Project Title: “A Machine Learning based Automated Load Model Parameterization Tool”
Project Summary:
In this project, the team will develop a machine learning based, automated Load Model Parameterization (LMP) tool for load disaggregation and load model parameterization. In the past, due to the lack of data, load composition and load model parameters are estimated only for critical cases such as peak or light load periods. As a result, only snap-shots of the system operation conditions are modeled in power system planning and operation studies. In the near future, the integration of distributed energy resources such as photovoltaics (PV), electric vehicles, controllable loads, or energy storage devices, will change not only steady-state load shapes but also load dynamic response characteristics. Energy efficiency and demand response (DR) programs targeting time-of-use rates have already created observable impacts on where system load peaks occur. However, the load impacts on voltage stability and transient stability are largely unknown due to the lack of dynamic load model parameters that reflect those new technology advancements. Therefore, in this proposed activity, we plan to use data collected from phasor measurement units (PMUs) and SCADA networks for load disaggregation and load model parameter derivation. The tool will have three main functions. First, it provides a clustering function to identify critical scenarios for dynamic responses. Second, it generates parameters for different load groups (e.g., ZIP, motor loads, and power electronics). Third, it automatically populates load model parameters for thousands of load nodes in feeders at both transmission and distribution levels for power flow or transient stability studies.
University Team Members:
Lead PI: Dr. Ning Lu – NC State University
Co-PI Members:
Dr. Ramtin Hadidi – Clemson University
Industry Advisors:
Kat Sico – Duke Enegy
Rebecca Rye – Dominion Energy