Peter Graf's research with NREL's Computational Science Center is focused on bringing state-of-the-art applied mathematics to bear on computational problems involving renewable energy. Graf has pursued this theme for nearly 20 years across the spectrum of scales from atomistic materials discovery to distributed power systems control. Current research is split roughly between artificial intelligence (AI) and quantum computing. 

In specific, Graf leads a team studying the applicability of novel, hybrid, "neurosymbolic" AI to NREL domain problems, especially in incorporating reinforcement learning into distributed control of energy systems, including the power grid, buildings operation, transportation networks, and wind farms.

Graf also leads NREL’s efforts in quantum computing algorithms for near-term quantum hardware, involving both adapting energy-related problems to known algorithms, such as quantum approximate optimization algorithm and variational quantum eigensolver, and exploring beyond these toward the ultimate capabilities of the quantum computational model as they relate to energy applications.

Past research includes inverse material design, multi-scale simulation and optimization of organic photovoltaics, multi-scale simulation and parameter estimation of Li-ion batteries, statistical loads estimation and stochastic optimization of wind turbines and wind farms, hierarchical simulation and optimization in systems biology, biomass characterization, multimodal sensor fusion for infrastructure-based perception and control of traffic, reinforcement learning for connected and autonomous vehicles control, and more. 

Graf attended Stanford University, where he graduated Phi Beta Kappa, with distinction, from the Symbolic Systems Program in 1989. After years as a C/C++ programmer, he entered graduate school and in 2003 received his doctorate in mathematics from the University of California at Berkeley. His thesis involved optimization of model reduction for systems of ordinary differential equations, under advisor Alexandre Chorin.

Research Interests

Optimization

AI

Neurosymbolic AI

Quantum computing

Multi-scale modeling

Parameter estimation

Education

Ph.D., Mathematics, University of California at Berkeley

B.S., Symbolic Systems, Stanford University

Featured Work