(4bl) Data-Driven Optimization Methods for the Design and Operation of Low-Carbon Energy and Chemical Production Systems

Fossil fuels, comprising of coal, oil and natural gas provide 85% of the total global energy supply. These fuels powered the industrial revolution of 1760-1840 and continue to heat our homes, run our vehicles, power manufacturing industry, and provide us with electricity. To limit the negative consequences of man-made climate change, it is critical to transition to an economy with low emissions. Numerous challenges still remain despite the past decades of intensive research toward sustainable energy. In the following, I elaborate on my specific research interests.

Low-carbon energy system design and operation

The power sector in many countries is rapidly evolving as low-carbon technologies to reduce greenhouse emissions. However, the large share of solar and wind power requires the implementation of energy storage systems and/or back-up fossil power generation in order to fill the time periods when no renewable energy is available. The crucial steps for economical, sustainable, and reliable power production are finding the optimal (a) combination of renewable resources given local weather conditions, (b) design, and (c) operating strategies of the plant. Towards this goal, my group will focus on developing and applying novel modeling strategies that consider variability in weather conditions and develop tailored solution methods for solving these models.

Modeling and optimization of strategies for decarbonizing chemical industry

Chemical production is responsible for a significant portion of greenhouse gas (GHG) emissions. Thus, decarbonizing the chemical industry would significantly impact the global CO₂ emissions. The sources of GHG emissions are the combustion of fossil fuels to meet the energy requirements and the production of feedstocks. To this end, I am interested in determining the optimal combination of feedstock (biomass, flue gas, etc.) and energy resources (PV, wind, concentrated solar power, etc.) for specific chemical production processes and study the the trade-offs between economic and environmental impacts. Furthermore, renewable energy resources such as solar and wind exhibit changing dynamics, nonlinearities, and uncertainties, whereas existing chemical processes require a steady-state operation. Therefore, I am also interested in applying efficient control strategies for safe, reliable, and profitable operation of chemical processes.

Hybrid data-driven optimization methods

Besides the aforementioned application areas, there are problems in physics, chemistry, automated machine learning, and operations research, where the function values are obtained as a result of computationally expensive simulations. Traditionally, these problems have been treated as black-box and solved using data-driven/derivative-free optimization methods. However, these approaches have limited success in solving high-dimensional problems, suffer from slow convergence, and cannot guarantee global optimality. To overcome these challenges, my group will investigate novel hybrid approaches that combine extractable model insights with data.

Research experience

My PhD research at Texas A&M University focused on the development of algorithms for a class of problems for which the analytical expressions and the derivatives of the objective and the constraints are unavailable. The algorithms relied on data generated by performing large-scale computationally expensive simulation, evaluation of legacy codes or experiments. Our research addressed several theoretical and algorithmic challenges for solving these problems. The performance of our methods was found to be competitive to existing approaches and enabled the optimization of highly complex nonlinear algebraic and partial differential equation models, which arise in process intensification and integration problems.

As part of my postdoctoral research at Princeton University and University of Wisconsin-Madison, I had the opportunity to collaborate with interdisciplinary research groups working in catalysis and material science to provide my expertise in process modeling and optimization, techno-economic and life-cycle analyses, and data-driven optimization. I have also worked on modeling and analyses of concentrated solar power (CSP) with thermochemical energy storage systems (TCES). Specifically, our analyses provided insights into optimal system design and material selection to improve the energy efficiency and costs of integrated CSP-TCES plants.

Teaching interests

The opportunities to teach and develop new courses have been a strong motivation for my pursuit of an academic career. To learn about the various teaching techniques, I completed Academy of Future Faculty program offered by Center for the Integration of Research, Teaching and Learning at Texas A&M University. I have also been a teaching assistant for two undergraduate-level courses (Fluid Mechanics and Mass Transfer Operations), which allowed me to deliver tutorials, grade assignments, and conduct office hours.

From my PhD and postdoc, I have developed expertise in optimization theory and algorithms, numerical methods, applied statistics, and process modeling and optimization. Therefore, I am best prepared to teach the following undergraduate-level courses:

• Chemical process control
• Process design and economics
• Introduction to chemical engineering
• Heat transfer operations
• Numerical methods for chemical engineers

In the era of big-data, companies are looking for candidates with strong data science and mathematical optimization skills. Therefore, I am also interested in developing two new advanced courses:

• Advanced process optimization
• Applied statistics and machine learning for chemical engineers

Selected Publications

[1] Bajaj I.; Iyer S.S; Hasan M. M. F. A trust region-based two phase algorithm for constrained black-box and grey-box optimization with infeasible initial point. Computers & Chemical Engineering. 2018, 116, 306-321.

[2] Bajaj I.; Hasan M. M. F. UNIPOPT: Univariate projection-based optimization without derivatives. Computers & Chemical Engineering. 2019, 127, 71-87.

[3] Bajaj I.; Hasan M. M. F. Deterministic global derivative-free optimization of black-box problems with bounded hessian. Optimization Letters. 2020, 14, 1011-1026.

[4] Bajaj I.; Hasan M. M. F. Global dynamic optimization using edge-concave underestimator. Journal of Global Optimization. 2020, 77(3):487-512.

[5] Chang H.*; Bajaj I.*; Motagamwala A. H.; Somasundaram A.; Huber G. W.; Maravelias C. T.; Dumesic J. A. Sustainable production of 5-hydroxymethyl furfural from glucose for process integration with high fructose corn syrup infrastructure. Green Chemistry. 2021, 23, 3277-3288.

[6] Peng X.*; Bajaj I.*; Yao M.; Maravelias C. T. Solid-gas thermochemical energy storage strategies for concentrating solar power: optimization and system analysis. In review.

[7] Chang H.; Bajaj I.; Huber G. W.; Maravelias C. T.; Dumesic J. A. Catalytic strategy for conversion of fructose to organic dyes, polymers, and liquid fuels. Green Chemistry. 2020, 22, 5285-5295.

[8] Arora A.; Iyer S. S.; Bajaj I.; Hasan M. M. F. Optimal methanol production via sorption enhanced reaction process. Industrial & Engineering Chemistry Research. 2018, 57(42), 14143-14161.

[9] Arora A.; Bajaj I.; Iyer S. S.; Hasan M. M. F. Optimal synthesis of periodic sorption enhanced reaction processes with application to hydrogen production. Computers & Chemical Engineering. 2018, 115, 89-111.

[10] Balasubramanian P.; Bajaj I.; Hasan M. M. F. Simulation and optimization of reforming reactors for carbon dioxide using both rigorous and reduced models. Journal of CO₂ Utilization. 2018, 23, 80-104.

[11] Iyer S. S.; Bajaj I.; Balasubramanian P.; Hasan M. M. F. Integrated carbon capture and conversion to produce syngas: novel process design, intensification and optimization. Industrial & Engineering Chemistry Research. 2017, 56(30), 8622-8648.