Expert Elicitation on Wind Farm Control
J. W. van Wingerden, P. A. Fleming, T. Göçmen, I. Eguinoa, B. M. Doekemeijer, K. Dykes, M. Lawson, E. Simley, J. King, D. Astrain, M. Iribas, C. L. Bottasso, J. Meyers, S. Raach, K. Kölle, G. Giebel
EExpert Elicitation on Wind Farm Control
J. W. van Wingerden , P. A. Fleming , T. G¨o¸cmen , I. Eguinoa ,B. M. Doekemeijer , K. Dykes , M. Lawson , E. Simley , J. King ,D. Astrain , M. Iribas , C. L. Bottasso , J. Meyers , S. Raach , K.K¨olle , G. Giebel Delft University of Technology, Delft, The Netherlands National Renewable EnergyLaboratory, Golden, Colorado, USA Technical University of Denmark, Roskilde, Denmark CENER, Sarriguren, Spain Technical University Munich, Munich, Germany KU Leuven,Leuven, Belgium sowento, Stuttgart, Germany SINTEF Energy Research, Trondheim,NorwayE-mail: [email protected]
Abstract.
Wind farm control is an active and growing field of research in which thecontrol actions of individual turbines in a farm are coordinated, accounting for inter-turbineaerodynamic interaction, to improve the overall performance of the wind farm and to reducecosts. The primary objectives of wind farm control include increasing power production,reducing turbine loads, and providing electricity grid support services. Additional objectivesinclude improving reliability or reducing external impacts to the environment and communities.In 2019, a European research project (FarmConners) was started with the main goal of providingan overview of the state-of-the-art in wind farm control, identifying consensus of researchfindings, data sets, and best practices, providing a summary of the main research challenges, andestablishing a roadmap on how to address these challenges. Complementary to the FarmConnersproject, an IEA Wind Topical Expert Meeting (TEM) and two rounds of surveys among expertswere performed. From these events we can clearly identify an interest in more public validationcampaigns. Additionally, a deeper understanding of the mechanical loads and the uncertaintiesconcerning the effectiveness of wind farm control are considered two major research gaps.
1. Introduction
Wind farm control (WFC) is an active and growing field of research in which the control actionsof individual turbines in a farm are coordinated, accounting for inter-turbine aerodynamicinteraction, to improve the overall performance of the wind farm and to reduce costs. Theprimary objectives of WFC include increasing power production, reducing turbine loads, andproviding electricity grid support services. Additional objectives include improving reliabilityor reducing external impacts to the environment and communities. In 2019, a European project(FarmConners ) was started with the main goal of providing an overview of the state of the artin WFC, identifying consensus of research findings, data sets, and best practices, providing asummary of the main research challenges, and establishing a roadmap on how to address thesechallenges. Complementary to the FarmConners project, an International Energy Agency (IEA)Wind Topical Expert Meeting (TEM) has similar objectives. In this paper these two initiativesare merged. See https://cordis.europa.eu/project/rcn/224282/factsheet/en and . a r X i v : . [ ee ss . S Y ] J un FC can be categorized accordingly to three distinct technologies used to achieve the primaryobjectives identified above:(i) The first technology is wake steering , where wake interactions are modified by redirecting thewakes in the wind farm. This technique could be used to either increase power productionby steering wakes away from downstream turbines or to reduce asymetric loading introducedby partial wakes.(ii) The second technology is axial induction control , where wake interactions and impactsare modified by derating upstream or uprating downstream turbines. Derating leads to areduction in the structural loads of the derated and downstream turbines while uprating canincrease power production. Both derating and uprating can provide the basis for supportinggrid services such as active power control.(iii) The third technology is wake mixing , where upstream turbines are dynamically uprated anddownrated on short timescales to induce additional wake recovery, minimizing wake lossesfurther downstream.The literature contains an abundance of WFC solutions that leverage one or more of thesetechnologies for one or more of the aforementioned objectives [4, 15, 16]. A popular methodof algorithm validation, much more cost-effective than field experiments, has been the use ofdedicated wind tunnel experiments. The potential of wake steering has been demonstrated onmultiple occasions. For example, Adaramola and Krogstad [1], Schottler et al. [19], and Bartlet al. [2] report gains of up to 12% for wake steering in two-turbine arrays in wind tunnelexperiments. Moreover, Campagnolo et al. [5, 6] and Park et al. [18] report gains of up to 33%for three-turbine arrays through wake steering in their wind tunnels. Bastankhah and Port´e-Agel [3] report gains of up to 17% for a five-turbine array using wake steering in a wind tunnel.Moreover, Campagnolo et al. [5, 6], among others, tested the potential of axial induction controlfor power maximization in their wind tunnel reporting no net gains. The third concept, wakemixing, is a novel technology and thereby has only been validated to a limited degree in theliterature. The concept demonstrated in simulation by Munters et al. [17] was validated byFrederik et al. [12] in a set of wind tunnel experiments. More recently, the helix wake mixingconcept was introduced by van Wingerden et al. [21], where individual pitch control is used totrigger wake mixing [13]. This concept is still to be validated through scaled experiments.A handful of publications have focused on validation of wind farm control algorithms throughfield experiments. Contrary to the findings of Campagnolo et al. [5], Van Der Hoek et al. [20]demonstrate axial induction control in field trials on a commercial wind farm, showing a smallbut positive increase in the power production of the farm compared to baseline operation.Fleming et al. [9] demonstrate wake steering on an offshore two-turbine array with positiveresults. Thereafter, Fleming et al. [10, 11] demonstrate wake steering at an onshore two-turbinearray surrounded by a complex topology. In these experiments, the benefits of wake steeringare common in the data, but some losses are measured, too. Howland et al. [14] demonstratewake steering at an onshore six-turbine array, showing significant gains in power productionfor particular situations (i.e., low wind speed, wake-loss-heavy wind directions, low turbulencelevels), although they report that their net gain over annual operation appears insignificantcompared to baseline operation. Doekemeijer et al. [8] then demonstrate wake steering ina complex onshore wind farm, containing different turbine types and more complicated wakeinteraction. Their results largely agree with the other field experiments: both gains and lossesin power production are measured. The authors conclude that, while the gains outnumber thelosses, more research is necessary on the topic of wake steering. At large, it is clear that moreresearch is necessary before WFC can be fully commercialized.To accelerate the industrialization of WFC, the primal objectives of the EuropeanFarmConners project are to develop an overview of the:i) state of the industry: What are the current capabilities of commercial WFC solutions?(ii) state of the academia: What are the current research findings, including theoretical results,overview of mathematical models and their capabilities, and the findings of high-fidelitycomputer simulations, wind tunnel studies, and field trials?(iii) consensus of research findings and best practices: What topics have consensus and whichquestions remain unanswered in the research? Moreover, a secondary objective is toestablish preferred nomenclature and define best practices for future research.(iv) research needs: Based on existing research and desired end goals, where is research mostneeded?In brief, for objective (i), the current commercial WFC solutions are dominated by steady-state wake steering where yaw misalignment is used to maximize the annual energy production(AEP). For objective (ii), recent articles [7, 4, 15] have reviewed the state of the research.Instead, this paper focuses on objectives (iii) and (iv) involving the consensus of research findings,best practices, and open research questions.The paper is structured as follows. Section 2 describes how information was gathered fromleading academia and industry in the field of WFC. Section 3 presents the findings. The paperis concluded in Section 4.
2. Methodology
To meet the aforementioned objectives, IEA Wind TEM th of September,2019, in Amsterdam with 47 experts from academia and industry to discuss the current statusof WFC. The TEM was preceded by two surveys about the current state of the art and researchneeds. The questionnaires were completed by over 50 WFC stakeholders. Figure 1 illustratesthe affiliation of the participants for both iterations of the questionnaire. Although the inputfrom academia and research institutes seems to have a higher share in the survey results, thereis still considerable participation from the industrial and commercial stakeholders.
0% 5% 10% 15% 20% 25% 30% 35% 40% 45%OtherWind farm operatorConsultantProject developerWind turbine manufacturer/OEMResearch instituteAcademia Round 1 Round 2
Figure 1: Affiliation of the participants for the two rounds of WFC surveysThere is a potential bias because of the high participation (more than 50%) of academiaand research institutes. Also, during the discussions in Amsterdam, it became apparent thathe group, to a large degree, shared research interests, focusing mainly on flow modelingand aerodynamical turbine interaction. Experts working on control algorithms, the electricalcomponents, and grid integration were underrepresented in the meeting. It can be assumedthat the same group that attended the TEM also answered the survey, causing a certain biastoward, for example, power maximization by flow control. The potential effect of the participantbackground is discussed further in Section 3, where applicable.During the TEM, the aforementioned questionnaires, as well as the open outcomes of currentresearch projects in the field, were used as a starting point for discussions during the meeting.Research on WFC has been separated into the aerodynamical topics and the aspects of gridintegration. These discussions revealed that there was no uniform consensus among participantson the definition of “grid services” as mentioned in the questionnaire. While some attendeesinterpreted grid services as “complying to grid codes,” others interpreted them as a “provision ofancillary services” and its related requirements. This should be considered when looking at theresults in Section 3. Accordingly, establishing a clear definition is recommended in this paper,especially considering the expected evolution of grid codes and electrical market services as thepenetration of wind energy in the energy mix increases over the coming years.
3. Results
The outcomes of the two-round survey are presented in this section.Before assessing the technical challenges of WFC, consensus must be reached on its definition.Therefore, the survey preceded with the question “What is wind farm control?” The resultsclearly point toward wake steering and axial induction control for power maximization and loadmitigation (i.e., dealing with flow phenomena in the wind farm using mathematical models andhandling uncertainty). Not surprisingly, because of the surveyed group, grid integration is ratedas less important using this definition.Moreover, one of the most pronounced outcomes of the survey is the common request forincreased collaboration in WFC, mainly in terms of i) access to data, ii) availability of models,and iii) availability of field tests. The main objectives of both the TEM in Amsterdam andthe European FarmConners project address exactly that: more engaged collaboration throughcommon dialogue platforms and open science principles.The survey is structured around three main queries, each elaborated next.
The most important benefits of WFC.
The clear majority of the WFC community seesthe increased energy production as the most important benefit of WFC implementation.Potential alleviation of turbine structural loads and lifetime extension follows as the second,where operations and maintenance (O&M) optimization comes as third in importance.Figure 2 shows that mitigation of the environmental impacts such as noise reduction andbird/bat collision is ranked lowest in terms of the benefits that WFC can potentially provide.
Expected AEP gain.
Figure 3 shows the responses for the question “How much of an increasein energy production is needed to justify implementation?” In this figure, we see thataround 60% of the respondents believe that a gain of 1.0% AEP gain is necessary to justifyimplementation. Looking at different stakeholder groups, however, we found that windfarm operators and turbine manufacturers had a lower threshold than 1.0% while academiatended to have a higher threshold. As part of the maturity process in the field, this findinghighlights the need to put the expected WFC benefits into better context by establishingmore realistic but still commercially appealing predictions of potential improvements.
Plausibility of the current results in the literature.
Figure 4 indicates the plausibility ofthe current benefits reported in the WFC literature by the stakeholders of the technology. .0 2.0 3.0 4.0 5.0 6.0 7.0 8.0RankIncreased Energy ProductionLoad Reduction/RebalancingGrid ServicesExtended Turbine LifeWind Farm DensificationO&M OptimizationNoise ReductionBird/Bat Collision 36 4 2 1 2 1 0 02 25 9 2 4 3 0 15 3 6 8 13 9 1 01 2 15 12 6 5 2 21 9 8 7 4 10 6 01 2 6 12 9 7 5 30 0 0 3 7 5 22 70 1 0 0 0 4 8 31
Figure 2: The most important benefits of WFC as perceived by the survey participants. Rankingon the x-axis refers to 1.0 – most important, 8.0 – least importantIt is seen that 50% of the survey participants find the reported results to be more than 60%believable, where the median of the answers is approximately 70% plausibility.
Thereafter, the survey participants were asked to indicate their agreement to several statementsin five categories from “Strongly agree” to “Strongly disagree.” These agreement classes arethen compiled into percentage distributions, as listed in Table 1. For example, 84% of the par-ticipants agree that the WFC technology will be broadly adopted in the future, and 78% ofrespondents highlight the need for standardization in the validation tools and methods.
Figure 3: Distribution of responses to the question: How much of an increase in energyproduction is needed to justify implementation? The x-axis shows the percentage gain andthe y-axis shows the cumulative percentage of questionnaire participants. uestion Agree Disagree
WFC will be broadly adopted at some point in the future 84% 0%There is a lack of standard reference validation tools and methods tobe openly used for certification and bankability purposes 78% 8%It is worth developing WFC if the only benefit is to AEP 74% 5%Wind farm flow control will be broadly adopted within 10 years 66% 9%There is a lack of reliable tools to evaluate the load impact of certainmodes of operation in WFC (e.g., yaw misalignment) 78% 15%It is worth developing WFC if the only benefit is to reduce the turbineloads 64% 11%Atmospheric condition measurements (e.g., lidars or other equip-ment) should be used as real-time inputs in future WFC applications 65% 13%WFC, as opposed to individual turbine control, is needed to providegrid services 60% 8%The theory of WFC is mature 21% 50%WFC is only applicable to newly designed wind farms 3% 80%Table 1: Response to various statements in five categories from “Strongly agree” to “Stronglydisagree.” This table shows the percentage distribution.Figure 5 shows statements to which most respondents agreed. Figure 6 shows statementsto which most respondents disagreed. Figure 7 shows statements in which no consensus wasfound. It appears that WFC is deemed interesting as long as at least one of the objectives(power maximization, load mitigation, electricity grid services) can be achieved. Moreover, therespondents consider experimental validation a priority over other research topics, and currentexperimental methods are considered insufficient. It is interesting to note that there is noconsensus on who will be primarily developing WFC algorithms—be it turbine manufacturers,
20 40 60 80 100% Believable024681012 N u m b e r o f r e s p o n d a n t s
20 40 60 80 100% Believable0.00%50.00%100.00% P e r c . o f r e s p o n d a n t s a t o r a b o v e Figure 4: Distribution of responses to the question: On a scale of 0–100, how believable doyou consider the reported energy yield gains of WFC methods? 100 – highly believable.
Left: histogram of the responses; right: cumulative distributioncademia, or an external entity.
Figure 5: Statements to which more survey participants agree than disagree
Figure 6: Statements to which more survey participants disagree than agree
An important objective of the FarmConners project is defining implementation barriers andassigning research priorities. The corresponding research results are shown in Figures 8 and 9.
Validation.
The survey participants consider the lack of validation as one of the main barriersto WFC on an industrial scale, as shown in Figure 8. The ranking of research priorities inFigure 9 shows a clear interest in public validation campaigns, which is in agreement withthe findings from Figure 5, among others. Moreover, the survey shows that there is a strongneed for consensus on widely acceptable validation methods.
Mechanical loads.
Among the research barriers listed in Figure 8, structural loads areconsistently ranked to be of medium importance. Accordingly, a deeper understandingof mechanical loads is considered one of the important research gaps in Figure 9. .0 0.2 0.4 0.6 0.8 1.0Wind turbine manufacturers are willing to help others implement wind farm control with their turbines.Study of hybrid (ie wind and solar) plants should be included in this overall topic areaAEP is the appropriate measure to determinethe success of a wind farm control schemeWind farm control will be primarily developed/provided/implemented by turbine manufacturersStrongly agree Agree Neutral Disagree Strongly Disagree
Figure 7: Statements without a clear trend among survey participants
Figure 8: Barriers preventing implementation of WFC, ranked by the survey participantsbetween 1 and 10 according to importance. 1 – most important, 10 – least important
Uncertainties concerning the effectiveness of WFC.
Finally, the understanding andquantification of statistical uncertainties is another major research gap according to thesurvey participants. This has only been explored to a very limited degree in WFC.The survey shows that the responding community considers topics such as the effect ofWFC on farm layout optimization, control for farm-farm interaction, and integration withother renewables and aeroacoustics not to be current research priorities that will acceleratethe implementation of industrial WFC.
4. Conclusions
WFC is an active and growing field of research where consensus of research findings, bestpractices, and identification of open research questions are needed. This paper has shown current .0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0RankValidation CampaignsUnderstanding Load ImpactUnderstanding/QuantifyingUncertaintiesDevelop Numerical ModelsReal-time Control StrategiesUnderstanding Dependenceof Atmospheric ConditionsDeveloping WidelyAccepted ValidationRobust Implementation(Curailment etc)Understanding how WFCimpacts optimal layoutGrid SupportFarm-Farm Cluster ControlIntegration WithStorage And SolarUnderstanding Sound Impact 13 5 5 4 4 4 1 2 3 4 1 0 03 8 7 6 7 6 3 2 2 1 0 1 04 6 5 9 7 3 1 4 2 1 1 1 17 4 6 6 5 4 4 3 4 1 0 1 13 8 1 9 7 3 2 6 3 0 3 0 05 8 4 2 4 7 5 5 2 0 0 1 27 2 4 2 5 3 6 4 2 0 6 3 11 2 3 5 0 4 13 2 3 5 3 3 02 0 5 3 2 6 3 4 8 9 1 2 10 2 2 0 3 1 4 2 6 12 8 4 11 0 3 0 1 2 3 3 8 11 8 2 30 1 1 0 1 0 0 2 1 0 9 11 190 0 0 0 0 0 0 5 1 1 5 16 16
Figure 9: Research priorities for WFC according to the questionnaire; the different topics areranked between 1 and 10 according to importance, 1 – most important, 13 – least importantareas of consensus, areas with further disagreement, and research gaps that need to be addressedin the go-to-market path of WFC. Reaching the common standard definition of concepts, metrics,and tools is recommended according to the survey results and TEM discussions. Moreover, astronger collaboration not only across institutions but also different disciplines working on thecontrol of wind power plants is desirable and one of the declared goals of FarmConners. Afterall, this survey has provided interesting insight into the beliefs and assumptions driving WFCresearch. Taking the outcomes of the TEM into account, a follow-up survey focusing on industrywill be conducted.
Acknowledgements
A majority of the authors have received funding from the European UnionsHorizon 2020 Research and Innovation Programme under Grant Agreement No.857844 as part of the FarmConners project. has received funding by theNetherlands Organization for Scientic Research (NWO) as part of his VIDI grantwith project number 17512 References [1] M S Adaramola and P A Krogstad. Experimental investigation of wake effects on windturbine performance.
Renewable Energy , 36(8):2078 – 2086, 2011.[2] J Bartl, F M¨uhle, and L Sætran. Wind tunnel study on power output and yaw momentsfor two yaw-controlled model wind turbines.
Wind Energy Science , 3(2):489–502, 2018.3] M Bastankhah and F Port´e-Agel. Wind farm power optimization via yaw angle control: Awind tunnel study.
Journal of Renewable and Sustainable Energy , 11(2):023301, 2019.[4] S Boersma, B M Doekemeijer, P M O Gebraad, P A Fleming, J Annoni, A K Scholbrock,J A Frederik, and J W van Wingerden. A tutorial on control-oriented modeling and controlof wind farms.
American Control Conference (ACC) , pages 1–18, 2017.[5] F. Campagnolo, V. Petrovi, C. L. Bottasso, and A. Croce. Wind tunnel testing of wakecontrol strategies.
Proceedings of the American Control Conference (ACC) , pages 513–518,2016.[6] F Campagnolo, V Petrovi, J Schreiber, E M Nanos, A Croce, and C L Bottasso. Windtunnel testing of a closed-loop wake deflection controller for wind farm power maximization.
Journal of Physics: Conference Series , 753(3), 2016.[7] B M Doekemeijer, P A Fleming, and J W van Wingerden. A tutorial on the synthesisand validation of a closed-loop wind farm controller using a steady-state surrogate model.
American Control Conference (ACC) , 2019.[8] B M Doekemeijer, S Kern, S Maturu, S Kanev, B Salbert, J Schreiber, F Campagnolo, C LBottasso, S Schuler, F Wilts, T Neumann, G Potenza, F Calabretta, F Fioretti, and J Wvan Wingerden. Field experiment for open-loop yaw-based wake steering at a commercialonshore wind farm in italy.
Wind Energy Science Discussions , 2020:1–22, 2020.[9] P A Fleming, J Annoni, J J Shah, L Wang, S Ananthan, Z Zhang, K Hutchings, P Wang,W Chen, and L Chen. Field test of wake steering at an offshore wind farm.
Wind EnergyScience , 2(1):229–239, 2017.[10] P A Fleming, J King, K Dykes, E Simley, J Roadman, A K Scholbrock, P Murphy, J KLundquist, P Moriarty, K Fleming, J van Dam, C J Bay, R Mudafort, H Lopez, J Skopek,M Scott, B Ryan, C Guernsey, and D Brake. Initial results from a field campaign of wakesteering applied at a commercial wind farm – part 1.
Wind Energy Science , 4(2):273–285,2019.[11] P A Fleming, J King, E Simley, J Roadman, A Scholbrock, P Murphy, J K Lundquist,P Moriarty, K Fleming, J van Dam, C Bay, R Mudafort, D Jager, J Skopek, M Scott,B Ryan, C Guernsey, and D Brake. Continued results from a field campaign of wakesteering applied at a commercial wind farm: Part 2.
Wind Energy Science Discussions ,2020:1–24, 2020.[12] J A Frederik, F Campagnolo, S Cacciola, R Weber, A Croce, C Bottasso, and J W vanWingerden. Periodic dynamic induction control of wind farms: proving the potential insimulations and wind tunnel experiments.
Wind Energy Science , 5(1):245257, 2020.[13] J A Frederik, B M Doekemeijer, S Mulders, and J W van Wingerden. On wind farm wakemixing strategies using dynamic individual pitch control.
The science of making torquefrom wind , 2020.[14] M F Howland, S K Lele, and J O Dabiri. Wind farm power optimization through wakesteering.
Proceedings of the National Academy of Sciences , 116(29):14495–14500, 2019.[15] A C Kheirabadi and R Nagamune. A quantitative review of wind farm control with theobjective of wind farm power maximization.
Journal of Wind Engineering and IndustrialAerodynamics , 192:45 – 73, 2019.[16] T Knudsen, T Bak, and M Svenstrup. Survey of wind farm control – power and fatigueoptimization.
Wind Energy , 18(8):1333–1351, 2015.[17] W Munters and J Meyers. Towards practical dynamic induction control of wind farms:analysis of optimally controlled wind-farm boundary layers and sinusoidal induction controlof first-row turbines.
Wind Energy Science , 3(1):409–425, 2018.18] J Park, S D Kwon, and K H Law. A data-driven approach for cooperative wind farmcontrol. In , pages 525–530, 2016.[19] J Schottler, A H¨olling, J Peinke, and M H¨olling.
Wind tunnel tests on controllable modelwind turbines in yaw . 2016.[20] D van der Hoek, S Kanev, J Allin, D Bieniek, and N Mittelmeier. Effects of axial inductioncontrol on wind farm energy production - a field test.