Subscribe: Wind Energy
Added By: Feedage Forager Feedage Grade B rated
Language: English
bearing  blade  design  load  loads  power  system  turbine  turbines  wake  wind power  wind turbine  wind turbines  wind 
Rate this Feed
Rate this feedRate this feedRate this feedRate this feedRate this feed
Rate this feed 1 starRate this feed 2 starRate this feed 3 starRate this feed 4 starRate this feed 5 star

Comments (0)

Feed Details and Statistics Feed Statistics
Preview: Wind Energy

Wind Energy

Wiley Online Library : Wind Energy

Published: 2017-12-01T00:00:00-05:00


Using wind velocity estimated from a reanalysis to minimize the variability of aggregated wind farm production over Europe


In this work we use the mean-variance portfolio optimization, using as input the power derived from the wind and density results of meteorological model simulations (ERA-Interim reanalysis), to minimize the variability of the wind power produced in a large region. The methodology involves selecting the placement of the wind farms on a high spatial resolution grid. We used the EU-28 region to check the method and perform sensitivity tests. We studied the influence of the ratio between the total installed power of the whole domain (Pt) and the maximum power that can be installed per cell (Pmi) on the variability of wind power yield. The results show that the reliability of the electrical system improves when Pmi grows and worsens when Pt grows. A quadratic fit relates the variability of the system and the aforementioned ratio. The optimization procedure tends to select groups of terrain cells where wind farms should be installed. These groups grow when more energy production is demanded of the system, but they roughly maintain their location. There is some evidence that in a larger region greater system reliability could be achieved. Most of the selected cells have either a high or a low capacity factor and those with the latter are crucial in enhancing system reliability.

Large-eddy simulations of wind-farm wake characteristics associated with a low-level jet


In this study, we performed a suite of flow simulations for a 12-wind-turbine array with varying inflow conditions and lateral spacings, and compared the impacts of the flow on velocity deficit and wake recovery. We imposed both laminar inflow and turbulent inflows, which contain turbulence for the Ekman layer and a low-level jet (LLJ) in the stable boundary layer. To solve the flow through the wind turbines and their wakes, we used a large-eddy simulation technique with an actuator-line method. We compared the time series for the velocity deficit at the first and rear columns to observe the temporal change in velocity deficit for the entire wind farm. The velocity deficit at the first column for LLJ inflow was similar to that for laminar inflow. However, the magnitude of velocity deficit at the rear columns for the case with LLJ inflow was 11.9% greater because of strong wake recovery, which was enhanced by the vertical flux of kinetic energy associated with the LLJ. To observe the spatial transition and characteristics of wake recovery, we performed statistical analyses of the velocity at different locations for both the laminar and LLJ inflows. These studies indicated that strong wake recovery was present, and a kurtosis analysis showed that the probability density function for the streamwise velocity followed a Gaussian distribution. In a quadrant analysis of the Reynolds stress, we found that the ejection and sweep motions for the LLJ inflow case were greater than those for the laminar inflow case.

Flexure pitch bearing concept for individual pitch control of wind turbines


The excessive use of individual pitch control (IPC) for fatigue load reduction is accompanied by the uncertainty of potential bearing failures. This problem, which is due to the small swivel angles associated with IPC, arises because of the rolling and sliding contacts that occur with the rolling element bearings that are typically used. The use of a flexure bearing is proposed as a way of bypassing this issue. The flexure bearing enables a certain range of motion to be exclusively provided by elastic deformation. This article presents a novel bearing concept that is based on the hypothesis that such a flexure bearing can handle the unfavorable load conditions associated with IPC better than a rolling element bearing. Methods for the dimensioning of the aforementioned flexure bearing are therefore presented. The loads, particularly the required elastic rotation angle of the flexure bearing, are determined first. A promising design for the flexure bearing itself is then chosen and adapted to meet the specific requirements of IPC. These methods are applied to develop an initial conceptual design of the novel bearing unit for a 3-bladed wind turbine of about 3.6 MW. The result demonstrates the feasibility of the concept, and a final discussion presents further opportunities of the design that will make this concept satisfy the special requirements of IPC.

A fast stochastic solution method for the Blade Element Momentum equations for long-term load assessment


Unsteady power output and long-term loads (extreme and fatigue) drive wind turbine design. However, these loads are difficult to include in optimization loops and are typically only assessed in a post-optimization load analysis or via reduced-order methods. Both alternatives yield suboptimal results. The reason for this difficulty lays in the deterministic approaches to long-term loads assessment. To model the statistics of lifetime loads they require the analysis of many unsteady load cases, generated from many different random seeds—a computationally expensive procedure. In this paper, we present an alternative: a stochastic solution for the unsteady aerodynamic loads based on a projection of the unsteady Blade Element Momentum (BEM) equations onto a stochastic space spanned by chaos exponentials. This approach is similar to the increasingly popular polynomial chaos expansion, but with 2 major differences. First, the BEM equations constitute a random process, varying in time, while previous polynomial chaos expansion methods were concerned with random parameters (ie, random but constant in time or initial values). Second, a new, more efficient basis (the exponential chaos) is used. This new stochastic method enables us to obtain unsteady long-term loads much faster, enabling unsteady loads to become accessible inside wind turbine optimization loops. In this paper we derive the stochastic BEM solution and present the most relevant results showing the accuracy of the new method.

Investigation of high-speed shaft bearing loads in wind turbine gearboxes through dynamometer testing


Many wind turbine gearboxes require repair or replacement well before reaching the end of their design life. The most common failure is bearing axial cracks, commonly called white etching cracks (WECs), which typically occur in the inner raceways of the high-speed parallel-stage rolling element bearings. Although the root causes of WECs are debated, one theory is that they are related to routine dynamic operating conditions and occasional transient events prevalent in wind turbines that can result in high bearing stress and sliding of the rolling elements. This paper examined wind turbine gearbox high-speed shaft bearing loads and stresses through modeling and full-scale dynamometer testing. Bearing outer race loads were directly measured and predicted using a variety of modeling tools in normal operations, misaligned conditions, and transient events particularly prone to bearing sliding. Test data and models of bearing loads were well correlated. Neither operational misalignment due to rotor moments nor static generator misalignment affected the bearing loads when compared with pure-torque conditions. Thus, it is not likely that generator misalignment is a causal factor of WECs. In contrast, during transient events, the bearings experienced alternating periods of high stress, torque reversals, and loads under the minimum requisite at high rotating speeds while showing indications of sliding, all of which could be related to the formation of WECs.

Measurement of spectra over the Bolund hill in wind tunnel


We have determined the normal Reynolds stresses and spectra of the wind velocity over a 1:115 scale mock-up of the Bolund hill. The experiment was run in a neutral boundary layer wind tunnel using 3-component hot-wire velocimetry, 2-component particle image velocimetry, and a high-precision traversing system. Spectra have been determined at different points along transects at 2 and 5 m height above ground level. The experiment was run for 270° wind direction and for two Reynolds numbers, Reh1=4.25×104 and Reh2=8.21×104, based on the maximum height of the hill and the free wind speed at this height. Our results show how the normalized power spectral density Siiσ=fSii/ui2‾ changes over the hill. The analysis of the normalized streamwise spectrum at 2 m height, just after the escarpment, reveals that part of the energy is concentrated in the interval of normalized frequencies nh≈0.01−0.02, which could be a signature of a weakened “flapping” phenomenon described in the literature for flows over forward facing steps. The departure of the spectra slope in the inertial subrange, from the value −5/3, was found to be correlated with the hill geometry.

A CFD study of coupled aerodynamic-hydrodynamic loads on a semisubmersible floating offshore wind turbine


The prediction of dynamic characteristics for a floating offshore wind turbine (FOWT) is challenging because of the complex load coupling of aerodynamics, hydrodynamics, and structural dynamics. These loads should be accurately calculated to yield reliable analysis results in the design phase of a FOWT. In this study, a high-fidelity fluid-structure interaction simulation that simultaneously considers the influence of aero-hydrodynamic coupling due to the dynamic motion of a FOWT has been conducted using computational fluid dynamics based on an overset grid technique. The DeepCwind semisubmersible floating platform with the NREL 5-MW baseline wind turbine model is considered for objective numerical verification with the NREL FAST code. A state-of-the-art computational model based on the coupled computational fluid dynamics and dynamic structure analysis is constructed and analyzed to solve multiphase flow, 6 degrees of freedom motions of OC4 semisubmersible FOWT. A quasi-static mooring solver is also applied to resolve the constraint motion of floater because of a 3-line mooring system. The influence of tower shadow on the unsteady aerodynamic performance and loads is also demonstrated. Finally, complex unsteady flow fields considering blade and tower interference effects among blade-tip vortices, shedding vortices, and turbulent wakes are numerically visualized and investigated in detail.

An experimental investigation on the wake interferences among wind turbines sited in aligned and staggered wind farms


An experimental investigation was conducted for a better understanding of the wake interferences among wind turbines sited in wind farms with different turbine layout designs. Two different types of inflows were generated in an atmospheric boundary layer wind tunnel to simulate the different incoming surface winds over typical onshore and offshore wind farms. In addition to quantifying the power outputs and dynamic wind loads acting on the model turbines, the characteristics of the wake flows inside the wind farms were also examined quantitatively. After adding turbines staggered between the first 2 rows of an aligned wind farm to increase the turbine number density in the wind farm, the added staggered turbines did not show a significant effect on the aeromechanical performance of the downstream turbines for the offshore case. However, for the onshore case, while the upstream staggered turbines have a beneficial effect on the power outputs of the downstream turbines, the fatigue loads acting on the downstream turbines were also found to increase considerably due to the wake effects induced by the upstream turbines. With the same turbine number density and same inflow characteristics, the wind turbines were found to be able to generate much more power when they are arranged in a staggered layout than those in an aligned layout. In addition, the characteristics of the dynamic wind loads acting on the wind turbines sited in the aligned layout, including the fluctuation amplitudes and power spectrum, were found to be significantly different from those with staggered layout.

Deep learning for automated drivetrain fault detection


A novel data-driven deep-learning system for large-scale wind turbine drivetrain monitoring applications is presented. It uses convolutional neural network processing on complex vibration signal inputs. The system is demonstrated to learn successfully from the actions of human diagnostic experts and provide early and robust fault detection on both rotor bearing, planetary and helical stage gear box bearings from analysis of multisensor vibration patterns using only a high-level feature selection. On the basis of data from 251 actual wind turbine bearing failures, we are able to accurately quantify the fleet-wide diagnostic model performance. The analysis also explores the time dependence of the diagnostic performance, providing a detailed view of the timeliness and accuracy of the diagnostic outputs across the different architectures. Deep architectures are shown to outperform the human analyst as well as shallow-learning architectures, and the results demonstrate that when applied in a large-scale monitoring system, machine intelligence is now able to handle some of the most challenging diagnostic tasks related to wind turbines.

Detection of faulty high speed wind turbine bearing using signal intensity estimator technique


Bearings are typically used in wind turbines to support shafts and gears that increase rotational speed from a low speed rotor to a higher speed electrical generator. For various bearing applications, condition monitoring using vibration measurements has remained a subject of intense study to the present day since several decades. Various signal processing techniques are used to analyse vibration signals and extract features related to defects. Statistical indicators such as Crest Factor (CF) and Kurtosis (KU) were reported as very sensitive indicators when the presence of the defects is pronounced, whilst their values may come down to the level of undamaged components when the damage is well advanced. Further, these indicators were applied to an acquired data from proposed diagnostic models, test rigs, and instrumentations that were specifically used for particular research tests, and thus, it is essential to undertake further investigations and analysis to assess the influence of other factors such as the structural noise and other operating conditions on the real-world applications. With this in mind, the present work proposes Signal Intensity Estimator (SIE) as a new technique to discriminate individual types of early natural damage in real-world wind turbine bearings. Comparative results between SIE and conventional indicators such as KU and CF are also presented. It was concluded that SIE has an advantage over the other fault indicators if sufficient data are provided.

A comparison of methods for assessing power output in non-uniform onshore wind farms


Wind resource assessments are used to estimate a wind farm's power production during the planning process. It is important that these estimates are accurate, as they can impact financing agreements, transmission planning, and environmental targets. Here, we analyze the challenges in wind power estimation for onshore farms. Turbine wake effects are a strong determinant of farm power production. With given input wind conditions, wake losses typically cause downstream turbines to produce significantly less power than upstream turbines. These losses have been modeled extensively and are well understood under certain conditions. Most notably, validation of different model types has favored offshore farms. Models that capture the dynamics of offshore wind conditions do not necessarily perform equally as well for onshore wind farms. We analyze the capabilities of several different methods for estimating wind farm power production in 2 onshore farms with non-uniform layouts. We compare the Jensen model to a number of statistical models, to meteorological downscaling techniques, and to using no model at all. We show that the complexities of some onshore farms result in wind conditions that are not accurately modeled by the Jensen wake decay techniques and that statistical methods have some strong advantages in practice.

Bend-bend-twist vibrations of a wind turbine blade


Dynamics of a wind turbine blade under bend-bend-twist coupled vibrations is investigated. The potential and kinetic energy expressions for a straight nonuniform blade are written in terms of beam parameters. Then, the energies are expressed in terms of modal coordinates by using the assumed mode method, and the equations of motion are found by applying Lagrange's formula. The bend-bend-twist equations are coupled with each other and have stiffness variations due to centrifugal effects and gravitational parametric terms, which vary cyclicly with the hub angle. To determine the natural frequencies and mode shapes of the system, a modal analysis is applied on the linearized coupled equations of constant angle snapshots of a blade with effects of constant speed rotation. Lower modes of the coupled bend-bend-twist model are dominantly in-plane or out-of-plane modes. To investigate the parametric effects, several blade models are analyzed at different angular positions. The stiffness terms involving centrifugal and gravitational effects can be significant for long blades. To further see the effect of blade length on relative parametric stiffness change, the blade models are scaled in size and analyzed at constant rotational speeds, at horizontal and vertical orientations. These studies show that the parametric stiffness effects should be taken into account when designing long blades.

Teeter design for lowest extreme loads during end impacts


Two bladed wind turbines are discussed as a possible turbine alternative for offshore use as they show a potential to save cost of energy. But compared to three-bladed turbines, their dynamic behavior is much more challenging. A possible solution to handle these larger dynamic loads is the use of a teeter hinge, which can significantly reduce fatigue loads. In contrast to that, extreme loads, coming from teeter end impacts, are often described as a problem for teetered turbines. There are different design parameters of the teeter system of a turbine, which have an influence on extreme loads during teeter end impacts. Despite numerous studies on teeter movement and load reduction potentials of operational loads, scientific literature does not give information about suitable load-reducing combinations of teeter design parameters and their influence on extreme loads. This paper, which is a summary of a PhD thesis, analyses which combination of teeter parameters has the largest load-reducing influence on extreme loads. Aeroelastic load simulations of the teetered turbine CART2 from the NREL test site and one of today's commercial two-bladed turbines, the SCD3MW from aerodyn (both pitch controlled upwind turbines), will be used.

Cover Image


The cover image, by Christian Santoni et al., is based on the Research Article Effect of tower and nacelle on the flow past a wind turbine, DOI: 10.1002/we.2130.

Issue Information


No abstract is available for this article.

Generating short-term probabilistic wind power scenarios via nonparametric forecast error density estimators


Forecasts of available wind power are critical in key electric power systems operations planning problems, including economic dispatch and unit commitment. Such forecasts are necessarily uncertain, limiting the reliability and cost-effectiveness of operations planning models based on a single deterministic or “point” forecast. A common approach to address this limitation involves the use of a number of probabilistic scenarios, each specifying a possible trajectory of wind power production, with associated probability. We present and analyze a novel method for generating probabilistic wind power scenarios, leveraging available historical information in the form of forecasted and corresponding observed wind power time series. We estimate nonparametric forecast error densities, specifically using epi-spline basis functions, allowing us to capture the skewed and nonparametric nature of error densities observed in real-world data. We then describe a method to generate probabilistic scenarios from these basis functions that allows users to control for the degree to which extreme errors are captured. We compare the performance of our approach to the current state-of-the-art considering publicly available data associated with the Bonneville Power Administration, analyzing aggregate production of a number of wind farms over a large geographic region. Finally, we discuss the advantages of our approach in the context of specific power systems operations planning problems: stochastic unit commitment and economic dispatch. Our methodology is embodied in the joint Sandia–University of California Davis Prescient software package for assessing and analyzing stochastic operations strategies.

Effect of tower and nacelle on the flow past a wind turbine


Large eddy simulations (LES) of the flow past a wind turbine with and without tower and nacelle have been performed at 2 tip speed ratios (TSR, λ=ωR/U∞), λ=3 and 6, where the latter corresponds to design conditions. The turbine model is placed in a virtual wind tunnel to reproduce the “Blind test 1” experiment performed at the Norwegian University of Science and Technology (NTNU) closed-loop wind tunnel. The wind turbine was modeled using the actuator line model for the rotor blades and the immersed boundary method for the tower and nacelle. The aim of the paper is to highlight the impact of tower and nacelle on the turbine wake. Therefore, a second set of simulations with the rotating blades only (neglecting the tower and nacelle) has been performed as reference. Present results are compared with the experimental measurements made at NTNU and numerical simulations available in the literature. The tower and nacelle not only produce a velocity deficit in the wake but they also affect the turbulent kinetic energy and the fluxes. The wake of the tower interacts with that generated by the turbine blades promoting the breakdown of the tip vortex and increasing the mean kinetic energy flux into the wake. When tower and nacelle are modeled in the numerical simulations, results improve significantly both in the near wake and in the far wake.

Aeroelastic multidisciplinary design optimization of a swept wind turbine blade


Mitigating loads on a wind turbine rotor can reduce the cost of energy. Sweeping blades produces a structural coupling between flapwise bending and torsion, which can be used for load alleviation purposes. A multidisciplinary design optimization (MDO) problem is formulated including the blade sweep as a design variable. A multifidelity approach is used to confront the crucial effects of structural coupling on the estimation of the loads. During the MDO, ultimate and damage equivalent loads are estimated using steady-state and frequency-domain–based models, respectively. The final designs are verified against time-domain full design load basis aeroelastic simulations to ensure that they comply with the constraints. A 10-MW wind turbine blade is optimized by minimizing a cost function that includes mass and blade root flapwise fatigue loading. The design space is subjected to constraints that represent all the necessary requirements for standard design of wind turbines. Simultaneous aerodynamic and structural optimization is performed with and without sweep as a design variable. When sweep is included in the MDO process, further minimization of the cost function can be obtained. To show this achievement, a set of optimized straight blade designs is compared to a set of optimized swept blade designs. Relative to the respective optimized straight designs, the blade mass of the swept blades is reduced of an extra 2% to 3% and the blade root flapwise fatigue damage equivalent load by a further 8%.

Individual pitch control for 2-bladed wind turbines via multiblade multilag transformation


Two-bladed wind turbines have regained the attention of the community thanks to the advantages in manufacturing cost provided by the lower number of blades and the ease of implementation of effective passive systems for load reduction (ie, teetering pin). Considering both teetering and nonteetering architectures, the dynamics of 2-bladed turbines is different from that of 3-bladed machines especially in terms of how multiples of the rotor frequencies in blade signals are translated on the fixed system. Such characteristics have hampered the adoption of active control laws for load mitigation based on individual pitch control, extensively studied for 3-bladed turbines. A basic element for control development allowing to capture the essence of the relationship between signals on the blades and the fixed system on 2-bladed turbines–represented by the Coleman transformation for 3-bladed turbines–has not been identified yet. The present paper tries to fill the gap, presenting an extended transformation–called multiblade multilag–applicable to turbines with an arbitrary number of blades, providing a systematic way to link rotor signals to fixed system signals, thus allowing the application of control algorithms for individual pitch control developed for 3-bladed turbines to the 2-bladed case. The paper addresses the problem first at a theoretical level, and subsequently providing applicative results from simulations on virtual models of teetering and nonteetering 2-bladed turbines. The proposed transformation algorithm and control laws allow to effectively reduce some relevant loads and motions respectively in the nonteetering and teetering scenarios, through a cyclic pitch input.

Analysis of reactive power strategies in HVDC-connected wind power plant clusters


Offshore wind power plants (WPPs) built near each other but far from shore usually connect to the main grid by a common high-voltage DC (HVDC) transmission system. In the resulting decoupled offshore grid, the wind turbine converters and the high-voltage DC voltage-source converter share the ability to inject or absorb reactive power. The overall reactive power control dispatch influences the power flows in the grid and hence the associated power losses. This paper evaluates the respective power losses in HVDC-connected WPP clusters when applying 5 different reactive power control strategies. The case study is made for a 1.2-GW–rated cluster comprising 3 WPP and is implemented in a combined load flow and converter loss model. A large set of feasible operating points for the system is analyzed for each strategy. The results show that a selection of simulations with equal wind speeds is sufficient for the annual energy production comparison. It is found that the continuous operation of the WPPs with unity power factor has a superior performance with low communication requirements compared with the other conventional strategies. The optimization-based strategy, which is developed in this article, allows a further reduction of losses mainly because of the higher offshore grid voltage level imposed by the high-voltage DC voltage-source converter. Reactive power control in HVDC-connected WPP clusters change significantly the overall power losses of the system, which depend rather on the total sum of the injected active power than on the variance of wind speeds inside the cluster.

Bird detection and species classification with time-lapse images around a wind farm: Dataset construction and evaluation


Collisions of birds, especially endangered species, with wind turbines is a major environmental concern. Automatic bird monitoring can be of aid in resolving the issue, particularly in environmental risk assessments and real-time collision avoidance. For automatic recognition of birds in images, a clean, detailed, and realistic dataset to learn features and classifiers is crucial for any machine-learning-based method. Here, we constructed a bird image dataset that is derived from the actual environment of a wind farm and that is useful for examining realistic challenges in bird recognition in practice. It consists of high-resolution images covering a wide monitoring area around a turbine. The birds captured in these images are at relatively low resolution and are hierarchically labeled by experts for fine-grained species classification. We conducted evaluations of state-of-the-art image recognition methods by using this dataset. The evaluations revealed that a deep-learning-based method and a simpler traditional learning method were almost equally successful at detection, while the former captures more generalized features. The most promising results were provided by the deep-learning-based method in classification. The best methods in our experiments recorded a 0.98 true positive rate for bird detection at a false positive rate of 0.05 and a 0.85 true positive rate for species classification at a false positive rate of 0.1.

Measurement of mechanical loads in large wind turbines: Problems on calibration of strain gage bridges and analysis of uncertainty


The paper discusses the complexity of calibration of strain gage full bridges applied to measure mechanical loads in large wind turbines, when direct application of calibration loads is not feasible. In particular, at first, it presents a generalized static-dynamic mechanical model which allows to calibrate the strain gage full bridges using its own unbalanced masses to generate known reference inputs. Then, the paper discusses the uncertainty associated to such a calibration, according to the ISO/IEC Guide 98-3:2008 “Guide to the Expression of Uncertainty in Measurement”. The uncertainty of the reference input and the following calibration is discussed, which is often larger than the target set by the standard IEC-61400-13 used for wind turbine type certification. The paper comments on the attainable range of calibration which is rather limited with respect to expected load range in operation. Even if calibration should take place in isothermal effects, this is not always the case in real world practice. Therefore, the thermal effects on strain gage bridges are also discussed, putting into evidence its influence on calibration uncertainty both for full bridges in T configuration and in parallel configuration.

A simple model of the wind turbine induction zone derived from numerical simulations


The induction zone in front of different wind turbine rotors is studied by means of steady-state Navier-Stokes simulations combined with an actuator disk approach. It is shown that, for distances beyond 1 rotor radius upstream of the rotors, the induced velocity is self-similar and independent of the rotor geometry. On the basis of these findings, a simple analytical model of the induction zone of wind turbines is proposed.

Tailoring incoming shear and turbulence profiles for lab-scale wind turbines


An active grid is used to generate a variety of turbulent shear profiles in a wind tunnel. The vertical bars are set to flap through varying angles across the test section producing a variation in the perceived solidity, resulting in a mean shear. The horizontal bars are used in a fully random operational mode to set the background turbulence level. It is demonstrated that mean velocity profiles with approximately the same shear can be produced with different turbulence intensities and local turbulent Reynolds numbers based on the Taylor microscale, λ. Conversely, it is also demonstrated that flows can be produced with similar turbulence intensity profiles but different mean shear. It is confirmed that the length scales and dynamics, the latter being assessed through the velocity spectra and probability density functions, do not vary significantly across the investigation domain. Such flows are of particular relevance for studies investigating the effect of in-flow conditions on obstacles where these studies wish to decouple the effects of turbulence intensity and mean shear, a feat previously unattainable in experimental facilities. Given that the power output of wind turbines is known to be a function of both mean shear and turbulence intensity, the experimental methodology presented herein is invaluable to the wind turbine model testing community who, at present, cannot exert such control authority over the in-flow conditions.

Statistical learning for wind power: A modeling and stability study towards forecasting


We focus on wind power modeling using machine learning techniques. We show on real data provided by the wind energy company Maïa Eolis that parametric models even following closely the physical equation relating wind production to wind speed are outperformed by intelligent learning algorithms. In particular, the CART-Bagging algorithm gives very stable and promising results. Besides, as a step towards forecast, we quantify the impact of using deteriorated wind measures on the performances. We show also on this application that the default methodology to select a subset of predictors provided in the standard random forest package can be refined, especially when there exists among the predictors one variable, which has a major impact.

Wind turbine performance decline in Sweden


We show that Swedish wind turbines constructed before 2007 lose 0.15 capacity factor percentage points per year, corresponding to a lifetime energy loss of 6%. A gradual increase of downtime accounts for around one third of the deterioration and worsened efficiency for the remaining. Although the performance loss in Sweden is considerably smaller than previously reported in the UK, it is statistically significant and calls for a revision of the industry practice for wind energy calculations. The study is based on two partly overlapping datasets, comprising 1,100 monthly and 1,300 hourly time series spanning 5–25 years each.