application. To optimize the system, multi-fidelity data obtained from a combination of experimental data and computational fluid dynamics simulations were used to create surrogate models through Gaussian Process Regressions (GPRs). Then, these models were embedded into a nonlinear optimization problem whereby the least-cost turbine dispatch strategy for a 24-hour period that meets a small commercial building power demand was found by choosing the H2-NG blends and electricity generation for multiple concurrent units while also considering purchasing electricity from the grid. Furthermore, we discuss the effects of a price on CO2 emissions on the optimal fuel blend and on-site electric generation.