Wind Turbine RUL Modeling Using Synchronized Load and SCADA Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating the remaining useful life of wind turbines are complex, unreliable, and require significant resources due to high uncertainty and high effort, especially when using simulation models and SCADA operating data, which often result in conservative RUL results.
Innovation Solution
A computer-implemented method for training a machine learning model using synchronized operation and load data sets from a reference wind turbine to determine the instantaneous turbine condition and predict the remaining useful life of another wind turbine with increased accuracy, employing artificial neural networks and regularization techniques to reduce overfitting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If simulation models are used to estimate RUL with SCADA operating data, then RUL estimation can be performed, but the results are conservative and unreliable due to high uncertainty
Solution Approach 1:
The patent creates a virtual copy of the wind turbine through digital twin technology, where a physics-based simulation model is continuously updated with real SCADA operating data. This digital copy replicates the actual turbine's behavior, allowing RUL estimation from the virtual model rather than direct measurement, thereby reducing uncertainty and improving both reliability and precision of RUL predictions
Solution Approach 2:
The system implements continuous feedback loops where real operating data from the wind turbine is fed back into the simulation model to update and refine the digital twin. This feedback mechanism allows the model to adapt to actual turbine conditions, reducing uncertainties over time and improving the accuracy and reliability of RUL estimates through iterative learning and calibration
2Measurement precision
If continuous load measurements are taken on all wind turbines throughout lifetime, then most accurate RUL prediction is achieved, but the costs are enormous and not feasible in practice
Solution Approach 1:
The patent develops a universal RUL prediction approach that works across multiple wind turbines of the same type using a common physics-based model framework. The digital twin technology allows one model structure to serve multiple turbines, reducing the need for extensive individual measurements on each turbine while maintaining accurate predictions through the model's ability to generalize across the turbine fleet
Solution Approach 2:
Instead of performing expensive continuous measurements on all turbines, the system creates virtual copies (digital twins) that simulate turbine behavior. These digital copies allow accurate RUL prediction without physical measurement infrastructure, eliminating the need for costly load measurement equipment while maintaining prediction accuracy through physics-based modeling and SCADA data integration
3Reliability
If machine learning models are trained with synchronized operation and load data, then RUL determination becomes more reliable and accurate, but the data processing complexity increases
Solution Approach 1:
The patent performs preliminary synchronization and preprocessing of operation and load data before feeding them to machine learning models. By pre-aligning the datasets in time and space, and preparing standardized input formats, the system reduces the computational complexity during model training and deployment while maintaining the reliability benefits of using synchronized multi-source data for RUL determination
Data Source
AI summary
The application relates to a method, in particular a computer-implemented method, for training a machine learning model usable for determining a remaining useful life of a wind turbine, including providing a plurality of operation data sets of a reference wind turbine, providing a plurality of load data sets of the reference wind turbine, wherein a load data set is based on at least one load parameter measured at the reference wind turbine, and generating a plurality of wind turbine training data sets for training a machine learning model by synchronously assigning a respective operation data set with a respective load data set.