Session: 16-01-01: Government Agency Student Poster Competition

Paper Number: 149872

149872 - Enhanced Digital Twin for Wind Turbines: An Adaptive Kalman Filter Model Based on Parameterized Linearization of System Dynamics 

Wind energy accounted for 10.2% of the electricity used in the U.S. in 2023 and is expected to grow to 21% within the next two years. Meeting this growth necessitates the development of new wind farms and the maintenance of aging systems. Turbines with a 20-year lifespan incur operation and maintenance costs that constitute 15-30% of the overall energy generation cost. Reducing these costs by timely and reliably identifying and addressing maintenance and structural integrity issues is critical for the economic competitiveness of wind energy, a critical factor for transition toward sustainable energy.

Currently, operators rely on data from a turbine's Supervisory Control and Data Acquisition (SCADA) system to identify issues by detecting anomalous trends. While SCADA tracks data such as wind speed, rotor speed, and power, it does not provide detailed information such as the forces experienced at specific points on the rotor or tower. To bridge this gap, advanced control and management systems are being developed to offer more detailed information, enabling operators to pinpoint maintenance issues more accurately and optimize repair times and costs.

Digital Twins (DT) have recently shown significant promise in providing real-time, detailed information about dynamic systems conditions using SCADA data. Specifically for wind turbines, the National Renewable Energy Laboratory (NREL) has developed a DT model using an augmented Kalman Filter (KF), achieving predictions within 10% of the actual turbine's performance. This model employs system, measurement, control, and direct transition matrices derived by averaging linearizations from OpenFAST – a modular, open-source software tool for simulating wind turbine dynamics – across different wind speeds and rotor displacements. Although this averaging simplifies the filter design for various operating conditions, it introduces errors because the matrices may not adequately reflect system dynamics for specific rotor angles and wind speeds during the course of exposure of the turbine to nonstationary winds. This simplification contributes to the relatively high error between the DT and the actual turbine response.

This research proposes a new approach to improve the accuracy and reliability of DTs by deriving the linearized state-space model matrices in a parameterized form with respect to wind speed and rotor displacement. A suit of surrogate models is explored for parameterizing system matrices including radial basis functions, Gaussian regression processes, and neural networks.  Conditional on wind speed and rotor displacement, the surrogate models are used to update state-space matrices at each time step to more closely reflect the true dynamics of the turbine system. Rotor speed is derived from the KF state vector, while turbulent wind speed is derived from the pitch angle, rotor velocity (from the KF measurement vector), and torque (another KF state). As the parameterization is with respect to the mean wind, the derived turbulent wind speed is processed using a band-pass filter to obtain the time-varying mean wind field.

The proposed methodology has been applied to NREL’s synthetic 5-MW onshore wind turbine and tested using a series of turbulent wind fields with varying mean wind profile to cover diverse operating regions. The enhancements in the predictive DT model will improve the performance of turbine control and management systems, enabling operators to reduce maintenance costs by reliably identifying issues before they become critical and costly.

Presenting Author: Ripley Vichot The Ohio State University

Presenting Author Biography: Hello, I am a fourth year civil engineering student at The Ohio State University. Over the summer I got to participate in the Resilient and Sustainable Infrastructure Systems for Smart Cities Research Experience for Undergraduates hosted by the Department of Civil, Environmental, and Geodetic Engineering at OSU. I am excited at the possibility to present the project topic I got to work on during that program!

Authors:

Ripley Vichot The Ohio State University
Minhyeok Ko The Ohio State University
Jieun Hur The Ohio State University
Abdollah Shafieezadeh The Ohio State University