(548c) Scalable Computing for Energy Transition

The energy system is undergoing transformative change, including integrating renewable resources and grid storage, expanding data centers, and electrifying transportation, buildings, and industry. At the core of this transition is the electrical grid, which connects diverse components that must be planned and operated reliably and efficiently. These decision-making problems—planning and operation—are inherently complex. They are influenced by uncertainties in renewable generation, fuel prices, and extreme weather, as well as by a wide array of spatial and temporal scales—from the microscopic physics of battery cells to large-scale energy storage—and by network interdependencies among generators, loads, renewables, and storage.

Scalable computing is one of the vital tools for addressing these challenges. In particular, advances in parallel computing and general-purpose GPU acceleration have enabled massive simulation, optimization, and machine learning tasks to be completed quickly enough for real-time decision-making. For example, our GPU-accelerated nonlinear optimization can compute the economic dispatch for the 70,000-bus eastern interconnection in 15 seconds—significantly faster than earlier methods that required more than three minutes. Furthermore, space-time decomposition enables tackling large-scale, long-term planning problems across large networks over multiple years. Combining GPU acceleration with decomposition techniques helps manage these extreme complexities in energy system decision-making.

This presentation explores how scalable computing supports decision-making across a variety of energy applications. We begin with an overview of GPU-based optimization and decomposition techniques, then illustrate their different use cases, including (i) power system operations with storage and electrochemical manufacturing, (ii) carbon capture process design that is robust against fluctuating power plant loads, (iii) operation of grid-scale batteries to balance their utilization and degradation optimally, and (iv) flexible operation of buildings and data centers. We conclude by examining the future impact of computing on addressing ever more complex decision-making problems in the energy transition.