Hewlett Packard Labs receives two DOE Fusion initiative awards projects

Meeting the challenges of climate change and building a carbon-free energy grid will require a combination of green energy solutions. Wind, solar, hydroelectric and nuclear (fission) power will all play a role in creating a carbon-free energy grid that powers the world of the near future. Potentially, humanity will also require energy derived from nuclear fusion – the very power source of stars like our sun, and one that has thus far vexed humanity’s greatest minds.

But we’ve made significant progress.

Scientists around the world have gotten closer to finding a way to harness the promising power of fusion. The U.S. Department of Energy (DOE) has recently awarded several multi-million-dollar grants as part of its “machine learning, artificial intelligence, and data resources for fusion energy sciences” initiative. Hewlett Packard Labs (Labs), the applied research organization within Hewlett Packard Enterprise (HPE), will have a groundbreaking role in two of these projects in collaboration with General Atomics and the University of Rochester. These projects will be led by HPE Fellow and Vice President Dr. Paolo Faraboschi, who leads the AI Research Lab at Hewlett Packard Labs.

Hewlett Packard Labs and General Atomics accelerate commercialization of fusion energy

The pace of fundamental research in plasma physics must accelerate for scientists to better understand how to create stable conditions in a fusion reactor relevant for energy production. Put simply, a Fusion Pilot Plant (FPP) must sustain plasmas at temperatures exceeding 100 million degrees Celsius – approximately 10 times the temperature at the center of the sun. Keeping that energy system stable has proved to be one of modern science’s greatest mysteries.

Artificial Intelligence (AI) and Machine Learning (ML) methods can help scientists achieve research goals that will help solve that mystery. However, AI workflows are complex and available data are not in a form directly suitable for their use to build AI models for the FPP design and control.

The DOE identified a need to develop infrastructures for managing, formatting, curating and enabling access to fusion data for development of AI/ML algorithms. It selected a multi-institutional team of data scientists led by General Atomics (GA) to develop a Fusion Data Platform (FDP) to advance fusion research. The FDP will provide researchers with access to high-quality fusion data for the efficient creation of reproducible AI and ML models that will support design and operation of FPPs in coming decades.

GA will partner with Labs, the San Diego Supercomputer Center (SDSC) at the University of California, San Diego (UC San Diego) and SapientAI to create the FDP, which the scientific community will ultimately tap to support rapid advancements in FPP designs. Specifically, Dr. Martin Foltin and his team at Labs will contribute expertise in AI/ML data and algorithms to make large-scale fusion data easier to access and analyze through the FDP.

Labs has developed a Common Metadata Framework that makes it easier to discover data that is suitable for building fusion models. It enables reproducible and reusable workflows to help bootstrap and focus fusion research and advances the development of certifiable models and management of their lifecycles. For mission-critical applications, it is important to capture model provenance (the entire history and processing steps that led to the creation of a model) for auditability and certifiability.

The work with General Atomics is crucial because it will accelerate the commercialization of fusion energy, and the team at Labs is excited to play an important role in helping to achieve objectives on the decadal vision roadmap toward fusion energy.

Hewlett Packard Labs revolutionizing Inertial Confinement Fusion with University of Rochester

To build upon the historic fusion ignition breakthrough in 2022 at the Lawrence Livermore National Laboratory (LLNL), scientists around the world have been actively advancing their research in inertial confinement fusion (ICF), or more colloquially “laser fusion.” The University of Rochester’s Laboratory for Laser Energetics (LLE) is a leader and a key partner in the ICF program. Scientists from LLE are actively engaged in developing laser-driven ICF technology, and their approach involves targeting multiple lasers at small spherical shells or pellets filled with fusion materials, such as deuterium and tritium, to initiate a fusion reaction. This process of firing lasers to produce large energy is a major scientific challenge, and expensive to setup, which limits the number of experiments. To address this challenge, researchers simulate the system to identify the optimal parameters of the experiment. There is an urgent need to bridge the gap between running computational simulations of this process at different fidelities and the experimental design, which deviate significantly. This requires using the power of AI to transfer the learning from these multi-fidelity simulations to physical experiments to help improve the design.

Labs’ AI/ML experts led by Senior Distinguished Technologist Soumyendu Sarkar will work directly with LLE’s fusion scientists to build predictive algorithms using transfer-learning and Reinforcement Learning. This will help accurately guide implosion design and predict experimental outcomes. The objective is to cost effectively run more laser simulations at different fidelities to advance the speed of research and improve the design of the physical lasers. 

Humanity needs carbon-free energy to continue to propel its evolution and innovation while protecting the planet from the devastating effects of climate change. It will take a variety of energy sources and the greatest thinkers to meet that generational and existential challenge. In these two fusion projects with GA and LLE, Labs will use its AI/ML expertise to embrace its role in securing a better, more sustainable future for all.