Session: 01-03: AI for Energy Sustainability III

Paper Number: 131440

131440 - A Data-Driven Surrogate Modeling Optimization Framework for the Economic Dispatch of Microturbines Using Hydrogen Blends 

Abstract: 

The growth of the commercial building and industrial sectors faces the challenge of satisfying demand for electrical power and heat, while simultaneously driving the carbon emissions down to desired levels for a decarbonized economy.  In this context, there is microturbine technology that can be used in small-scale energy production applications, which can be adapted to use a blend of Hydrogen (H2) and Natural Gas (NG). This in principle has the benefit of being able to regulate the carbon emissions, while presenting the operational challenge of selecting an appropriate ratio of fuels, that achieves the desired energy generation. However, the operation of the turbine is significantly influenced by its costs which might vary significantly over time. Therefore, least-cost operation of the system under emission and demand considerations needs to be studied. In order to achieve least-cost operation it is key having access to an accurate representation of the turbine, that can be used to predict the outputs for a wide range of input conditions. Then, this can be embedded into an optimization algorithm to balance its operation in the face of exogenous information, e.g., Hydrogen cost, demand levels, power imports, etc., for an arbitrary time horizon. 

In this manuscript, we present the optimal dispatch problem for a C65 Capstone microturbine, that can deliver up to 65kW of power using H2-NG fuel blend. This problem has been built upon a data-driven surrogate model of the turbine [1], which has been created using a multi-fidelity Gaussian process (GP) regression that merges low-fidelity numerical simulation data with high-fidelity experimental data of the system. This model is then used by an optimization algorithm to assess the best possible input conditions to the turbine, namely kW set-point and H2 percentage, for a sequence multiple time periods. Moreover, to address instances with high electricity consumption, this problem is extended to consider multiple independent units with their own operational variables, that can aggregate their power generation alongside electricity imports from the grid.  Because of the nature of the multi-fidelity GP framework, the resulting multi-period multi-unit problem is nonlinear, which can be solved with state-of-the-art nonlinear optimization algorithms with a few considerations based on the availability of exact derivatives of the surrogate model. 

The optimization framework is then used in the context of a small commercial building load data from NREL’s ComStock, alongside historical electricity data from PJM Interconnection. These creates scenarios in which turbine electricity production is in competition with imports from the grid. We show the conditions in which significant H2 use is feasible, moreover we study the implementation of a penalty on the CO2 emissions associated with both the turbines and the grid imports, i.e., an explicit carbon tax, and its effects on operational variables. 

[1] Bhattacharya, C, Christopher, J, Thierry, D, Biruduganti, M, Supekar, S, & Dasgupta, D. "Data-Driven Surrogate Modeling of Microturbine Combustors Burning Hydrogen Blends." Proceedings of the ASME Turbo Expo 2023: Turbomachinery Technical Conference and Exposition. Volume 9: Microturbines, Turbochargers, and Small Turbomachines; Oil and Gas Applications. Boston, Massachusetts, USA. June 26–30, 2023. V009T18A011. ASME.

Presenting Author: david thierry argonne national laboratory

Presenting Author Biography: David is a Postdoctoral Appointee at the Argonne National Laboratory. Its background focused on nonlinear optimization applications, including optimization of PDAEs, sensitivity analysis, and model predictive control. At its current appointment he focussed on optimization of an array of Microturbines, and multi period investment selection for decarbonization of industry.

Authors:

david thierry argonne national laboratory
Supekar Sarang argonne national laboratory
Chandrachur Bhattacharya Argonne National LAboratory
Joshua Christopher Argonne National Laboratory
Munidhar Biruduganti Argonne National LAboratory
Debolina Dasgupta Argonne National Laboratory
Sibendu Som ARGONNE NATIONAL LABORATORY