AI Grid Dispatch System GreenGrid Deployed in California: Renewable Energy Curtailment Drops from 12% to 1.8%

The California Independent System Operator announced full deployment of the AI grid dispatch system GreenGrid in October. Developed jointly by Google DeepMind and CAISO, the system uses deep reinforcement learning to predict wind and solar generation 72 hours ahead, combined with demand forecasting to dynamically dispatch battery storage systems and industrial flexible loads.

In Q3 2028 operational data, GreenGrid reduced Californias renewable energy curtailment rate from 12% a year earlier to 1.8%, equivalent to absorbing approximately 4.2 terawatt-hours of additional clean electricity. CAISO stated this improvement corresponds to roughly 2 million tons of avoided CO2 emissions.

GreenGrids dispatch decision latency is 800 milliseconds, approximately 100 times faster than traditional rule-based dispatch systems. The system has been placed on the 2029 technology evaluation lists of grid operators in Germany and Australia.