Economic NMPC for a Reversible Solid Oxide Cell

Sakshi Naik, Y. Zhao, Douglas Allan, S. Lee, Lorenz T. BieglerIndustrial 2026 · DOI

TL;DR

This paper introduces an economic nonlinear model predictive control (E-NMPC) framework to optimize the operation of reversible solid oxide cells (rSOCs) for electricity grid integration. The E-NMPC approach demonstrably reduces hydrogen consumption in fuel cell mode and slightly improves hydrogen production in electrolysis mode compared to traditional set-point tracking methods, while also considering the integration of battery storage for enhanced flexibility.

What Problem Does This Paper Address?

Reversible solid oxide cells (rSOCs) are promising technologies that can switch between generating electricity (fuel cell mode) and producing hydrogen (electrolysis mode). This flexibility is crucial for integrating with the electricity grid, where real-time prices (locational marginal pricing - LMP) fluctuate. However, rSOCs are complex systems with tightly coupled thermal, electrochemical, and mechanical dynamics. Determining the optimal operating strategy to maximize economic benefit (e.g., minimizing hydrogen costs when electricity prices are high, and maximizing hydrogen production when prices are low) is a significant challenge. Traditional control methods, like set-point tracking, may not fully exploit the economic potential of these systems, especially during mode transitions.

Methods

The core of this work is the development and application of an Economic Nonlinear Model Predictive Control (E-NMPC) framework. Unlike standard NMPC that aims to track set-points, E-NMPC directly optimizes an economic objective function over a future time horizon.

Key methodological aspects include:

  • Detailed rSOC Flowsheet Model: A comprehensive dynamic model of an rSOC system, including the stack and balance-of-plant equipment (heat exchangers, blowers, etc.), was used. This model captures the complex interactions and transient behaviors crucial for effective control.
  • Economic Objective Function: The E-NMPC formulation directly minimizes hydrogen consumption in fuel cell mode and maximizes hydrogen production in electrolysis mode, while still meeting electricity production/consumption targets dictated by the LMP.
  • Variable Transformations: To improve numerical stability and handle potential singularities (especially during mode switching), specific variable transformations were employed, such as reformulating the recycle ratio definition.
  • Discretization Strategies: Careful consideration was given to discretization schemes for heat exchangers to ensure numerical stability and prevent non-physical oscillations. A spatially symmetric Lagrange-Legendre collocation scheme was found to be effective.
  • Battery Integration: The framework was extended to include a simplified battery model, allowing the system to draw from or supply to a battery to buffer electricity demand and further optimize economic performance.

Datasets

The study utilized hourly Locational Marginal Pricing (LMP) data from the NREL database. This data was used to simulate realistic electricity market conditions, specifically a section of LMP data from January in the CAISO (California Independent System Operator) region, chosen to induce frequent switching between fuel cell and electrolysis modes.

Key Insights and Ideas

  • Economic Optimization Outperforms Tracking: The E-NMPC framework consistently demonstrates superior economic performance compared to traditional set-point tracking NMPC, achieving better hydrogen utilization and production economics.
  • Mode Switching is a Critical Challenge: The complex dynamics during transitions between fuel cell and electrolysis modes pose significant numerical challenges for control. The paper highlights the importance of robust modeling and control strategies to handle these transitions smoothly and efficiently.
  • Battery Integration Enhances Flexibility: Incorporating a battery, even with a simplified model, allows the rSOC system to better manage electricity commitments and further improve economic outcomes by acting as an energy buffer.
  • Trade-offs Exist: While E-NMPC offers economic benefits, there's a trade-off between aggressive economic optimization (lower penalty on tracking terms) and system stability. Aggressive optimization can lead to surplus hydrogen being purged rather than utilized.

Key Results

  • Hydrogen Consumption Reduction: In fuel cell mode, E-NMPC reduced hydrogen consumption by approximately 4% compared to T-NMPC-Power, while maintaining the same electricity output.
  • Hydrogen Production Improvement: In electrolysis mode, E-NMPC yielded a marginal increase in hydrogen production compared to set-point tracking NMPC.
  • Energy Reallocation: E-NMPC was able to reallocate electricity from balance-of-plant equipment (feed and sweep heaters) to the rSOC stack itself, leading to more efficient operation. For instance, E-NMPC consumed significantly less electricity for the feed heater (35 kWh vs. 1240 kWh for T-NMPC-Power).
  • Battery Impact: With a 1 MW battery, hydrogen consumption was reduced by an additional 1.8%. A 10 MW battery showed a more substantial reduction in hydrogen consumption (around 26% compared to tracking NMPC), though this was partly due to draining the battery without explicit recharging provisions.

Limitations and Future Work

  • Simplified Battery Model: The battery model used is simplified, lacking detailed terminal constraints on recharging. Future work should involve more rigorous battery modeling to analyze hybrid systems more accurately.
  • Discrete Decisions: The paper acknowledges that the decision of when to switch the destination of the out stream (e.g., for hydrogen compression vs. purging) is a discrete decision that was not explicitly modeled within the scope of this work.
  • Computational Delay: The authors plan to explore advanced-step NMPC strategies to reduce computational delays, which are crucial for real-time control.
  • Cyclic Steady State: For systems with frequent switching, reformulating the optimization problem to target cyclic steady states rather than single steady states is suggested.
  • Model Mismatch and Disturbances: Future work will also focus on compensating for model mismatch and external disturbances.
  • Utility Plant Extension: The developed E-NMPC formulations can be extended to optimize operations in utility plants with multiple power generating assets.

Summary updated

Abstract

Reversible solid oxide fuel cells (rSOCs) offer the flexibility to operate in tandem with the electric grid by switching between fuel cell and electrolysis modes based on real-time electricity prices. However, their complex, tightly coupled dynamic behavior poses significant challenges in determining optimal operating strategies. In this work, we present an economic nonlinear model predictive control (E-NMPC) framework to optimize the operation of rSOCs. The proposed E-NMPC is applied to a detailed rSOC flowsheet model that includes a utility scale rSOC module as well as balance-of-plant equipment necessary for thermal management. Our results demonstrate that in fuel cell mode, the E-NMPC strategy reduces hydrogen consumption compared to conventional set-point tracking NMPC, while maintaining the same level of electricity output. Also, in electrolysis mode, the E-NMPC yields a marginal improvement in hydrogen production. In addition, we explore the integration of a battery with the rSOC system to enhance flexibility in meeting electricity production and consumption targets.
BibTeX
@inproceedings{naik2026economic,
  author = {Sakshi Naik and Y. Zhao and Douglas Allan and S. Lee and Lorenz T. Biegler},
  title = {{Economic NMPC for a Reversible Solid Oxide Cell}},
  booktitle = {Industrial},
  year = {2026},
  doi = {10.1021/acs.iecr.5c04217},
}
Figure