Decreasing Wind Speed Extrapolation Error via Domain-Specific Feature Extraction and Selection

Vassallo, Daniel; Krishnamurthy, Raghavendra; Fernando, Harindra J. S.

doi:10.5194/wes-2019-58

Cited by 5 publications

(7 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The input features used for the wind speed extrapolation are listed in Table 2. As wind speeds often show a diurnal cycle in response to atmospheric stability (Barthelmie et al, 1996;Zhang and Zheng, 2004), we have included multiple variables to capture the diurnal variability in the atmospheric boundary layer: Obukhov length, TKE, and time of day. To preserve the cyclical nature of time of day (i.e., hour 23 and hour 0 being close to each other), we calculate the sine and cosine 1 of the normalized time of day and include these two input features to represent time in the learning algorithm.…”

Section: Random Forestmentioning

confidence: 99%

“…using lidar measurements at a flat terrain site in Saudi Arabia. Finally, Vassallo et al (2019) tested the performance of deep neural networks in extrapolating wind speed as a function of different input features, both in complex terrain and offshore, using lidar data. In all cases, the machine-learning models are compared against traditional extrapolation techniques like the power or logarithmic law, and considerable improvements in extrapolation accuracy using machine-learning techniques have generally been found.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

The importance of round-robin validation when assessing machine-learning-based vertical extrapolation of wind speeds

Bodini

Optis

2020

Wind Energ. Sci.

View full text Add to dashboard Cite

Abstract. The extrapolation of wind speeds measured at a meteorological mast to wind turbine rotor heights is a key component in a bankable wind farm energy assessment and a significant source of uncertainty. Industry-standard methods for extrapolation include the power-law and logarithmic profiles. The emergence of machine-learning applications in wind energy has led to several studies demonstrating substantial improvements in vertical extrapolation accuracy in machine-learning methods over these conventional power-law and logarithmic profile methods. In all cases, these studies assess relative model performance at a measurement site where, critically, the machine-learning algorithm requires knowledge of the rotor-height wind speeds in order to train the model. This prior knowledge provides fundamental advantages to the site-specific machine-learning model over the power-law and log profiles, which, by contrast, are not highly tuned to rotor-height measurements but rather can generalize to any site. Furthermore, there is no practical benefit in applying a machine-learning model at a site where winds at the heights relevant for wind energy production are known; rather, its performance at nearby locations (i.e., across a wind farm site) without rotor-height measurements is of most practical interest. To more fairly and practically compare machine-learning-based extrapolation to standard approaches, we implemented a round-robin extrapolation model comparison, in which a random-forest machine-learning model is trained and evaluated at different sites and then compared against the power-law and logarithmic profiles. We consider 20 months of lidar and sonic anemometer data collected at four sites between 50 and 100 km apart in the central United States. We find that the random forest outperforms the standard extrapolation approaches, especially when incorporating surface measurements as inputs to include the influence of atmospheric stability. When compared at a single site (the traditional comparison approach), the machine-learning improvement in mean absolute error was 28 % and 23 % over the power-law and logarithmic profiles, respectively. Using the round-robin approach proposed here, this improvement drops to 20 % and 14 %, respectively. These latter values better represent practical model performance, and we conclude that round-robin validation should be the standard for machine-learning-based wind speed extrapolation methods.

show abstract

Section: Random Forestmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

The importance of round-robin validation when assessing machine-learning-based vertical extrapolation of wind speeds

Bodini

Optis

2020

Wind Energ. Sci.

View full text Add to dashboard Cite

show abstract

“…Tests were performed with different input features and different heights at various site locations to confirm that bias from a given site and/or measurement height was removed. The Keras library, built on TensorFlow, was utilized to construct the ANN model (Abadi et al, 2016;Chollet et al, 2015).…”

Section: Neural Network Architecturementioning

confidence: 99%

Decreasing wind speed extrapolation error via domain-specific feature extraction and selection

2020

Self Cite

View full text Add to dashboard Cite

Abstract. Model uncertainty is a significant challenge in the wind energy industry and can lead to mischaracterization of millions of dollars' worth of wind resources. Machine learning methods, notably deep artificial neural networks (ANNs), are capable of modeling turbulent and chaotic systems and offer a promising tool to produce high-accuracy wind speed forecasts and extrapolations. This paper uses data collected by profiling Doppler lidars over three field campaigns to investigate the efficacy of using ANNs for wind speed vertical extrapolation in a variety of terrains, and it quantifies the role of domain knowledge in ANN extrapolation accuracy. A series of 11 meteorological parameters (features) are used as ANN inputs, and the resulting output accuracy is compared with that of both standard log-law and power-law extrapolations. It is found that extracted nondimensional inputs, namely turbulence intensity, current wind speed, and previous wind speed, are the features that most reliably improve the ANN's accuracy, providing up to a 65 % and 52 % increase in extrapolation accuracy over log-law and power-law predictions, respectively. The volume of input data is also deemed important for achieving robust results. One test case is analyzed in depth using dimensional and nondimensional features, showing that the feature nondimensionalization drastically improves network accuracy and robustness for sparsely sampled atmospheric cases.

show abstract

“…An alternative, useful approach would be to consider off‐the‐shelves forecast data such as ECMWF‐HRES, as pointed out by B, which could be used for additional validation of the results. We agree that the functional form of extrapolation, that is, the power law, has been subject to increased scrutiny and alternative modeling strategies have been currently sought to decrease the extrapolation error. One of the most recent work on the topic (Vassallo, Krishnamurthy, & Fernando, 2020) successfully applied a feedforward neural network, providing up to 52% improvement in extrapolation over the power law. The use of such an approach is however here problematic as (i) no observational data at high altitude were available at the time of this work and (ii) even assuming such data were available, more than 80,000 neural networks should be trained for each location, unless a full domain convolutional neural network approach would be deployed. In our siting work (Giani et al, 2020) we constrained the search of sites to locations without a complex terrain precisely to avoid these effects.…”

Section: Geoscience Engineering and Wind Energy Aspectsmentioning

confidence: 99%

Rejoinder to the discussion on A high‐resolution bilevel skew‐t stochastic generator for assessing Saudi Arabia's wind energy resources

et al. 2020

View full text Add to dashboard Cite

We would like to thank all discussants for their discussions on different aspects of our paper. We were particularly pleased to see comments across the entire spectrum of our work, from theoretical properties of the model to geoscience, engineering and wind energy aspects that could provide a pathway to implement a wind energy portfolio in Saudi Arabia, truly reflecting the mission of Environmetrics. Throughout this rejoinder, we denote the discussion of Adelchi Azzalini as A, Sándor Baran as B, Emilio Porcu, Jonas Rysgaard and Valerie Eveloy as PRE and Andew Zammit-Magion as Z. This rejoinder covers the different points raised by the discussants in different sections. Section 2 discusses theoretical and methodological aspects of our model. Section 3 focuses on inference and also shows new results with a Bayesian implementation with Stan (Carpenter et al., 2017). Section 4 replies to discussion points on validation. Finally, Section 5 covers aspects related to geoscience, engineering and wind energy stemming from this work.

show abstract

Decreasing Wind Speed Extrapolation Error via Domain-Specific Feature Extraction and Selection

Cited by 5 publications

References 30 publications

The importance of round-robin validation when assessing machine-learning-based vertical extrapolation of wind speeds

The importance of round-robin validation when assessing machine-learning-based vertical extrapolation of wind speeds

Decreasing wind speed extrapolation error via domain-specific feature extraction and selection

Rejoinder to the discussion on A high‐resolution bilevel skew‐t stochastic generator for assessing Saudi Arabia's wind energy resources

Contact Info

Product

Resources

About