In this talk, we introduce an optimization framework with new solution strategies to enable joint operations of renewable-driven power transmission systems and electrified chemical process heating. In this optimization framework, electrified chemical plants are modeled as microgrids that accommodate diverse energy sources for process heating. We then integrate the microgrid operation model with multi-agent unit commitment to derive a large-scale mixed-integer linear programming (MILP) formulation to coordinate the operations of electrified chemical plants and the power system [2]. Nevertheless, directly implementing this MILP formulation faces several practical limitations and computational challenges. First, when directly solving this joint operation formulation, state-of-the-art solvers such as Gurobi often fail to converge in a reasonable amount of time. Furthermore, this formulation implicitly assumes no information sharing barrier between power systems and chemical plant stakeholders. However, in reality, most plant managers are hesitant about directly sharing their process data with an independent system operator (ISO) due to its sensitive and private nature. Such information sharing barrier would impede decarbonization of process heating in chemical industries.
To tackle these challenges, we develop effective decomposition schemes that solve the joint problem in decentralized fashion. First, we implement a two-stage solution methodology based on Benders’ decomposition followed by MILP refinement. This allows for improved computational efficiency and solution tractability. Second, to preserve data privacy, we explore an Alternating Direction Method of Multiplier (ADMM) approach for dualizing the power flow and process heating demand constraint. This leads to a joint formulation that consists of process heating subproblems for each chemical plant and can be integrated with the power system operations problem in a decentralized manner. In this iterative ADMM scheme, power system stakeholder only needs to share the power flow variables corresponding to the bus associated with the electrified chemical plant, whereas transmission ISO only needs to receive the objective function value obtained from the plant-level operational subproblem [3]. We also study the convergence and computational performance of ADMM-based decentralized algorithms and obtain theoretical guarantees.
To validate the effectiveness of these computational advances, we consider a case study of clean olefins production in Texas using electrified steam cracking. We survey all 26 ethylene plants in Texas in terms of their feedstocks, production capacities, and local wind and solar profiles. We consider heterogeneous scenarios of electrification level across different chemical plants. And for each plant, its associated bus in ERCOT is identified within the benchmark 2000-bus synthetic grid dataset ACTIVSg2000 [4]. By comparing the computational time and optimal solution obtained using only Benders’ decomposition with those using Bender’s decomposition as well as ADMM-based decentralized algorithms, we find out that, apart from preserving data privacy, the latter strategy leads to much faster convergence to the optimal solution. In addition to operating cost minimization, we also calculate the carbon emissions under different electrification levels to evaluate the trade-offs between economic and environmental considerations. This allows us to identify practical “sweet spots” that balance economic performance with sustainability goals. Overall, findings of this case study will provide valuable, actionable insights for chemical and power systems stakeholders seeking to design and operate future-generation clean energy systems in the evolving energy landscape.
[1] U.S. DOE, “Industrial decarbonization roadmap,” September 2022. [Online]. Available: https://www.energy.gov/eere/doe-industrial-decarbonization-roadmap
[2] S. Ghasemi Naraghi, T. Kareck, R. Reed, P. Ramanan, and Z. Jiang. Decarbonization of Steam Cracking for Clean Olefins Production: Microgrid Planning and Operation. In: Optimization of Sustainable Process Systems: Multiscale Models and Uncertainties. John Wiley & Sons, Inc. (2025)
[3] S. Boyd, N. Parikh, E. Chu, B. Peleato, and J. Eckstein, “Distributed optimization and statistical learning via the alternating direction method of multipliers,” Found. Trends Mach. Learn., vol. 3, no. 1, pp. 1–122, Jan. 2011.
[4] H. Li, J. H. Yeo, A. L. Bornsheuer and T. J. Overbye, “The Creation and Validation of Load Time Series for Synthetic Electric Power Systems,” in IEEE Transactions on Power Systems, vol. 36, no. 2, pp. 961-969, March 2021, doi: 10.1109/TPWRS.2020.3018936.