DC+: A Fast, Voltage-Aware Alternative To DC Load Flow
summary
A voltage‑sensitive linearization of AC load flow retains voltage magnitudes, angles and reactive power while behaving like a single Newton step around an AC base case, capturing violations far more accurately than classical DC. Implementation uses a Woodbury update with one LU decomposition and preadded empty buses to enable bus split actions. GPU code reaches about 1,000,000 load flows/s for large grids; a CPU flavor achieves ~25,000/s (Q limits, distributed slack) and typically converges in one–two iterations, falling back to full Newton for a few hard cases.
AI-written summary · about
chapters (7)
- 0:00Introduction to DC Plus
- 6:49Project background and goals
- 9:55Core ideas behind DC Plus
- 14:13Research results and accuracy
- 18:25Features and capabilities
- 20:51Collaboration and future steps
- 21:42Q&A session
related talks
31mGridFM - Towards Foundation Models for the Electric Grid
12mEfficient Workflow Serving for Diffusion Models with Many Adapters
21mGPU-accelerated Hydrology Algorithms for On-prem Computation: Flow accumulation,(…)
26mTowards an AI Assistant for Tertiary Voltage Control
14mKairox: Adaptive GPU-CPU Hybrid LLM Inference via Online Neuron Balancing
13mARK: GPU-driven Code Execution for Distributed Deep Learning
18mFAST: An Efficient Scheduler for All-to-All GPU Communication
19mYou Might Not Need 50 Diffusion Steps