As power systems grow in complexity, the size of optimization problems is outpacing what traditional workflows were designed to handle. This article examines how GPU-based algorithms can change the computational envelope for large-scale energy optimization, with benchmarks on Brazil’s dispatch LP.

SDDP has been the central tool for hydrothermal dispatch planning for over three decades, but its structural assumptions shape the modeling choices available in practice. This article explores what becomes possible when reinforcement learning is paired with optimization to relax those constraints.

Translating a research paper into production software typically takes months of engineering effort. This article reports a deliberate experiment in AI-assisted development, using a stochastic generation-expansion solver as a test case for what frontier coding agents can deliver today.

Understanding which hydrological operational constraints generate the highest system costs is essential for more efficient planning. A methodology developed in partnership with ONS (National Electric System Operator) quantifies their economic impact, supporting the objective prioritization of flexibility measures and guiding more efficient investment decisions.

The SINAIS framework, developed in partnership with ONS (National Electric System Operator), introduces a multidimensional approach to assessing grid health. Using a set of five models, it quantifies energy reliability, ramping flexibility, and resilience during peak demand periods. By capturing tail risks through Conditional Value at Risk (CVaR) and simulating real-time reserve adjustments, SINAIS provides a data-driven roadmap for the energy transition.