Ge Vernova Inc. · NYSE:GEV
Ge Vernova — Technical Presentation (July 2024)
6 slides from Ge Vernova's technical presentation, July 2024
Slide 1: Using ML training computations for grid stability in 2050
Using ML training computations for grid stability in 2050 Scott Evans* Tapan Shah* *: GE Vernova Research
Slide 2: Motivation
Motivation • Zero-carbon regime in 2050--> high renewable mixes • Increase vulnerability of power grids to black outs • Solar/Wind are prone to under/over-utilization • Push for Controllable loads • Battery storage • Hydrogen • Vehicle-to-grid etc. Some prior work in using compute-centers as control
Slide 3: Proposed Approach
Proposed Approach Machine Information Thermodynamic Learning Work Work https://openai.com/research/ai-and-compute Explore Machine Learning (information work) as a controllable loads against battery storage and hydrogen conversion
Slide 4: Comparison with Battery and Hydrogen
Comparison with Battery and Hydrogen
Slide 5: Scenario Evaluation
Scenario Evaluation Scneario 1: Grid stabilization by moving jobs Scenario 2: Grid stabilization by modulating quality
Slide 6: Conclusion and next steps
Conclusion and next steps • Preliminary research at concept stage • Transferring bits more efficient than transferring electricity to achieve grid stability. • Machine Learning as computational load has multiple tunable knobs eg. number of epochs, desired accuracy, training locations etc. • Batterie
GE Vernova — Technical Presentation (July 2024) on 6ix