Research

Which low-carbon technology mixes deliver reliable, affordable energy, and who pays for them

Research of Honglin Li: optimization and techno-economic analysis of hydrogen, nuclear and firm low-carbon power for large new loads such as AI data centers.

I build optimization and techno-economic models for decisions about low-carbon energy systems: what to build, where, how to run it, and under which market and policy conditions it pays off.

My doctoral research asked one question at three scales: when does flexibility, from hydrogen or from demand, add value to a low-carbon energy system? The answer at the national scale raised the question behind my current work: what can carry the capital cost of new firm, low-carbon supply, now that AI data centers need hundreds of megawatts around the clock?

Each project page shows the problem, the hard parts, the design choices and the findings with their limits.

Research path

  1. 2019–2021 · M.S. Foundations Operations research for supply chains with on-site microgrids Supply chains and microgrids
  2. 2021–2024 · Ph.D. Plant Hydrogen with wind and solar plants Wind–hydrogen Solar–hydrogen Offshore wind–hydrogen
  3. 2022–2025 · Ph.D. State Hydrogen across the Texas grid under uncertain renewable build-out Texas hydrogen
  4. 2025–2026 · Ph.D. and INL National fleet Hydrogen and grid services for U.S. nuclear plants Nuclear–hydrogen
  5. 2025– · UT Dallas and Kansas State Data-center campus and nation Firm low-carbon supply, including small modular reactors, for AI data centers SMR cogeneration SMR perspective

Supporting work Carbon-aware demand Wind power uncertainty Cost of delayed grid connection

Chapter 1 · Ph.D., The University of Texas at Dallas, and Idaho National Laboratory

Flexibility from plant to nation

One question at three scales: when does flexibility, from hydrogen or from demand, add value to a low-carbon energy system? Each step up in scale changed who makes the decision, which uncertainty matters most, and where flexibility gets its value.

  1. Plant

    Hydrogen makes wind output firmer and more valuable

    I extended an open optimization model with electrolysis, hydrogen storage and fuel cells, and co-optimized equipment sizes with hourly operation for a wind farm, then for solar and offshore wind plants. Two of the studies showed that the cost per kilogram of hydrogen and the cost of the whole system can move in opposite directions. These models took the site as given, so they could not say where to build across a region.

    Wind–hydrogenSolar–hydrogenOffshore O&M

  2. State

    Coordinating hydrogen supply and demand under uncertainty

    For Texas, I built the inputs from public market disclosures and a two-stage stochastic model that sizes hydrogen assets in eight weather zones and schedules them hour by hour across scenarios of demand and renewable build-out. The model showed how the price of gas-fired backup power changes where green hydrogen goes.

    Texas hydrogen

  3. National fleet

    Making firm low-carbon supply flexible

    During my internship at Idaho National Laboratory, I co-developed an hourly model that splits a reactor's output among grid sales, high-temperature electrolysis, storage and reserves, linked to a lifetime cash-flow analysis of 42 reactors in seven U.S. markets. Operated as baseload-only plants at 2024 prices, nearly all had negative net present value; with hydrogen integration, including the assumed hydrogen production credit, a majority reached positive net present value.

    Nuclear–hydrogen

The question that led to data centers

The fleet study showed that hydrogen, with the assumed production credit, and grid services can change the economics of existing reactors, while new reactors depend on capital cost and policy support. What could carry the capital cost of new firm, low-carbon supply? Large AI data centers, which need hundreds of megawatts around the clock, often where the grid cannot connect them for years, made that question concrete.

Chapter 2 · Research Associate, The University of Texas at Dallas; Postdoctoral Fellow, Kansas State University

Firm low-carbon power for AI data centers

Small modular reactors can supply firm power at the scale of an AI campus. I study when such supply can compete with the alternatives, at the scale of one campus and of the United States.

  1. Framing

    What a credible assessment must capture

    In a co-first-authored Perspective, I wrote the techno-economic analysis: reliability requirements, wide first-of-a-kind cost ranges, service-based metrics and co-products such as heat all change the comparison.

    SMR perspective

  2. One campus

    Where SMR cogeneration becomes competitive

    For a hyperscale campus in Texas, I co-optimized reactor output, turbine steam extraction for absorption cooling and grid exchange, hour by hour. Capital cost and policy set the competitiveness boundary far more than operating choices do, and an output-based credit discourages using steam for cooling. Accepted in Energy Conversion and Management.

    SMR cogeneration

  3. United States

    Data-center demand and SMR commercialization

    A national study asks whether data-center demand can help small modular reactors move from their first units to wider commercial deployment. Manuscript in preparation.

    Upcoming study

Proposal · Co-PI · under review (2026)

Secure power for Kansas data centers

Kansas State University Secure Energy Initiative, seed funding

Kansas is seeing data-center proposals of several hundred megawatts. The proposal asks how small modular reactors, their recoverable heat, data-center cooling and grid connections can be planned together, so that computing and cooling continue through disruptions at a cost the region can bear, with community priorities in siting and design.

  1. Community and policy preferences and local conditions
  2. Physical models of reactor heat, cooling and data-center operation
  3. Integrated planning of locations, capacities and timing under uncertainty
  4. Kansas demonstration at county level, with stakeholder review

My role: coupled models, optimization, economics and software, in an interdisciplinary team spanning electrical, nuclear and civil engineering, water research and political science.

Supporting capabilities

Side studies that built methods and skills used across the main line.

  • Operations research foundations

    Facility location, production planning and microgrid sizing for supply chains that trade energy with the grid (M.S.).

    Supply chains and microgrids

  • Carbon signals and demand

    Re-timing electricity use with hourly emission intensity across U.S. regions, and what that accounting can and cannot show.

    Emission-informed load regulation

  • Uncertainty quantification

    Multi-fidelity Gaussian processes that combine the turbine's theoretical power curve with scarce measurements for probabilistic wind power.

    Wind power curves Plant-level wind uncertainty

  • Interconnection economics

    What waiting for grid energization costs a data-center developer, other ratepayers and society, and how staging recovers part of it.

    Cost of delayed energization

Methods and toolkit

Methods I use across these studies:

Optimization

Linear, mixed-integer and stochastic optimization

Pyomo linopy AMPL Gurobi HiGHS

Techno-economic analysis

Discounted cash flow, levelized costs, tax-credit modeling

Carbon accounting

Life-cycle and carbon accounting

Simulation

Discrete-event and Monte Carlo simulation

Surrogate models

Gaussian-process and machine-learning surrogates

Open model frameworksDocumented, runnable frameworks that accompany my research.Software

Looking ahead

I want to keep working on how low-carbon energy systems should be planned as electricity demand grows: how large new loads can be supplied reliably and affordably, how to invest when demand and technology costs are uncertain, and who bears the costs. The models behind each study are designed to be reused, checked and extended.