Figure 1 - available via license: Creative Commons Attribution 3.0 Unported
Content may be subject to copyright.
Source publication
A multi-level hyperparameter optimization framework is performed to calibrate analytical wake models in the context of multiple wind farms within the Belgian-Dutch offshore cluster. The calibration, applied on the TurbOPark model with Gaussian wake profile, is performed on different scales. Initially, calibration focused solely on internal wake eff...
Contexts in source publication
Context 1
... this study, the calibration of the wake model is conducted using turbines located in the Belgian offshore zone, as illustrated in Figure 1, while computation of the wake model encompasses all wind turbines in both the Belgian and Dutch offshore zones, as shown in Figure 2. The Belgian offshore zone is characterized by its high power density, at 9.5MW/km 2 , with a rated capacity of 2.2GW. In contrast, the Dutch offshore zone has a lower power density of 4.8MW/km 2 and a rated capacity of 1.5GW. ...
Context 2
... configuration offers a comprehensive view of wakes in a densely populated wind farm region, allowing for a detailed analysis of both intra-and inter-farm wake effects for one of the world's largest concession zones to date. (i) Internal wake calibration: Initially, the calibration focuses solely on the internal wake effects of each wind farm and is performed for each wind farm depicted in Figure 1. (ii) Internal and external wake calibration: At this stage, the calibration accounts for both the internal wake effects within each wind farm and the external wake effects from neighbouring wind farms, as depicted in Figure 2. ...
Context 3
... Internal and external wake calibration: At this stage, the calibration accounts for both the internal wake effects within each wind farm and the external wake effects from neighbouring wind farms, as depicted in Figure 2. This calibration is performed on the wind farms as shown in Figure 1. ...
Context 4
... (iii) Cluster wake calibration: In this phase, calibration takes into account both internal and external wake effects, similar to the previous step. Additionally, the cost function is modified to encompass all wind farms from Figure 1 within one cost function, instead of having separate cost functions per wind farm. (iv) Cluster wake calibration including blockage: This calibration considers internal and external wake effects, along with a blockage model. ...
Context 5
... performance of the wake model calibration in a cluster wake scenario is evaluated by examining data from north-west wind directions, specifically between 300 and 330 degrees and wind speeds between 7.0 and 9.0 m/s. The results are depicted in Figures 10 and 11. 10 shows the relative error between the calibrated wake model and SCADA data, while Figure 11 presents a scatter plot comparing active power from SCADA with the wake model predicted power. ...
Context 6
... performance of the wake model calibration in a cluster wake scenario is evaluated by examining data from north-west wind directions, specifically between 300 and 330 degrees and wind speeds between 7.0 and 9.0 m/s. The results are depicted in Figures 10 and 11. 10 shows the relative error between the calibrated wake model and SCADA data, while Figure 11 presents a scatter plot comparing active power from SCADA with the wake model predicted power. Here, the predictions from the wake model are based on the tuning parameter, A, that have been individually calibrated for each considered timestamp on the entire cluster. ...
Context 7
... quantifying inflow heterogeneity are visualized in Figures 12 and 13. Figure 12 reveals a consistent trend for the turbines operating in the freeflow, with turbines parallel to the wind direction showing a pattern where the calibration framework underestimates active power from north-west and overestimates the active power from the south-east. ...
Context 8
... quantifying inflow heterogeneity are visualized in Figures 12 and 13. Figure 12 reveals a consistent trend for the turbines operating in the freeflow, with turbines parallel to the wind direction showing a pattern where the calibration framework underestimates active power from north-west and overestimates the active power from the south-east. This suggests that the wind farm experiences significant and consistent heterogeneous inflow. ...
Context 9
... suggests that the wind farm experiences significant and consistent heterogeneous inflow. This is further supported by Figure 13, which shows that under homogeneous inflow conditions the model fails to capture these inflow characteristics, with an R 2 value equal to 0.40, compared to 0.90 in the cluster wake case. The increasing overestimation between the model and SCADA data from south-east to north-west indicates that this discrepancy is not solely due to blockage. ...
Similar publications
The global rise in offshore wind farms underscores the need to cut costs and optimise energy production. As turbines increase in size and wind farms become more concentrated, mitigating downstream wake effects is crucial for operational efficiency. LiDAR technology, offering advantages like eliminating the need for meteorology masts, has been exten...
This work presents a robust methodology for calibrating analytical wake models, as demonstrated on the velocity deficit parameters of the Gauss–curl hybrid model using 4 years of time series supervisory control and data acquisition (SCADA) data from an offshore wind farm, with a tree-structured Parzen estimator employed as a sampler. Initially, a s...
Citations
... All models are optimized for a constant turbulence intensity of 0.06 and a shear exponent of 0.12. [13,15], is performed on all wind farms depicted in Figure 1. The analysis within this paper focuses on wind directions ranging between 300 and 330 degrees and wind speeds between 7.0 and 9.0 m/s. ...
... The analysis within this paper focuses on wind directions ranging between 300 and 330 degrees and wind speeds between 7.0 and 9.0 m/s. As noted in van Binsbergen et al. (2024) [15], this particular wind direction tends to experience a limited amount of heterogeneous inflow and has demonstrated promising results with the Gaussian TurbOPark model by [9]. The calibration procedure is performed on one year of SCADA (Supervisory Control and Data Acquisition) data. ...
... and Ω ∈Ω [13,15] can be consulted. ...
This study benchmarks the performance of multiple analytical wake models using a multi-level hyperparameter optimization framework for calibrating models with SCADA data in the Belgian-Dutch offshore zone. The calibration targets wind coming the northwest (300 to 330 degrees) with wind speeds ranging from 7 to 9 m/s, a wind direction where wakes are clustered across Belgian wind farms. Six wake models are evaluated, namely those described by Jensen (1983), Bastankhah (2014), Niayifar (2016), Zong (2020), Nygaard (2020), and Pedersen (2022). Relative and accumulated relative error between the calibrated wake models and SCADA data are analyzed both on turbine, farm, and cluster level. Statistical moments of the residual error between each model and SCADA data for individual turbines are presented in boxplots for each wind farm and are analyzed using kernel density estimates. Additionally, the calibrated tuning parameters are used to calculate wake losses, demonstrating a strong overlap across models. The analysis reveals that the top-hat TurbOpark model shows the best performance, followed by its Gaussian variant. The Gaussian models by Niayifar (2016) and Zong (2020), as well as the top-hat model by Jensen (1983) all show relatively good performance, while the Gaussian model by Bastankhah (2014) has the worst performance after calibration. While some models perform better than other models, all models show similar trends post-calibration, indicating that with proper calibration, any model can be viable, given its inherent limitations are recognized and managed.
This paper introduces a novel model for predicting wind turbine power output within a wind farm at a high temporal resolution of 30 seconds. The wind farm is represented as a graph, with Graph Neural Networks (GNNs) used to aggregate selected input features from neighboring turbines. A temporal component is added by feeding a timeseries of input features into the graph and utilizing a hybrid GNN-LSTM model architecture. This approach sequentially extracts temporal features with the LSTM component and spatial features with the GNN component. Our model is integrated into a Normal Behavior Model (NBM) framework for analyzing power loss events in wind farms. The results show that both the Spatial and Spatio-Temporal GNN models outperform traditional data-driven power curve methods, with the Spatio-Temporal GNN demonstrating superior performance due to its ability to capture both spatial and temporal dynamics. Additionally, we illustrate the model’s effectiveness in detecting and analyzing instances of reduced performance and its ability to identify various types of abnormal events beyond what is recorded in standard status logs.
