SOURCE-LINKED INTELLIGENCE

ReliAble Online LeaRning in Power ConverTErs to Unlock Flexibility from Motor AppliCaTions

CORDIS · observation · Publication date unknown

ized framework that harnesses advanced edge-computing in modern power electronic converters to enable motors to self-quantify and aggregate their flexibilities at scale. By rethinking fundamentals of machine learning and system identification, we introduce several key breakthroughs: 1) Reinventing ideas from experimental design theory to filter high-value signals from noise and enable efficient real-time learning in resource-constrained edge devices; 2) Combining physics-informed, offline-trained Bayesian priors with the novel concept of temporal posterior fusion, to robustly learn knowledge across various motor mission profiles and preventing catastrophic forgetting; 3) Developing new probabilistic flexibility envelopes to quantify flexibility with formal uncertainty bounds, while ensuring operational safety; 4) Establishing a fully decentralized aggregation mechanism that will enable millions of motors to self-organize and provide grid services with minimal central intervention, eliminating reliance on historical datasets and centralized processing, and aligning with strict data-pr

Read original source ↗ Open in workspace

recordType

award

status

SIGNED

region

EU

value

2000000

unit

EUR