loading page

Few-Shot Machine Learning at the Grid-Edge: Data-Driven Impedance Models for Model-Free Smart Inverters
  • +5
  • Yufei Li ,
  • Yicheng Liao ,
  • Minjie Chen ,
  • Xiongfei Wang ,
  • Lars Nordström ,
  • Prateek Mittal ,
  • H. Vincent Poor ,
  • Liang Zhao
Yicheng Liao
Author Profile
Minjie Chen
Princeton University, Princeton University

Corresponding Author:[email protected]

Author Profile
Xiongfei Wang
Author Profile
Lars Nordström
Author Profile
Prateek Mittal
Author Profile
H. Vincent Poor
Author Profile
Liang Zhao
Author Profile

Abstract

The future electric grid will be pervasively supported by a large number of smart inverters distributed at the grid edge, whose dynamics are critical for grid stability and resiliency. The operating conditions of these inverters may vary across a wide range, leading to various impedance patterns and complicated grid-inverter interaction behaviors. Existing analytical impedance models require thorough and precise understandings of system parameters and make numerous assumptions to reduce the system complexity. They can hardly capture complete electrical behaviors of physical systems when inverters are controlled with sophisticated algorithms or performing complex functions. Real-world impedance acquisitions across multiple operating points through simulations or measurements are expensive and impractical. Leveraging the recent advances in artificial intelligence and machine learning, we present the InvNet, a few-shot machine learning framework that is capable of characterizing inverter impedance patterns across a wide operation range when only limited impedance data for each inverter is available. The InvNet is capable of extrapolating from physics-based models to real-world models and from inverters to inverters. Comprehensive evaluations were conducted to verify the effectiveness of the proposed approach in various application scenarios. All data and models were open-sourced. We showcase machine learning and neural networks as powerful tools for modeling black-box characteristics of sophisticated grid-edge energy systems and their capabilities of analyzing behaviors of larger-scale systems that cannot be described via traditional analytical methods.