Research

The Regional Energy Deployment System (ReEDS) is an open-source power sector capacity expansion model developed at the National Renewable Energy Lab (NREL) for modeling the future scenarios for the U.S. power grid. Currently, some instances of ReEDS, a large-scale linear program, can take hours to solve. We documented the mathematical formulation of the model and implemented a progressive hedging algorithm based on a spatial decomposition of the U.S. into 18 transmission groups.
Alex Derenchuk – Mines
Dr. Alex Zolan – NREL

Market participants in wholesale power markets are exposed to volatile real-time prices for power. Large excursions (both positive and negative) from average prices, caused by transmission congestion and reserve shortages, are common. Using data from Oklahoma Gas and Electric (OGE) and the Southwest Power Pool (SPP), we identify the causes, locations, and timing of price spikes in the SPP grid region and analyze their potential impact on member utilities such as OGE. We then investigate whether strategic battery storage deployment at specific grid locations in critical periods could dampen or eliminate these spikes, and utilize decision theory to identify optimal policies for battery storage deployment and operation. This work is supported by a plus-up grant from ARPA-E.
Alex Derenchuk – Mines
Prof. Daniel Bienstock – Columbia University
Prof. Yury Dvorkin, Tomás Tapia – Johns Hopkins University

Figure 1: Arvada Tower Location
The tower for suggested use is pictured above. Heliostats would be located on the beam, which
is north of the tower.

Figure 2: Key layouts for each heliostat-size-design-method pairing.

Figure 3: Receiver profiles for 200 kWth field (left) and max receiver heat field (right) at receiver
optical height of 33m at design point DNI of 633 W/m2
and equinox sun position. Legend values
are maximum, median, and minimum flux intensities.

The scale-up of concentrating solar technologies requires adequate on-sun test facilities to support on-sun demonstration of components that approximate plant-scale performance conditions and enable developers and researchers to economically evaluate the potential of new designs and concepts. The limited number of test facilities for central tower receiver technologies at or above the 100-kWth scale presents an impediment to the development of receiver concepts for concentrating solar power generation and solar reactors. Adaptation of available tower structures for solar receiver testing provides a low-cost pathway to expand demonstration capabilities of new receiver concepts, to explore solar field control technologies, and, in general, to accelerate development cycles of concentrating solar technologies. This study examines the possible adaptation of an existing 60-m high tower facility in the Front Range region of Colorado as a 200-kWth scale receiver and heliostat field test facility. The steel tower studied in this project stands at the Elevate Quantum site in Arvada, Colorado, recently acquired by the Colorado School of Mines. Although currently unused, the tower remains structurally robust to support multi-ton receiver assemblies at various heights along the tower where there are structural supports for receiver and balance of plant equipment. The tower is adjacent to a plateau for siting a heliostat field with adequate land area for reflecting close to 1 MWth of solar radiation toward a north-facing receiver location on the tower. The tower and surrounding site offer an exciting opportunity to establish a concentrating solar test facility for emerging heliostat and receiver technologies. This study uses NREL’s SolarPILOT to examine the viability of the tower, surrounding environs, and weather to offer up to 500 kWth of thermal power over enough days per year to support effective research and development. SolarPILOT generates heliostat layouts to achieve desired solar fluxes on a 1m x 1m receiver for varying degrees of flux uniformity for different tower placements. SolarPILOT simulation analyzes the cost-to-power ratio and other performance metrics for each receiver height and heliostat field layout. Two sizing approaches provide a basis for calculating the heliostat field to achieve a desired mean receiver flux and uniformity. The smallest layout method generates a field of heliostats with dense packing and an aiming strategy with a focused thermal flux profile. The spaced layout method produces a field layout with more heliostats and a dispersed aiming strategy. The smallest layout field has greater optical efficiency than the field from the spaced layout method, but with much lower flux uniformity which may be less ideal for performing receiver demonstration tests. The field for the spaced layout includes more heliostats and provides more uniform flux to the receiver, but at the expense of increased spillage around the receiver due to the aiming strategy. For both sizing methods, we compare the performance of 3m x 3m heliostats and smaller 1m x 1m heliostats. Available receiver locations on the top half of the tower offer a suitable range of solar flux capture to test 1m x 1m with mean fluxes of 250 kW/m2 from which final location can be selected based on demonstration test objectives. Future work includes consideration of additional heliostat and receiver designs, and the incorporation of additional testing costs and accessibility requirements.

Andrew Aikman, Sebastian Dogue, Dr. Gregory Jackson, and Dr. Alexandra Newman

Figure 1: Analysis Locations.
Each location considered in this analysis is shown above, with its corresponding marginal operating emissions rate region. PSCO – Public Service Company of Colorado, CAISO_NORTH – California Independent System Operator: Northern California, CAISO_SANBERNARDINO – California Independent System Operator: San Bernardino, CAISO_SANDIEGO – California Independent System Operator: San Diego, ERCOT_SOUTHTX – Energy Reliability Council of Texas: Southern Texas, NYISO_LONG – New York Independent System Operator: Long Island, FPL – Florida Light & Power, SPP_WESTNE: Southwest Power Pool: Western Nebraska

Figure 2: Results
Distributions of battery quantity, relative cost savings, and relative emissions savings across all Lake County homes are displayed above. Each column reflects the local utility electricity pricing structures of each location.

Lithium-ion battery energy storage systems can provide cost and CO2 emissions savings to residential stakeholders by shifting energy usage. Lifetime cost savings can be achieved without co-located generation due to decreases in capital costs, storage-based incentive programs, and large differences in peak and off-peak pricing. We consider five control strategies to address both electricity cost and emissions goals, and create two sizing methodologies for stand-alone residential battery energy storage systems; we rely on a derivative of REopt, a mixed-integer linear program consisting of decision variables representing system energy capacity and power, state-of-charge and hourly discharge, and an objective including capital, operating, and emissions costs. The Life-Cycle-Cost approach uses incentive programs to balance upfront capital expenditures with the present worth of lifetime electricity cost savings, and only induces the purchase of batteries under net profitable circumstances. Our Iterative-Heuristic approach adds battery modules until the marginal benefit diminishes below a specified tolerance. We determine optimal system designs for each home in a modeled electrified and retrofitted lowincome community in eight locations, with local electric demand profiles, electricity pricing and grid emissions rates. Results demonstrate that systems consisting of two to five batteries can provide mean energy cost savings of 20.5% for cost-minimizing strategies and mean emissions savings of 28.8% for emissions-minimizing strategies across all locations. Although opportunities are limited, there are locations in California where stand-alone residential battery energy storage can save up to $8,000 (29% of baseline electricity consumption costs) over a ten-year lifetime.

Andrew Aikman, Karlyle Munz, Dr. Paulo Cesar Tabares-Velasco, and Dr. Alexandra Newman

We analyze an integrated freight logistics system for mining operations in the United States. Our study focuses on optimizing the fleet size and cost over multiple years at daily fidelity, within a constrained multi-modal transportation network. We formulate and solve a system of multi-commodity network flow model that minimizes operational costs while satisfying material needs for mining and processing operations. The model captures dynamic supply and demand, transportation capacity constraints, storage limits, and detailed transload information. We provide quantitative trade-offs between transport modes and offer recommendations on transfer locations, storage policies, and routing protocols for the time horizon of interest.

Water heating is the second-largest contributor to energy consumption in residential buildings. However, while conventional electric water heaters rely on resistive heating, Heat Pump Water Heaters (HPWHs) leverage ambient air to heat water, achieving up to four times greater energy efficiency. Unfortunately, their larger size and slower recovery rates limit adoption in space-constrained residences such as multi-family and manufactured housing. To address these challenges, this project develops a Model Predictive Control (MPC) strategy that anticipates hot water demand and preheats the tank accordingly. This approach enables small-volume HPWHs to maintain sufficient hot water supply while minimizing energy consumption. The results demonstrate the feasibility of HPWHs as an efficient, compact solution for energy-conscious, space-limited residential applications.

Janelle Domantay

Paulo Tabares

Tyrone Vincent

Blue Force logistics in contested environments require plans that remain effective under uncertain and adaptive Red Force actions. This research develops a contested logistics framework that integrates robust optimization with an iterative Blue–Red paradigm. In each iteration, a column-generation procedure proposes optimal Blue strategies, after which an adversarial subproblem identifies the Red tactics that would produce the greatest operational impact to Blue. The resulting adversarial information is then incorporated back into the master model to progressively improve Blue resilience across iterations.

Justin Kilb, Luke Messer, Bobby Provine, Gabe Hake, Brandon Werling, Caleb Fluker

This research develops a Monte Carlo discrete event simulation framework to estimate the value of soiling information for concentrating solar power plants. The model couples stochastic representations of soiling, weather, and measurement error with a dynamic cleaning dispatch policy that links heliostat reflectance to annual energy production and O&M costs. Applied to two representative case studies, the framework shows that both the frequency and accuracy of reflectance measurements can substantially affect plant performance, with more frequent surveys delivering energy gains that far exceed their cost.

Justin Kilb

This research introduces a two-phase optimization framework that selects pre-stage locations and dispatch times for relief supplies as disaster forecasts evolve. A risk-aware prepositioning model balances delivery time against facility disruption risk, while a scenario-based timing model trades off improved forecast accuracy against lost lead time. Together, they provide practical guidance on when to move which supplies to minimize closing time and unmet demand in humanitarian operations.

Justin Kilb

Growing data center electricity demand presents new challenges for continuous and reliable power supply. Hybrid concentrating solar power-photovoltaic systems with thermal energy storage can deliver baseload electricity; incorporating electric resistance heaters enables power-to-heat conversion that improves system flexibility. We extend the National Renewable Energy Laboratory’s Hybrid Optimization and Performance Platform to evaluate electric heater integration in hybrid systems with a 30 MW load– representative of data center demand. Three high-solar resource locations in the U.S. are chosen: Daggett, CA, Fort Stockton, TX, and Phoenix, AZ, utilizing 2024 weather and price data. Systems are optimized for installed cost, levelized cost of energy, and reliability. Results show that while battery storage achieves the lowest missed load (1–6\%), electric-heater-based configurations reduce capital costs and levelized cost of energy by up to 20\% but operate with slightly lower reliability, missing approximately 5\% of the load. Electric heater operation occurs during less than 2\% of annual hours, primarily when surplus PV energy is available and thermal energy storage is not saturated. Overall, electric heaters provide a lower-cost, but lower-efficiency, complement to batteries, expanding the portfolio of viable technologies for hybrid systems in continuous power applications.

We extend an existing mixed-integer linear programming model that determines a cost-minimal set of conventional and renewable technologies and their corresponding dispatch at hourly fidelity to include: (i) multiple years of electrical demand, (ii) multiple investment opportunities, and (iii) technology degradation. Solving the monolithic formulation proves to be computationally inefficient. We utilize a Lagrangian decomposition methodology in which yearly subproblems are solved individually (and in parallel). An upper bound derived via optimization-based heuristics assesses solution quality against a linear-programming-based lower bound. Across twelve case studies involving decades-long planning horizons and projected electrical loads in distributed-generation (college campus) settings, the decomposed model achieves objective values within 0.7\% of the monolithic optimum and maintains optimality gaps below 1\% with solve times that are a small fraction of those exhibited by the monolith. Our approach enables scalable, multi-period energy planning over extended horizons that would otherwise be computationally prohibitive.

Service-oriented businesses rely heavily on labor and expertise, making employee compensation structures central to their financial success. Despite this importance, compensation scales are rarely optimized. The challenge is to design a wage structure that balances business profitability, employee fairness, and competitive pricing. We develop and apply a mixed-integer nonlinear optimization model to determine a tiered compensation structure. The model: (i) ensures that our business partner achieves a desired gross profit per employee, (ii) minimizes deviations in employee wages from the original compensation scale, and (iii) incorporates an overall service price increase. We then refine the optimized output through managerial adjustments to better align with organizational goals of equity across pay tiers. Applying the model to data from a service-oriented business in a major metropolitan area, the optimized scale improves average gross profit per commission tier by 29\%, while employee wages deviate between $-6$\% and +4\% relative to those earned on the original scale. To address managerial priorities of equitable pay at lower tiers, adjustments narrow wage deviations to between $-4$\% and +8\%, but reduce the average gross profit gain. This trade-off highlights the value of optimization as a decision-support tool, rather than to provide a solution implemented verbatim: the model provides a profit-maximizing baseline, while management can apply equity or retention considerations through targeted modifications. Overall, the approach enables service businesses to strategically balance profitability and employee satisfaction, offering a practical framework for data-driven compensation plans.

As space agencies and industry plan long-duration or multi-agent missions, they will increasingly rely on teams of autonomous assets. This PhD project contributes to autonomous decision-making methods that allow Mobile Autonomous Robotic Swarms (MARS) to coordinate work.

The research focuses on decentralized task allocation: how heterogeneous robots decide which agent should do which job. The project combines tools from operations research, multi-robot systems, and distributed optimization to design and analyze algorithms that let swarms:

• share information over communication networks,
• allocate and reallocate tasks when appropriate, and
• exceed overall mission performance compared to traditional centralized planners.

The work supports Lunar Outpost’s MARS framework, developed for end-users/customers at the U.S. Air Force Research Laboratory and U.S. Space Force. By advancing robust, scalable task allocation for autonomous swarms, this PhD contributes to the broader goal of enabling robotic workforces to construct and sustain critical infrastructure in space and on Earth extreme environments.

Joseph Kenrick

Management science researchers have long been asking an important question: “how does team performance depend on team composition?” This performance is a result of a collective team behavior that depends on the composition of the team; thus, it is important to consider the different roles that team members play when investigating team performance. Team roles, or the “style of behavior of a team member” is a topic of research interest as a first step towards investigating team performance and composition. This research uses as a case study a professional basketball league. We develop a novel clustering methodology to identify archetypes in the NBA. We then formulate an optimization model that considers these archetypes and the roles they play within a team context to assist executives in their team-building process. Ultimately, we aim to incorporate uncertainty in these decision-making models given the competitive nature of the player acquisition process in the NBA.

Independent Systems Operators welcome bids from power plants in the day-ahead electricity market in the seven different power markets across the United States. Ancillary services complement energy production to ensure the reliability of a power system and contribute to the goal of matching energy supply with demand at different points in time. Typically, these services respond over short time frames on the order of one second to ten minutes, and include spinning reserves, non-spinning reserves, regulation up, and regulation down.  The extent of ancillary services procurement varies across different power markets, and may make concentrated solar power more attractive the more it can provide system reliability in the way more conventional energy sources do.

Kehinde Abiodun, Karoline Hood

Jesse Wales

As concentrating solar power (CSP) systems become more common, the need for experienced operators arises. We extend a dispatch optimization model to provide operator decision support in a real-time setting. This model considers a more detailed plant representation, including solar resource quality and the resulting power cycle efficiency. We combine the updated CSP dispatch optimization model with other technology models to evaluate the design of more complex hybrid renewable energy plants. A flexible execution framework facilitates the analysis and optimization of future renewable energy installations.

William Hamilton, John Cox

As the demand for water within the South Platte Basin grows, we seek to mitigate the shortage that will ensue by the optimal placement of additional reservoir storage while including water transfers via pipeline.

A problem in the mining industry is production scheduling, or determining when, if ever, notional three-dimensional blocks of ore should be mined. Often lacking in underground production scheduling models is the consideration of heat, accumulated largely by the equipment used for the execution of underground activities such as development, extraction, and back-filling. To correct this, we attribute a specific heat load to each mining activity based on equipment use, autocompression, broken rock and strata rock. We simulate engines used in diesel equipment under the same conditions as those in the mine airway to accurately characterize heat and emissions. We incorporate heat considerations as a knapsack constraint in an integer programming model to generate more realistic schedules; adhering to them could increase revenue by lowering refrigeration costs for the mine and also ensure the health and safety of workers. Future research will allow for the exploration of alternative
fuels and evaluation of their impact on the environment in working areas.

Aaron Swift, John Ayaburi

Concentrated Solar Power (CSP) systems with thermal storage are capable of responsive operation as market or weather conditions vary, but are burdened with an abundance of possible operational schedules that satisfy the engineering requirements of the system. The primary goal of the project is to increase net revenues from electricity sales by anticipating and producing power during high-value time periods, while reducing revenue loss due to maintenance and long-term component wear and tear. This project will extend prior work on System Advisor Model, in CSP dispatch optimization, and in direct normal irradiance forecasting to provide a software tool to be used in situ at operating facilities to execute optimal operational strategies. The reliability of the steam generation systems (SGS) has been the most noted issue regarding availability. The heat exchangers within the SGS have required premature maintenance resulting from manufacturing defects (e.g., tube to tubesheet welding), subjugation to improper water quality during operation, and process design regarding excessive temperature gradients during operation. Thermo-mechanical stress resulting from transient thermal loading and temperature gradients is a main contributor to premature leak-failure within shell-and-tube heat exchangers (STHX). Currently, there is a lack of industry standards or codes to understand the relationship between geometry, operations, and thermo-mechanical stress within complex zones of STHXs. Capturing the underlying physics of operation through well-informed engineering models enables us to populate an optimization model with accurate, policy-dependent operating costs. This proposed work develops a predictive modeling tool that evaluates the thermo-mechanical stress within complex zones of the STHXs across a wide range of off-design operations with considerations of various control schemes representative of a CSP plant.

Phillip Buelow, Anna Perez, Karoline Hood

The National Basketball Association (NBA) is divided into two conferences, each of which is comprised of fifteen teams. At the end of the regular season, the top eight teams from each conference, based on winning percentage, compete in the playoffs. An integer-programming model determines when a team has guaranteed its position in the playoffs, or, conversely, when it has been eliminated before the completion of the regular season. At the end of the regular season, there are instances in which teams’ winning percentages are tied. Ties are broken using seven independent criteria, and we implement these by determining: (i) when a team has been eliminated from the playoffs; and (ii) how many games a team must win in order to clinch a playoff position. The results are published on the RIOT website so fans can follow their favorite teams’ playoff standings. We compare the time at which (and day on which) these results are published as the NBA official standings; in many cases, RIOT notifies
the public prior to the NBA by, on average, 4.1 games. We also describe a scenario in which the NBA erroneously reported the Boston Celtics had clinched a playoff spot and a show that the Golden State Warriors had clinched a playoff spot before the official announcement by the NBA.

We compare our results against what is posted on the NBA website for the 2017-2018 regular season. The figure shows the date on which the respective information source determined when a given team either clinched first place in the conference, clinched a playoff position, or was eliminated from the playoffs for the Eastern Conference. Cells highlighted in green show RIOT outperforming the NBA’s published results and blue represents the case in which the two tie, where we define outperformance as the ability of a model to determine earlier that a given team had either clinched or been eliminated. https://s2.smu.edu/~olinick/riot/detail_nba_numbers.html  

Mark Husted

A mathematical model designs and operates a hybrid power system consisting of diesel generators, photovoltaic cells and battery storage to minimize fuel use at remote sites subject to meeting variable demand profiles, given the following constraints: power generated must meet demand in every time period; power generated by any technology cannot exceed its maximum rating; and best practices should be enforced to prolong the life of the technologies. We solve this optimization model in two phases: (i) we obtain the design and dispatch strategy for an hourly load profile, and (ii) we use the design strategy, derived in (i), as input to produce the optimal dispatch strategy at the minute-level. Our contributions consist of: combining a year-long hourly optimization procurement strategy with a minute-level dispatch strategy, and using a high-fidelity battery model at the minute level derived from electrochemical engineering principles that incorporate temperature and voltage transient effects. We solve both phases of the optimization problem to within 5% of optimality and demonstrate that solutions from the minute-level model more closely match the load, more closely capture battery and generator behavior, and provide fuel savings from a few percent to 30% over that provided by the hour-level model for the tested scenarios.

Mark Husted

Combined Heat and Power Integration

We study a multi-objective design and dispatch optimization model of a solid-oxide fuel cell assembly for unconventional oil and gas production. Fuel cells are galvanic cells which chemically convert hydrocarbon-based fuels to electricity. The Geothermic Fuel Cell concept involves utilizing heat from fuel cells during electricity generation to provide thermal energy required to pyrolyze kerogen into a mixture of oil, hydrocarbon gas and carbon-rich shale coke.

Gladys Anyenya

Large underground hard rock mines are complex industrial projects whose economic success depends on intelligent design and detailed scheduling. The choice of ore extraction method influences decisions regarding the type of infrastructure and sequence of activities for an underground mining operation. Current industry practice selects a single ore extraction method based on ore body characteristics (e.g., grade, depth, and rock hardness), and, as a function of this method, produces an extraction schedule. Considering multiple methods requires navigating geotechnical complexities associated with the interaction of different mining methods; a feasible schedule must consider both intra- and inter-mining-method precedence. We are researching a methodology that combined formal optimization with heuristics to design an underground mine consisting of two extraction methods in order to maximize profitability via its corresponding production schedules over the life of the mine. A heuristic determines by which method an area of the mine is extracted and an optimization model determines the time at which each activity takes place. We empirically demonstrate that the optimization-based heuristic generates solutions that improve profitability.

Peter Nesbitt