IN MAY, THE US DEPARTMENT of Energy (DOE) announced $27 million in funding for nine projects towards developing ‘digital twin’ technology to reduce operations and maintenance costs in advanced reactors.
The funding is part of the Generating Electricity Managed by Intelligent Nuclear Assets (GEMINA) programme, managed by US DOE’s Advanced Research Projects Agency-Energy (ARPA-E). ARPA-E was set up in 2009 to fund potentially high-impact energy technologies that are too early for private sector investment. It funds R&D for all types of clean energy technology, but has only been active in nuclear energy for about three years. ARPA-E has funded 25 nuclear projects since 2017, to the tune of about $70 million, says Dr Rachel Slaybaugh, GEMINA programme director, and nuclear engineering professor at the University of California, Berkeley.
GEMINA is hoping to reduce operation and maintenance (O&M) costs for the next generation of nuclear plants by “up to ten times,” in order to make them more economical, flexible and efficient. The project teams plan to do this by developing digital twins and associated technologies for a range of advanced reactor designs. The digital twins will be used to strategically design future O&M frameworks, and to create tools to increase flexibility, increase autonomy and speed up design iteration. “The way you design a system now determines how it will be operated and maintained,” says Slaybaugh. “If you don’t think strategically about O&M from the outset, it might set you up to have a system that is not competitive.” That is born out by the digital twin plan: there is a limit to the extent that digital technologies can drive down costs in existing reactors, because they are not equipped with the digital sensors needed to gather data to support advanced predictive maintenance techniques.
Improving the economic case of nuclear generation sources is critical to the future of the nuclear industry. In 2019, the US nuclear fleet generated a record 809TWh of electricity, around fifth of the USA’s total, but more than half of the country’s emissions-free electricity, according to the US Nuclear Energy Institute (NEI). Investing in projects and R&D that will make the nuclear fleet more efficient and cost-effective is critical to ensuring nuclear energy continues to power the US for years to come, according to Mark W. Menezes, Under Secretary of Energy.
Nuclear power’s average total generating cost has decreased by around a third, from $44.57/MWh in 2012 to $30.42/MWh in 2019, the NEI says, citing figures from the Electric Utility Cost Group.
While the focus is often on levelised cost of energy (LCOE), this does not give the whole picture. Operating costs are as important, particularly in the US, where reactors have closed down for economic reasons in recent years. In fact, O&M costs make up almost all of the generating cost for the current nuclear fleet, says Slaybaugh, and fixed O&M costs for nuclear are about an order of magnitude higher than those for natural gas (see Figure 1).
GEMINA is focused on novel digital technologies to achieve significant and sustainable reductions in O&M costs. These advances will lay the groundwork for a future where advanced reactors operate with a staffing plan and fixed O&M costs more competitive with those of other generation sources.
“Reducing operations and maintenance costs are essential to increasing the economic competitiveness of nuclear energy,” says Pradeep Ramuhalli, technical lead for the Oak Ridge National Laboratory (ORNL) effort in one of the project funding recipients. “Digital twin technology provides a means for nuclear power plants to reduce their O&M costs by enabling risk-informed decisions on plant operations, system reconfiguration, and predictive maintenance planning that optimise cost without compromising safety.”
The nine projects that received funding under the GEMINA programme in May are listed in the table. They cover a range of advanced reactor technologies, including molten salt reactors (MSRs), high-temperature reactors (HTRs), and boiling water reactors (BWRs). Funding amounts range from $899,825 to $6 million.
“We have picked several different reactor types to hit the different categories,” Slaybaugh said when questioned about opportunities for collaboration on strategies for O&M between different reactor vendors. “Teams have a balance of universities and national labs that can take the cutting- edge findings and make them available. Each team also has a reactor developer company on it. With this configuration, there is a commercial path forward and major lessons can be shared.”
Modelling maintenance for MSRs
The University of Michigan received almost $5.2 million to develop physics-based, model-centric, and scalable capabilities to achieve unprecedented integrated state awareness for advanced reactor power plants. Modules will include: a scalable digital twin; a maintenance proactive evaluator to monitor usage and assess the health conditions and maintenance needs of advanced reactors; an operations intelligent controller to achieve autonomous control during normal and accident conditions; and an O&M deep supervisor to supervise O&M conditions. The team intends to validate the product using a molten salt loop operating at the University of Michigan and apply it to Kairos Power’s fluoride salt-cooled HTR design to demonstrate how the proposed capability can be used to optimise plant design.
In partnership with Kairos Power, the MARS project led by Argonne National Laboratory aims to reduce the operating and management costs of Kairos’ reactor design through advanced sensing and instrumentation. With $2.2 million in funding, scientists will develop new sensors that can handle the high temperatures and chemical environment inside the reactor, and will develop algorithms using machine learning to analyse sensor data, helping to automate reactor monitoring. “We’re trying to make the new reactors as inexpensive and safe as possible,” says Nathaniel Hoyt, manager of the Process Simulation and Safeguards group at Argonne. “Through advanced sensors and automation, we can reduce the amount of money it takes to operate and monitor nuclear reactors.” The proposed methods aim to achieve O&M costs on the order of $2/MWh.
In another project, Moltex Energy is aiming to develop a multi-physics plant digital twin environment for its Stable Salt Reactor-Wasteburner design. As part of the project a non-nuclear separate-effects test loop will support the digital twin development, to validate larger uncertainties in flow conditions providing a virtual test environment for simulating operations and maintenance strategies in SSR-W.
A two-year project led by MIT in collaboration with MPR Associates aims to gain better understanding of the behaviour of radioactive materials produced in the cores of MSRs to predict and reduce O&M costs. The team will use MIT’s research reactor to irradiate molten salts containing fuel in materials planned for MSR construction, while monitoring the released radioactivity. This data is needed for the development of digital twins.
MIT will also work closely with the Electric Power Research Institute (EPRI) on a proof of concept study that will explore moving away from the traditional ‘maintain and repair’ approach. The EPRI-led project will examine a ‘replace and refurbish’ model, in which components are designed with shorter, more predictable lifetimes — similar to the approach adopted by the commercial airline industry.
“This collaborative project will take a fresh look at reducing O&M cost by allowing nuclear technology to better adapt to the ever-changing energy market conditions,” says MIT’s project-lead Koroush Shirvan, professor in the Department of Nuclear Science and Engineering. MIT’s role is to identify cost-reducing pathways that would be applicable across a range of advanced reactor technologies.
Developing digital twins for BWRX-300
GE Research and MIT have been awarded grants to lead the project teams that will develop digital twin technology, the key to employing artificial intelligence and advanced modelling controls.
The GE Research-led team (which includes Exelon Generation, Oak Ridge National Laboratory, the University of Tennessee-Knoxville and GE Hitachi [GEH]) will build a digital twin of BWRX-300 critical components and use predictive technologies (ie, artificial intelligence) to make risk-informed decisions. Exelon, which operates the largest US fleet of nuclear plants, will provide historical data to inform the model and O&M cost reduction targets for advanced reactors. Abhinav Saxena, a senior AI scientist at GE Research and project leader, says that GE has developed and deployed over 1.2 million digital twins within an array of products and sectors.
“We’re excited at the prospect of applying GE’s digital twin technology and our novel concept of Humble AI to advanced nuclear reactors,” Saxena says. “Humble AI is part of a new lexicon of AI terms emerging, as AI becomes integrated into critical industrial infrastructure where safety, reliability and performance are paramount. It will allow us to deliver improved performance and services, while maintaining and even enhancing safety and reliability.”
GE has been already been piloting its Humble AI technology with wind turbines in the field and gas turbine combustion, which have resulted in higher efficiency and energy output. The goal is to bring those same benefits, along with reduced operational and maintenance costs and more plant automation, to the nuclear sector.
In a second project an MIT-led team (which includes GE Research and GEH) will demonstrate predictive maintenance approaches and model-based fault system detection techniques.
The digital twins will address mechanical and thermal fatigue failure modes, which drive O&M activities.
Cutting costs for HTRs
Two research projects focus on high-temperature gas- cooled reactor technologies.
Framatome will develop two novel digital twins for use with Metroscope, a software package that connects digital twins and their associated fault libraries and monitors them, using an algorithm to detect problems early on. Argonne National Laboratory is designing a passive cooling system for the digital twins.
Led by Argonne scientist Darius Lisowski, the project will simulate natural air and water circulation reactor cooling, using data from Argonne’s Natural Convection Shutdown Heat Removal Test Facility. This data will be used to develop Framatome’s digital twins and improve their reliability and maturity. Framatome will then use the digital twins to compare the passive cooling system with a typical cooling circuit, to identify the superior system for the steam cycle high-temperature gas-cooled reactor.
US-based X-energy has also been awarded $6 million for a digital twin project that aims to reduce the fixed O&M cost of its Xe-100 HTR reactor design to $2/MWh.
The project will use human factor engineering, probability risk assessment, hazard analysis, and security and maintenance evaluations to identify areas for optimisation. X-energy will also use advanced technologies such as automation, robotics, remote and centralised maintenance, and monitoring to optimise staffing plans and ensure optimal plant operation.
The team will develop two modelling frameworks to evaluate and validate these solutions. The Immersive Environment Toolset is a multi-disciplinary 3D model that will use virtual reality to test techniques. The digital twin framework synthesizes information from the operating plant and assimilates data from across the fleet to provide a holistic understanding of the asset.
What’s next?
Initially it is hoped that the combination of physical and computational experiments will be enough to create accurate digital models for a range of designs.
Then, as more sophisticated experimental facilities and demonstration reactors are built, more accurate data can be generated and the models improved.
The DOE is designing a versatile test reactor, which could enter operation by 2026, while Oklo, Terrapower, Karios and NuScale are planning for reactor demonstrations to come online as early as mid-2025. “Hopefully the ideas developed in this programme will be proven out in the experimental facilities,” says Slaybaugh.
The goal is ultimately for the tools developed to become standard and adopted by advanced reactors in the future, for example, as licensable software packages, she adds.
