Data-driven virtual sensor for online loads estimation of drivetrain of wind turbines

Abstract

Data-driven approaches have gained interest recently in the field of wind energy. Data-driven online estimators have been investigated and demonstrated in several applications such as online loads estimation, wake center position estimations, online damage estimation. The present work demonstrates the application of machine learning algorithms to formulate an estimator of the internal loads acting on the bearings of the drivetrain of onshore wind turbines. The loads estimator is implemented as a linear state-space model that is augmented with a non-linear feed-forward neural network. The estimator infers the loads time series as a function of the standard measurements from the SCADA and condition monitoring systems (CMS). A formal analysis of the available data is carried out to define the structure of the virtual sensor regarding the order of the models, number of states, architecture of neural networks. Correlation coefficient of 98% in the time domain and matching of the frequency signature are achieved. Several applications are mentioned and discussed in this work such as online estimation of the forces for monitoring and model predictive control applications.

On the Use of Indirect Measurements in Virtual Sensors for Renewable Energies: A Review

“…Regarding hybrid methods, we have observed a growing interest among researchers in hybrid approaches that demonstrate excellent performance by combining the advantages of data-based methods with a foundational layer of physics. The vast majority of these methods rely on sequential Bayesian updating, although recent attempts have been made to integrate state-space models with neural networks, as demonstrated by the application mentioned in the work of [119]. Other methods that blend physics with machine learning exist but have not yet been exploited enough in the context of virtual sensors applied to renewable energies, such as PIGP and PINN, which have the ability to incorporate the knowledge of physical laws governing a dataset into the learning process.…”

“…The performance evaluation in a numerical case study showed moderate to high correlations between estimated and actual measurements, with fatigue damage estimation errors ranging from 5% to 15%. Kamel et al [119] developed a datadriven virtual sensor for estimating the internal loads on the drivetrain bearings of WT using SCADA data (generator rotation and power, gear displacements, and bearing acceleration). They first identified a linear state-space model, and instead of applying a Kalman filter as is usually done, they attempted to explore a linear state-space model augmented by a FNN to predict the error between the dynamic response of the linear model and the actual system response.…”