Dynamic Model Identification for Inverter-Based Resources

Historically, synchronous generators, invented in 1880s, are the main work horse for electricity generation. To date, generic models of synchronous machines for power grid dynamic analysis are well developed and indispensable for power grid dynamic analysis. The current power grid is going through a significant transition.

Penetration of inverter-based resources (IBR), e.g., wind, solar, and batteries, is going up. For example, the U.S. Energy Information Administration’s hourly electric grid monitor shows that on May 2, 2020, 59% of power was generated by wind energy in Electric Reliability Council of Texas.

IBRs significantly change grid dynamic characteristics. Unprecedented dynamic phenomena appeared in power grids. Examples include subsynchronous oscillations that occurred in 2009 and 2017 in Texas wind farms, low-frequency oscillations in an offshore wind farm that caused Great Britain power disruption on 9 August 9 2019, subcycle overvoltage dynamics which triggered large-scale solar photovoltaic (PV) tripping in California in 2017 and 2018, and instantaneous ac overcurrent dynamics which caused another large-scale PV tripping event in California in July 2020.

High penetration of IBRs leads to a significant change in the modeling practice of the bulk power system industry. It is expected that generic models with standard structures will be used for grid dynamic evaluation as a future trend. On the other hand, models provided by original equipment manufacturers (OEMs) are black boxes.

Real-code models provided by OEMs are usually in dynamic-linked library format with input and output specified while internal details not released. Though these real-code models have been benchmarked by hardware experiments, detailed structures and parameters are unknown. Thus, there is an urgent need: How to find the generic model’s parameters based on what we have, which are essentially black boxes?

The objective of the proposed research is to address the urgent need to design a gray-box model identification framework for IBRs. This project makes two significant innovations, including (i) first principle based IBR gray-box model structure design and linear time invariant model derivation; and (ii) measurement-based IBR model structure parameter estimation relying on black-box model identification and advanced computing algorithms.

The notable innovation is the integration of the two core technologies from two different fields: Power Electronics and System Identification. The two core technologies are: admittance-based black-box model measurement and identification and gray-box model identification through optimization problem formulation and solving. The project employs both computer simulation and hardware experiments for validation.

This project tackles real-world IBR modeling challenges and can generate significant impact to the current power grid industry. The tackled problem falls into the category of gray-box model identification and this research will also generate reference values to many other domains (e.g., automotive systems and airplane systems) that use gray-box model identification approaches for model building and control design.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.