Thesis/6-broader-impacts/v1.tex

\section{Broader Impacts}

\textbf{Who cares and why now:} The nuclear industry, datacenter operators, and clean energy advocates all face the same economic constraint: high operating costs driven by staffing requirements. AI infrastructure demands—growing exponentially—have made this constraint urgent.

Nuclear power presents both a compelling application domain and an urgent
economic challenge. Recent interest in powering artificial intelligence
infrastructure has renewed focus on small modular reactors (SMRs), particularly
for hyperscale datacenters requiring hundreds of megawatts of continuous power.
Deploying SMRs at datacenter sites minimizes transmission losses and
eliminates emissions from hydrocarbon-based alternatives. Nuclear power
economics at this scale, however, demand careful attention to operating costs.

The U.S. Energy Information Administration's Annual Energy Outlook
2022 projects advanced nuclear power entering service in 2027 will cost
\$88.24 per megawatt-hour~\cite{eia_lcoe_2022}. Datacenter electricity demand is
projected to reach 1,050 terawatt-hours annually by
2030~\cite{eesi_datacenter_2024}. Nuclear power supplying this demand would
generate total annual costs exceeding \$92 billion. Within this
figure, operations and maintenance represents a substantial component. The EIA
estimates that fixed O\&M costs alone account for \$16.15 per megawatt-hour,
with additional variable O\&M costs embedded in fuel and operating
expenses~\cite{eia_lcoe_2022}. Combined, O\&M-related costs represent
approximately 23--30\% of the total levelized cost of electricity, translating
to \$21--28 billion annually for projected datacenter demand.

\textbf{What difference will it make?} This research directly addresses the
\$21--28 billion annual O\&M cost challenge through high-assurance autonomous
control, making small modular reactors economically viable for datacenter power.

Current nuclear operations require full control room staffing for each
reactor—whether large conventional units or small modular designs. For large reactors producing 1,000+ MW, staffing costs spread across substantial output. Small modular reactors producing 50-300 MW face the same staffing requirements with far lower output, making the per-megawatt cost prohibitive. These staffing requirements drive the economic challenge
that threatens SMR deployment for datacenter applications. Synthesizing provably correct hybrid controllers from formal
specifications automates routine operational sequences that currently require
constant human oversight. A fundamental shift from direct operator
control to supervisory monitoring becomes possible, where operators oversee multiple autonomous
reactors rather than manually controlling individual units.

The correct-by-construction methodology is critical for this transition.
Traditional automation approaches cannot provide sufficient safety guarantees
for nuclear applications, where regulatory requirements and public safety
concerns demand the highest levels of assurance. Formally verifying both the
discrete mode-switching logic and the continuous control behavior, this research
will produce controllers with mathematical proofs of correctness. These
guarantees enable automation to safely handle routine operations---startup
sequences, power level changes, and normal operational transitions---that
currently require human operators to follow written procedures. Operators will
remain in supervisory roles to handle off-normal conditions and provide
authorization for major operational changes, but the routine cognitive burden of
procedure execution shifts to provably correct automated systems that are much
cheaper to operate.

SMRs represent an ideal deployment target for this technology. Nuclear
Regulatory Commission certification requires extensive documentation of control
procedures, operational requirements, and safety analyses written in structured
natural language. As described in our approach, these regulatory documents can
be translated into temporal logic specifications using tools like FRET, then
synthesized into discrete switching logic using reactive synthesis tools, and
finally verified using reachability analysis and barrier certificates for the
continuous control modes. The infrastructure of requirements and specifications
already exists as part of the licensing process, creating a direct pathway from
existing regulatory documentation to formally verified autonomous controllers.

Beyond reducing operating costs for new reactors, this research will establish a
generalizable framework for autonomous control of safety-critical systems. The
methodology of translating operational procedures into formal specifications,
synthesizing discrete switching logic, and verifying continuous mode behavior
applies to any hybrid system with documented operational requirements. Potential
applications include chemical process control, aerospace systems, and autonomous
transportation, where similar economic and safety considerations favor increased
autonomy with provable correctness guarantees. Demonstrating this approach in
nuclear power---one of the most regulated and safety-critical domains---will
establish both the technical feasibility and regulatory pathway for broader
adoption across critical infrastructure.


These broader impacts answer the final Heilmeier questions: 

\textbf{Who cares?} The nuclear industry, datacenter operators, and anyone facing high operating costs from staffing-intensive safety-critical control. 

\textbf{Why now?} AI infrastructure demands have made nuclear economics urgent, and formal methods tools have matured to the point where compositional hybrid verification is achievable. 

\textbf{What difference will it make?} Enabling autonomous control with mathematical safety guarantees addresses a \$21--28 billion annual cost barrier while establishing a generalizable framework for safety-critical autonomous systems.

The next section presents the timeline for achieving these outcomes through a structured 24-month research plan.