Virtual fatigue diagnostics of wake-affected wind turbine via Gaussian Process Regression

Avendaño-Valencia, Luis David; Abdallah, Imad; Chatzi, Eleni

doi:10.1016/j.renene.2021.02.003

Cited by 48 publications

(16 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is possibly related to the worse performance of the model under wake. Avendaño-Valencia et al (2021) worked in this direction, concluding that the fatigue life of OWTs under freestream inflow can be quite distinct from OWTs under wake (Avendaño-Valencia et al, 2021). This can be further verified by inspecting Fig.…”

Section: Fatigue Rate (Del) Estimationmentioning

confidence: 56%

“…Similarly, Avendaño-Valencia et al (2021) has used Gaussian process regression time series modelling to evaluate the influence so-called EOPs (environmental and operational parameters) have on the features of the vibration response of the wind turbine blades. Also applied to estimate the blade root flapwise damage equivalent loads (DELs), Schröder's (2020) work has emphatically demonstrated how a surrogate model based on ANNs outperforms other surrogate models, such as polynomial chaos expansion and quadratic response surface, in computational time, model accuracy and robustness, further applying it to connect wind farm loads to turbine failures (Schröder, 2020).…”

Section: Use Of Machine Learning In (Offshore) Windmentioning

confidence: 99%

See 1 more Smart Citation

Data-driven farm-wide fatigue estimation on jacket-foundation OWTs for multiple SHM setups

et al. 2022

View full text Add to dashboard Cite

Abstract. The sustained development over the past decades of the offshore wind industry has seen older wind farms beginning to reach their design lifetime. This has led to a greater interest in wind turbine fatigue, the remaining useful lifetime and lifetime extensions. In an attempt to quantify the progression of fatigue life for offshore wind turbines, also referred to as a fatigue assessment, structural health monitoring (SHM) appears as a valuable contribution. Accurate information from a SHM system can enable informed decisions regarding lifetime extensions. Unfortunately direct measurement of fatigue loads typically revolves around the use of strain gauges, and the installation of strain gauges on all turbines of a given farm is generally not considered economically feasible. However, when we consider that great numbers of data, such as supervisory control and data acquisition (SCADA) and accelerometer data (of cheaper installation than strain gauges), are already being captured, these data might be used to circumvent the lack of direct measurements. It is then highly relevant to know what is the minimal sensor instrumentation required for a proper fatigue assessment. In order to determine this minimal instrumentation, a data-driven methodology is developed for real-world jacket-foundation offshore wind turbines (OWTs). In the current study the availability of high-frequency SCADA (1 Hz) and acceleration data (>1 Hz) as well as regular 10 min SCADA is taken as the starting point. Along these measured values, the current work also investigates the inclusion of an estimate of the quasi-static thrust load using the 1 s SCADA using an artificial neural network (ANN). After data collection all data are transformed to features on a 10 min interval (feature generation). When considering all possible variations a total of 430 features was obtained. To reduce the dimensionality of the problem this work performs a comparative analysis of feature selection algorithms. The features selected by each method are compared and related to the sensors to decide on the most cost-effective instrumentation of the OWT. The variables chosen by the best-performing feature selection algorithm then serve as the input for a second ANN, which estimates the tower fore–aft (FA) bending moment damage equivalent loads (DELs), a valuable metric closely related to fatigue. This approach can then be understood as a two-tier model: the first tier concerns itself with engineering and processing 10 min features, which will serve as an input for the second tier that estimates the DELs. It is this two-tier methodology that is used to assess the performance of eight realistic instrumentation setups (ranging from 10 min SCADA to 1 s SCADA, thrust load and dedicated tower SHM accelerometers). Amongst other findings, it was seen that accelerations are essential for the model's generalization. The best-performing instrumentation setup is looked at in greater depth, with validation results of the tower FA DEL ANN model showing an accuracy of around 1 % (MAE) for the training turbine and below 3 % for other turbines, with a slight underprediction of fatigue rates. Finally, the ANN DEL estimation model – based on two intermediate instrumentation setups (combinations of 1 s SCADA, thrust load, low quality accelerations) – is employed in a farm-wide setting, and the probable causes for outlier behaviour are investigated.

show abstract

Section: Fatigue Rate (Del) Estimationmentioning

confidence: 56%

Section: Use Of Machine Learning In (Offshore) Windmentioning

confidence: 99%

Data-driven farm-wide fatigue estimation on jacket-foundation OWTs for multiple SHM setups

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Ziegler et al 10 present a methodology for wind turbines that opts for a regression based prediction when measured data are available at the location of interest as an alternative to the updated finite element model approach discussed previously. Avendano-Valencia et al 11 use Gaussian process (GP) regression to provide a probabilistic fatigue estimate also based on the DEL concept, predicting the equivalent load using readily available data such as wind speed and turbulence intensity. Virtual loads monitoring of aircraft have used global parameters such airspeed and local measurements of, for example, acceleration to build regression models for predicting loads elsewhere.…”

Section: Introductionmentioning

confidence: 99%

Distributions of fatigue damage from data-driven strain prediction using Gaussian process regression

Gibson

Rogers

Cross

2023

Structural Health Monitoring

View full text Add to dashboard Cite

Fatigue is a leading cause of structural failure; however, monitoring and prediction of damage accumulation remains an open problem, particularly in complex environments where maintaining sensing equipment is challenging. As a result, there is a growing interest in virtual loads monitoring, or inferential sensing, particularly for predicting strain in areas of interest using machine learning methods. This paper pursues a probabilistic approach, relying on a Gaussian process (GP) regression, to produce both strain predictions and a predictive distribution of the accumulated fatigue damage in a given time period. Here, the fatigue distribution is achieved via propagation of successive draws from the posterior GP through a rainflow count. The establishment of such a distribution crucially accounts for uncertainty in the predictive model and will form a valuable element in any probabilistic risk assessment. For demonstration of the method, distributions for predicted fatigue damage in an aircraft wing are produced across 84 flights. The distributions provide a robust measure of predicted damage accumulation and model uncertainty.

show abstract

“…Although presented only marginally (within selected 10‐min intervals), Noppe et al 8 investigated the thrust load time‐series via ANN based surrogates, utilizing both high‐fidelity aeroelastic simulations and high frequency (1Hz) SCADA from field experiments earlier. Avendaño‐Valencia et al 9 proposed a surrogate model based on Gaussian Process Regression with Bayesian hyperparameters calibration to estimate short‐term fatigue Damage Equivalent Loads output of the higher fidelity aero‐servo‐elastic simulations. Lastly, Tarpø et al 10 implemented supervised principal component analysis (PCA) to estimate strain response on an offshore turbine and concluded that the developed virtual sensors had higher reliability and robustness compared to the actual strain gauges for that monitoring campaign.…”

Section: Introductionmentioning

confidence: 99%

Virtual sensors for wind turbines with machine learning‐based time series models

Dimitrov

Göçmen

2022

Wind Energy

View full text Add to dashboard Cite

Modern wind turbines have multiple sensors installed and provide constant data stream outputs; however, there are some important quantities where installing physical sensors is either impractical or the sensor technology is not sufficiently advanced.An example of such a problem is, for example, sensing the shape and location of wake-induced wind deficits caused by upwind turbines-a feature which would have relevant application in wind farm control; however, it is hard to detect physically due to the need of scanning the airflow in front of the turbine in multiple locations. Another control-related example is monitoring and predicting the blade tip-tower clearance. A "virtual sensor" can be created instead, by establishing a mathematical relationship between the quantity of interest and other, measurable quantities such as readings from already available sensors (e.g., SCADA, lidars, and met-masts).Machine Learning (ML) approaches are suitable for this task as ML algorithms are capable of learning and representing complex relationships. This study details the concept of ML-based virtual sensors and showcases three specific examples: blade root bending moment prediction, detection of wind turbine wake center location, and forecasting of blade tip-tower clearance. All examples utilize sequence models (Long Short-Term Memory, LSTM) and use aeroelastic load simulations to generate wind turbine response time series and test model performance. The data types used in the examples correspond to channels that would be available from high-frequency SCADA data combined with a blade and tower load measurement system. The resulting model performance demonstrates the feasibility of the ML-based virtual sensor approach.

show abstract

Virtual fatigue diagnostics of wake-affected wind turbine via Gaussian Process Regression

Cited by 48 publications

References 18 publications

Data-driven farm-wide fatigue estimation on jacket-foundation OWTs for multiple SHM setups

Data-driven farm-wide fatigue estimation on jacket-foundation OWTs for multiple SHM setups

Distributions of fatigue damage from data-driven strain prediction using Gaussian process regression

Virtual sensors for wind turbines with machine learning‐based time series models

Contact Info

Product

Resources

About