Vertical axis wind turbine

Abstract

A vertical axis wind turbine (VAWT) (1) comprising a turbine (102) for coupling to a generator (101), the turbine (102) rotatable about a vertical turbine rotation axis and comprising a plurality of blade devices (2) disposed around the vertical turbine rotation axis coupled via at least one turbine rotor arm (105) to a rotor of the VAWT, each blade device (2) comprising at least one blade (6) and at least one blade actuator (7) coupled to the blade (6) configured to vary a pitch angle of the at least one blade, the turbine further comprising : a sensor system (3) including a blade sensor system, and a control system (4) connected to the sensor system and to the blade actuator (7) in a feedback loop for real-time control of the blade pitch angle, the control system configured to receive measurement data from the sensor system and to output a target blade pitch angle to the blade actuator (7). The blade sensor system (8) further comprises at least one blade deformation sensor (10) arranged on the blade and a plurality of pressure sensors (9) arranged on a surface of the blade, on opposite sides of the blade relative to a chord of the blade, the at least one deformation sensor and plurality of pressure sensors connected to the control system and outputting at least a portion of said measurement data, the control system comprising software configured to compute a fluid flow state around the blade based on the measurement data output by the pressure sensors, the control system further configured to compute said target blade pitch angle based at least on said fluid flow state and an output of said blade deformation sensor as a function of a control objective, said control objective stored in the control system, input into the control system, or computed in the control system as a function of at least a wind condition impinging upon the VAWT.

Description

[0001] P3049PC00

[0002] VERTICAL AXIS WIND TURBINE

[0003] The present invention relates to a vertical axis wind turbine.

[0004] Vertical axis wind turbines (VAWTs) are quieter, safer for wildlife, and capable of higher energy density than horizontal axis wind turbines. VAWTs however have lower efficiency and less structural stability in high winds compared to horizontal axis turbines. These drawbacks are due in particular to problems of dynamic stall and aerodynamic phenomenon causing the creation, growth, and separation of a large and coherent vortex.

[0005] To improve the efficiency of VAWTs and overcome some of the aforementioned drawbacks, it is known to control blade pitching in an open-loop manner in which the blade pitching follows pre-defined kinematics. The drawback of open-loop control is that the blade pitching does not adapt well to varying wind conditions. To overcome this drawback, it is known, at least theoretically, to perform a real-time feedback blade pitch control system based on the real-time flow velocity around the blade and using a computational fluid dynamics method with to evaluate the performance of the pitch control system on improving VAWT performance, as described in Journal of Wind Engineering and Industrial Aerodynamics « A novel real-time feedback pitch angle control system for vertical-axis wind turbines » Chen et al. 2019. A weakness of such methods is the reliance of the computational model on the measurement of flow data from an external measurement point, which is complex, as well as the intensive computation of flow velocity required for real-time control. Moreover dynamic forces leading to vibration and factors adversely affecting structural integrity of the VAWT blades may not be reliably deduced from the aforementioned computational fluid dynamics method.

[0006] In view of the foregoing, it is an object of this invention to provide a vertical axis wind turbine with high performance and efficiency in conjunction with reduced wear and a long service time

[0007] It is advantageous to provide a vertical axis wind turbine which can operate efficiently over a large range of wind speeds.

[0008] It is advantageous to provide a vertical axis wind turbine which is economical to produce, install, and operate.

[0009] It is advantageous to provide a vertical axis wind turbine which is reliable and robust.

[0010] Objects of this invention have been achieved by providing a vertical axis wind turbine according to claim 1. Dependent claims set forth various advantageous features of embodiments of the invention. P3049PC00

[0011] Disclosed herein is a vertical axis wind turbine (VAWT) comprising a turbine for coupling to a generator, the turbine rotatable about a vertical turbine rotation axis and comprising a plurality of blade devices disposed around the vertical turbine rotation axis coupled via at least one turbine rotor arm to a rotor of the VAWT, each blade device comprising at least one blade and at least one blade actuator coupled to the blade configured to vary a pitch angle of the at least one blade, the turbine further comprising : a sensor system including a blade sensor system, and a control system connected to the sensor system and to the blade actuator in a feedback loop for real-time control of the blade pitch angle, the control system configured to receive measurement data from the sensor system and to output a target blade pitch angle to the blade actuator.

[0012] The blade sensor system further comprises at least one blade deformation sensor arranged on the blade and a plurality of pressure sensors arranged on a surface of the blade, on opposite sides of the blade relative to a chord of the blade, the at least one deformation sensor and plurality of pressure sensors connected to the control system and outputting at least a portion of said measurement data, the control system comprising software configured to compute a fluid flow state around the blade based on the measurement data output by the pressure sensors, the control system further configured to compute said target blade pitch angle based at least on said fluid flow state and an output of said blade deformation sensor as a function of a control objective, said control objective stored in the control system, input into the control system, or computed in the control system as a function of at least a wind condition impinging upon the VAWT.

[0013] In an advantageous embodiment, the pressure sensors include at least one pressure sensor on each side of the blade within a distance from a leading edge of the blade ranging from 5% to 30% of a chord length of the blade.

[0014] In an advantageous embodiment, there are a plurality of pressure sensors arranged in a distributed manner chordwise between the leading edge and a trailing edge of the blade.

[0015] In an advantageous embodiment, there are at least three pressure sensors arranged in said distributed manner chordwise between the leading edge and a trailing edge of the blade.

[0016] In an advantageous embodiment, there are a plurality of pressure sensors arranged in a distributed manner spanwise between a first extremity and a second extremity of the blade. P3049PC00

[0017] In an advantageous embodiment, there are at least three pressure sensors arranged in said distributed manner spanwise between the first extremity and the second extremity of the blade.

[0018] In an advantageous embodiment, the at least one blade deformation sensor comprises an inertial measurement sensor (IMU).

[0019] In an advantageous embodiment, there are a plurality of said blade deformation sensors mounted on each blade.

[0020] In an advantageous embodiment, the at least one blade deformation sensor is positioned in proximity to said plurality of pressure sensors and forms therewith a sensor set extending in a band chordwise across the blade surface.

[0021] In an advantageous embodiment, the sensor set comprises a flexible circuit with at the least one blade deformation sensor and the plurality of pressure sensors connected thereon.

[0022] In an advantageous embodiment, the pressure sensors are MEMs pressure sensors.

[0023] In an advantageous embodiment, the blade device comprises a pair of said blades coupled to a centrally positioned said blade actuator, each blade extending from the blade actuator to a free end extremity.

[0024] In an advantageous embodiment, the at least one blade deformation sensor and at least some of said plurality of pressure sensors are mounted proximate said free end extremity of each blade, distal from the blade actuator.

[0025] In an advantageous embodiment, the sensor system further comprises a wind speed sensor connected to the control system, a measurement output of the wind speed sensor wind constituting or partially forming said wind condition impinging upon the VAWT.

[0026] In an advantageous embodiment, the blade actuator comprises a rotary electrical motor mounted on said turbine rotor arm and coupled to a rotor of the blade device.

[0027] In an advantageous embodiment, the control system comprises an individual blade controller for each blade device.

[0028] In an advantageous embodiment, the blade controller is mounted proximate, on, or in, the blade actuator. P3049PC00

[0029] In an advantageous embodiment, the control system comprises control software including an initialization module, a calibration module and a real-time blade control module.

[0030] In an advantageous embodiment, the initialization module is configured to characterise and discretise the state space in which the VAWT typically operates.

[0031] In an advantageous embodiment, the calibration module is configured to optimise the blade pitch angle value associated the each discrete node obtained from the initialization module.

[0032] In an advantageous embodiment, the real-time blade control module is configured to compute the optimal blade pitch angle at any instant by performing a weighted average of the optimal blade pitch angle values obtained in the calibration module, with weights proportional to the distance between the current flow state and the discrete nodes in the state space.

[0033] Also disclosed herein is a method of controlling a VAWT according to any preceding embodiment comprising the steps of : inputting measurement signals from the sensor system, in particular the blade deformation sensors and blade profile pressure sensors into the control system , computing a flow state from the pressure sensor measurement signals, computing a power spectrum from the blade deformation measurement signals, combining and processing the measurement signals to generate a state vector s^t), defining a control objective based at least in part on a wind speed, defining a control law adapted to achieve said defined objective, computing a blade pitch angle in the control system using the state vector and the defined control law.

[0034] Embodiments of the invention advantageously provide improved performance of vertical-axis wind turbines, in particular to increase their electricity production, extend their lifespan, and broaden their range of operating wind conditions.

[0035] Further objects and advantageous features of the invention will be apparent from the claims, from the detailed description, and annexed drawings, in which:

[0036] Figure la is a perspective schematic representation of a vertical axis wind turbine (VAWT) according to an embodiment of the invention;

[0037] Figure lb is a schematic simplified side view of a portion of the VAWT according to an embodiment of the invention; P3049PC00

[0038] Figure 2a is a schematic block diagram of a sensor and control system of a VAWT according to an embodiment of the invention;

[0039] Figure 2b is a simplified flow diagram of the sensor and control system of a VAWT according to an embodiment of the invention;

[0040] Figure 3a is a schematic top view of a VAWT illustrating certain parameters related to velocity and pitch angle of blades of a VAWT;

[0041] Figures 3b and 3c are plots of parameters illustrated in figure 3a;

[0042] Figure 3d is an illustration of wind flow over a blade at different angles of rotation about a turbine axis and figure 3e is a plot of blade efficiency as a function of the rotation angle by way of example;

[0043] Figure 3f illustrates parameters related to a pitch angle of a turbine blade;

[0044] Figure 3g illustrates a plot of an average power spectrum of blade deformation sensors of a turbine blade according to an embodiment of the invention;

[0045] Figure 3h illustrates a 2D plot of an optimal pitch angle determination using a state vector and selected control law of a blade pitch control method according to an embodiment of the invention;

[0046] Figure 4a is a simplified cross-sectional top view of a blade of a VAWT according to an embodiment of the invention;

[0047] Figure 4b is a simplified schematic side view of a sensor set of a blade of a VAWT according to an embodiment of the invention;

[0048] Figures 4c to 4e are side perspective views of a blade of a VAWT according to an embodiment of the invention illustrating the position of pressure sensors distributed on the blade surface of different variants;

[0049] Figure 5 is a schematic illustration of a control method of a blade pitch angle of a VAWT according to an embodiment of the invention;

[0050] Figure 6 are illustrations of an experimental set up of a blade of a VAWT with pressure sensors;

[0051] Figure 7 shows plots of lift coefficient respectively drag coefficient relative to t / T of the experimental blade of figure 6;

[0052] Figure 8 is a schematic overview diagram of a control process of a VAWT according to an embodiment of the invention; P3049PC00

[0053] Figures 9a to 9e are schematic flow charts of algorithms of a control system of a VAWT according to embodiments of the invention;

[0054] Figure 10 is a plot of power versus wind speed of a VAWT and associated examples of actuation objectives of a VAWT according to an embodiment of the invention.

[0055] Referring to figures, starting with figures la and lb, a vertical axis wind turbine 1 comprises a turbine 102 supported on a tower 103 , the turbine coupled to an electrical generator 101 to convert wind energy to electrical energy. The turbine 102 comprises a turbine rotor (not visible) connected to a hub 104 rotating about a turbine axis At, a plurality of blade devices 2, and at least on turbine rotor arm 105 interconnecting the plurality of blade devices 2 to the hub 104. In the illustrated embodiment, each blade device is mounted a radial end of a turbine rotor arm. However, in variants it is possible to have the blade devices mounted on a ring shaped support that in turn is connected to the hub via one or more turbine rotor arms, such that the number of turbine rotor arms does not necessarily match the number of blade devices 2.

[0056] The turbine rotor couples the turbine 2 to the generator 101 which may be positioned at a bottom end of the tower 103, or alternatively adjacent the hub 104 or at any position between the hub and the bottom of the tower.

[0057] The number of blade devices 2 is preferably from 2 to 5, possibly more, preferably arranged equidistant from each other around the turbine rotor axis. It may however be noted that various other blade arrangements are possible within the scope of the invention for instance having pairs of blades arranged symmetrically around the turbine rotor axis however with a distance between successive blades that is not equidistant. Also, although it is preferred to have equidistant separation between blades and each of the blades having an identical shape, it is possible to have a first and second set of blades with different profiles or dimensions.

[0058] In the illustrated embodiments there are three blade devices 2 equidistant from each other and at a fixed radius from the turbine rotor axis At. Each blade device 2 comprises at least one blade 6 coupled to a blade actuator 7.

[0059] The blade actuator may for instance comprise a rotary electrical motor coupled to a pivot arm of the blade directly, or via a gear or belt transmission. The electrical motor may for instance comprise a brushless DC motor or a stepping motor. In variants, the drive force may however be hydraulic or pneumatic instead of electric. In variants, the blade actuator may have other configurations such as a linear actuator coupled via a linkage or arm to the blade to rotate the blade about a blade pivot coupling. P3049PC00

[0060] In a preferred embodiment, the blade 6 is rotatably mounted to the turbine rotor arm 105, rotatable about a blade axis Ab, the blade device having a blade rotor coupled to the blade actuator 7 configured to pivot the blade 6 about the blade axis Ab thereby changing the pitch angle of the blade.

[0061] Each blade device has its own blade actuator 7 such that the pitch angle of each blade may be individually actuated.

[0062] In a preferred embodiment, the blade axis Ab extends vertically however within the scope of the invention it is possible to have a slight inclination angle of the blade rotation axis relative to a vertical direction.

[0063] In an embodiment, each blade device 2 is coupled to a turbine rotor arm 105 substantially at a central vertical position such that there are a pair of blades 6 extending in opposite directions from the coupling of the blade device to the turbine rotor arm 105. The blade actuator 7 may be positioned at the central vertical position between the pair of vertically extending blades. Other per se known blade and turbine rotor arm arrangements may however be provided. For instance each blade may be coupled to a turbine rotor arm positioned at opposing vertical bottom and top ends of the blade. In such case, the blade actuator may be positioned either on the top turbine rotor arm or the bottom turbine rotor arm. For VAWTs having a long blade span height, it is also possible to have more than two turbine rotor arms coupling to each blade device, for instance three rotor arms, where one is in the middle of the span height and the other two at or proximate ends of the blades.

[0064] The turbine 102 of the VAWT 1 according to embodiments of the invention, includes a sensor system 3 and a control system 4 connected to the sensor system 3 and to the blade actuators 7, configured to control the rotation of the blade about its axis or alternatively to control a flap or other flow engaging device of the blade that is configured to change the effective pitch angle or profile of the blade.

[0065] The sensor system 3 comprises a plurality of blade sensors 8, a general wind sensor or anemometer 13 a turbine angular position or rotation sensor 14, and a turbine torque sensor 16.

[0066] The anemometer 13 may be connected to the tower 103 and positioned above or below the blade device 2, or alternatively not connected to the tower 103, positioned at a certain distance horizontally from the turbine 102, configured to measure the general wind speed. It may be noted that a plurality of VAWTs may be connected to a single anemometer. P3049PC00

[0067] The turbine torque sensor 16 is configured to measure an aerodynamic force acting on the blade devices 2 that drives the turbine 102 due to the wind energy. In a first variant, the torque applied on the turbine by the blade devices 2 may be measured as a whole, for instance by the torque applied on the generator

[0068] 101. In a second variant the torque applied on the turbine by the blade devices 2 may be performed by measuring the torque applied by each blade device 2 on the hub 104 of the turbine 102. In the first variant, the generator torque may be measured by various per se known means, including for instance a measurement based on the electric parameters of the generator. In the second variant, the turbine torque sensor 16 may comprise one or more strain gauges 16a, 16b, 16c placed for instance on the turbine rotor arm 105 and / or on the blade rotor proximate the coupling of the blade rotor to the turbine arm. It may be noted that other force measurement sensors measuring the force or pressure applied by the blade against the turbine rotor arm 105, for instance capacitive sensors or other forms of pressure sensors, for instance positioned at the coupling between the blade rotor and the turbine rotor arm, may be employed to measure the aerodynamic force applied by the blade device on the turbine rotor. Various force sensing sensors are per se known and could be employed within the scope of the invention. In one of the preferred embodiments, the use of strain gauges provides a simple, reliable and easy to control force sensing measurement device.

[0069] The turbine angular position sensor 14 measures the angular position of the turbine, and thus allows to determine turbine rotational speed and the position of each blade device relative to general wind direction. Various per se known sensors may be used to measure the rotational position of the turbine

[0070] 102. The turbine angular position sensor may be mounted on the hub 104, or on the generator 101, or at any position along the turbine rotor (not visible).

[0071] The blade sensors 8 according to an aspect of the invention include a plurality of blade profile pressure sensors 9 having measurement points arranged on the surface of the blade, on both sides of the blade (relative to the chord). The blade sensors 8 according to an aspect of the invention further include at least one blade deformation sensor 10, preferably a plurality of blade deformation sensors 10. In an advantageous embodiment, the at least one or plurality of blade deformation sensors 10 are positioned in proximity to at least some of the blade profile pressure sensors.

[0072] The blade sensors 8 further comprise a blade angular position sensor 11 configured to measure the rotational angle of the blade, whereby the blade angular position sensor 11 may be integrated in the blade actuator 7, for instance in the form of an optical encoder or any other form of per se known rotation sensor for rotary components.

[0073] The blade sensors 8 may further comprise a blade torque sensor 12 which may be implemented for instance as a strain gauge mounted on a portion of the rotor of a blade proximate the blade actuator 7 or P3049PC00 may be integrated in the blade actuator, various per se known torque sensors on or in electrical motors being per se known. It may be noted that the blade torque sensor 12 is optional but may contribute to control of the blade pitch angle.

[0074] The blade profile pressure sensors 9 on each side of the blade profile relative to the chord comprise at least first pressure sensors 9a positioned proximate the leading edge 24a of the blade 6, preferably within a range of 5% to 25% of the chord length ChL from the leading edge 24a. It may be noted that although pressure sensors should be positioned on both sides of the blade, they do not need to be necessarily symmetrically positioned about the chord plane.

[0075] Preferably, in order to measure more accurately a distribution of flow around the blade profile, there are on each side of the blade surface at least a second pressure sensor, preferably at least three or more pressure sensors 9 spaced apart from each other extending chordwise between the leading edge 24a and the trailing edge 24b of the blade.

[0076] The pressure sensors 9 may optionally be provided at a plurality of different heights, for instance two, three or more different heights between a first extremity 28a and a second extremity 28b of each blade as schematically illustrated in figures 4e and 4f. In these variants, both the chordwise distribution of pressure and spanwise distribution of pressure on a blade may be measured in order to obtain data on the pressure distribution over the entire blade profile, thus enabling computation of an accurate flow profile over the span length of each blade.

[0077] The preferred spanwise position of the pressure sensors may vary depending on the configuration of the blade device, in particular the position of the turbine rotor arm 105 coupling to the blade device 2. In a configuration with a turbine rotor arm 105 coupling to the blade device at a spanwise central position as illustrated in figure 1, there is at least one set of pressure sensors 9 arranged on each blade 6 of each blade device 2, preferably positioned proximate the free blade extremity 28b and distal from the blade actuator 7 and centre coupling.

[0078] In a preferred embodiment, the blade deformation sensor 10 comprises or consists of an inertial measurement sensor. The inertial measurement sensor measures acceleration of the blade at the position of placement of the sensor, which allows, inter alia, to capture dynamic deformations, in particular vibration of the blade resulting from aerodynamic forces acting upon the blade. Dynamic structural deformations, which may include bending of the blade relative to the blade pitching axis Ab and torsion of the blade about its pitching axis Ab, are a source of material fatigue and can cause structural failure of the blade after a certain duration. One of the important objectives is to improve performance and P3049PC00 efficiency of the VAWT however without adversely affecting service lifetime, since overall performance should take into account material use and maintenance downtime.

[0079] Instead of inertial measurement sensors, or in combination with inertial measurement sensors, the blade deformation sensors may comprise a stress measurement sensor, for instance in the form of a strain gauge, mounted on the blade. A stress measurement sensor also allows to capture vibration and other structural deformations of the blade, whereby in contrast to an inertial measurement, static deformation can also be measured with a strain gauge.

[0080] The blade deformation sensors 10 may be placed on each side of the blade profile relative to the chord similar to the pressure sensors 9. However, in variants, the blade deformation sensors 10 may be positioned only on one side of the blade since the dynamic deformation of the blade can be measured on only one side of the blade. Moreover, the number of blade deformation sensors may be different from the number of pressure sensors 9, in particular less blade deformation sensors are needed to provide accurate and reliable vibration other dynamic stress measurement data of the blade dynamic behaviour. In embodiments, the blade deformation sensors 10 comprise at least a first blade deformation sensor 10 positioned proximate the leading edge 24a of the blade 6, preferably within a range of 5% to 25% of the chord length ChL from the leading edge 24a.

[0081] Preferably, in order to measure more accurately a dynamic and optionally static deformation of the blade, there may be at least a second blade deformation sensor, possibly three or more deformation sensors 10, spaced apart from each other extending chordwise between the leading edge 24a and the trailing edge 24b of the blade. A plurality of blade deformation sensors allows to increase redundancy and improve the signal quality with averaging.

[0082] The deformation sensors 10 may optionally be provided at a plurality of different heights, for instance two, three or more different heights between a first extremity 28a and a second extremity 28b of each blade as schematically illustrated in figure 4f. In this variant, both the chordwise and spanwise deformations of the blade may be measured in order to obtain more data over the entire blade profile, for more accurate computation of vibration or stress conditions.

[0083] The preferred spanwise position of the deformation sensors may vary depending on the configuration of the blade device, in particular the position of the turbine rotor arm 105 coupling to the blade device 2. The optimal position of the deformation sensors is in a zone of the blade where the largest amplitude deformations can be captured. In a configuration with a turbine rotor arm 105 coupling to the blade device at a spanwise central position as illustrated in figure 1, and the use of inertial measurement sensors, there is at least one set of inertial measurement sensors 10 arranged on each blade 6 of each P3049PC00 blade device 2, preferably positioned proximate the free blade extremity 28b and distal from the blade actuator 7 and centre coupling.

[0084] Measurement data output by the blade deformation sensors 10 in combination with the blade profde pressure sensors 9, both directly mounted on each blade, and input into the control system, may advantageously be used to compute flow conditions and anticipate instabilities, in particular dynamic stall conditions, with greater accuracy and reliability than the use of pressure sensors alone, or deformation sensors alone. In addition, conditions adversely affecting service lifetime of the turbine 102 can be monitored and reduced.

[0085] An accurate estimation of the flow allows accurate modelling and anticipation of dynamic stall conditions, taking into account the latency in the development of such conditions for efficient realtime control of the blade pitch angle.

[0086] The control system 4 comprises a general turbine control 18 and individual blade controllers 20. Each of the blade controllers 20 controls individually a blade actuator 7 coupled to a blade 6 of the blade device 2.

[0087] The control system 4 is connected to and receives input from the sensors system 3, and outputs a control signal to each blade actuator 7 for separate individual control of each blade actuator 7.

[0088] The control system comprises control software 22 that receives data from the sensors and processes this data to control the blade rotational angle. The control software 22 may include an initialization module 22a, a calibration module 22b and a realtime blade control module 22c. It may be noted that various algorithms may be integrated into a single software code, the presentation of modules herein intended to describe the functional architecture of the software and not necessarily distinct and separable software modules. In other words, the term “module” mentioned herein is intended to describe a functional aspect of the control software 22.

[0089] Embodiments of the invention allow optimal control of the pitch angle of each blade in realtime taking into account unsteadiness of the incoming flow of air as well as the occurrence of dynamic stall and efficiency drop, these two phenomenon being briefly described below with reference to figures 3a to 3f:

[0090] Unsteadiness of the VAWT’s flow

[0091] The orientation between a blade and the incoming flow has a strong influence on the flow response around the blade and the aerodynamic loads. In a VAWT, this orientation is dynamically changing throughout the rotation and is responsible for the large flow unsteadiness. The effective incoming flow direction perceived by the blade Ueff is a combination of the blade speed and the incoming P3049PC00 flow speed (figure 3a), and varies during the rotation. The orientation between the blade and the perceived incoming flow direction is the angle of attack aeff, and is presented on the plot of figure 3b over a blade rotation. It may be noted that the effective angle in Fig. 3b is for a potential flow, without viscosity nor vorticity, and the actual flow and effective angle in realistic conditions is much more complex, mostly because of flow separation and dynamic stall. The angle of attack can reach +-35° and more in high wind conditions. As a result from the vertical axis of rotation, the blade is alternatively opposing the incoming wind and going in the same direction. The magnitude of the velocity Ueff perceived by the blade consequently varies a lot during the rotation (figure 3c).

[0092] Dynamic stall and efficiency drop

[0093] The shorter lifetime and lower efficiency of VAWTs compared to HAWTs are rooted in the dynamic stall, an unsteady aerodynamic phenomenon which consists of the creation, growth and eventually separation of a large and coherent vortex called the leading edge vortex. This vortex occurrence is due to the fast variations of angle of attack as mentioned above. The development of the leading edge vortex is illustrated in figure 3d, which shows the blade at various azimuthal positions and the magnitude of the vorticity field. The dynamic stall effect on efficiency is highlighted on the plot of figure 3e.

[0094] Changing the blade's orientation during the rotation significantly impacts the dynamic stall and the overall flow behavior. In figure 3f the unactuated blade has an angle of attack a. The actuation changes the blade angle about its own axis, or pitch angle, to modify the orientation. The new orientation becomes aeff= a + aPitch.

[0095] The purpose of implementing real-time blade control on a vertical-axis wind turbine (VAWT) is ultimately to optimize the regulation of the aerodynamic force experienced by the blade and transmitted to the turbine rotor such that the control objectives are attained. The overarching control objective in most operating conditions will be to seek the highest efficiency in converting wind to electrical energy, although this may be accompanied by other control objectives such as reducing noise emissions, limiting component stress and wear, managing operating conditions at low or high wind velocities and safety.

[0096] The aerodynamics of a VAWT blade operating in an inviscid flow can be modelled as a combination of a pitching (pivot about its own axis) and surging blade (varying oncoming flow velocity, Greenberg’s Force Prediction for Vertical-Axis Wind Turbine Blades David Bensason, Sebastien Le Fouest, Anna Young, and Karen Mulleners AIAA Journal 2022 60:7, 4467-4470) using Greenberg’s theory (Greenberg J., “Airfoil in Sinusoidal Motion in a Pulsating Stream,” NACA TN 1326, 1947). Here, the aerodynamic force experienced by the wind turbine blade is correlated to the turbine geometry (fixed) and the dynamic oscillation of the blade’s effective angle of attack and incoming flow velocity, as well P3049PC00 as their time derivatives. Assuming the turbine operates in an inviscid flow, the effective angle of attack and incoming flow velocity are determined solely by geometry.

[0097] The inviscid flow assumption is a significant limitation to theoretical models of wind turbine aerodynamics. At low tip-speed ratios A = — < 2.5, a VAWT blade’s effective angle of attack exceeds ^co a critical angle, known as the static stall angle, even under steady wind conditions. Extended excursions of the effective angle of attack beyond this critical angle cause dynamic stall — a phenomenon in which airflow fully separates from the blade. This leads to a significant drop in aerodynamic performance and heavy transient force fluctuations. Dynamic stall is a viscous flow phenomenon, rendering inviscid flow models invalid.

[0098] The aerodynamic force experienced by a VAWT blade operating in real-world conditions remains correlated to the effective angle of attack a and incoming flow velocity Ueff . However, there are no satisfactory theoretical models for accurately predicting either of these parameters in viscous flows. Moreover, no current model relates the aerodynamic force to a blade undergoing pitching and surging motions beyond the critical stall angle.

[0099] Embodiments of the present invention address these shortcomings by integrating both blade deformation sensors on the blade devices and blade profile pressure sensors that allow to compute real-time data about both the effective angle of attack (a) and the incoming flow velocity (t / eff)- This data is processed by a data-driven algorithm, which is optimised to regulate the aerodynamic force experienced by the wind turbine blade with a given objective, improving performance and stability, while reducing wear, in all wind conditions.

[0100] Conventional practical implementations of blade pitch control in vertical-axis wind turbines, generally involves sensors to measure undisturbed free-stream wind speed and direction. The turbine’s rotational frequency and blade azimuthal positions are also monitored. Based on this information, a pre-determined pitch profile is instructed to the wind turbine blades. The advantages are an improvement of the turbine coefficient of power (efficiency) at steady wind conditions and low cost of implementation. The drawbacks are poor representation of the effective flow conditions at the turbine’s blade level, failure to maintain structural integrity of the turbine in gusty wind conditions and limited performance improvements. P3049PC00

[0101] In embodiments of the invention, the instrumentation of the blades with embedded sensors to measure the effective flow conditions and stress on the blade provides data that may be processed by algorithms to develop and optimize control laws, enabling the regulation of aerodynamic forces in accordance with predefined performance objectives.

[0102] Direct measurement of the effective angle of attack and flow velocity acting on a vertical-axis wind turbine (VAWT) blade presents significant challenges, since parameters vary along both the chord and span of the blade, meaning any point measurement provides only partial information. Parameter values for steady wind conditions and fixed blade positions can be determined unambiguously, however in dynamic operating conditions where the blade may be subjected to vortex-dominated flows and the effective angle of attack and incoming flow velocity are functions of both time and space, it is advantageous to capture data on flow conditions over the blade surface for accurate real-time blade pitch control. The sensor system of the VAWT according to embodiments of the invention is configured to provide sufficient information about the effective wind and flow conditions, as well as vibration and structural forces acting on the blades such that data-driven algorithms may advantageously deduce information about the flow around each blade that may advantageously include:

[0103] Table 1

[0104] The importance of each information is defined as follows:

[0105] • level 1: essential to perform real-time blade control.

[0106] • level 2: significant improvement to the quality of the real-time blade control.

[0107] • level 3: allows optimal real-time blade control.

[0108] The pressure sensors 9 mounted on the blades 6 may include surface pressure taps that can be piezoresistive, capacitive or fibre optic, and microelectromechanical system sensors (MEMS). Pressure P3049PC00 measurement sensors may also be indirect and include heat flux sensors, which include hot-film sensors, thin-film heat flux sensors and MEMS-based thermal sensors. The table below presents pressure sensor setups according to embodiments of the invention and their benefits for real-time blade pitch control : Table 2

[0109] Using a combination of the pressure sensor setups #2 and #3 in the table 2 above, all the flow physics mentioned in table 1 are captured for a 2D cross-section of the turbine blade. In case of a wind turbine operating in a vertically sheared flow, the blades will experience a significant spanwise variation in pressure distribution. Additional spanwise measurements can be acquired using multiple setups such as those presented in table 2. Although spanwise variation of blade pitch is not possible with a rigid blade, the control system algorithms can compute the best compromise based on multiple spanwise measurements for optimal performance for each blade.

[0110] The airflow computation over the blade surface using the pressure sensors 9 may advantageously be combined with the measurements of the blade deformation sensors 10 and optionally also the turbine P3049PC00 torque sensor 16 to optimise the control settings for various given objectives, such as maximizing torque or minimizing structural wear.

[0111] Experimental example

[0112] Referring to figure 6, an experimental setup of a blade with pressure sensors, to experimentally study the pressure sensor setup of a VAWT blade according to embodiments of the invention, is illustrated. The application is a pitching blade, which consists of a NACA 0021 blade facing the flow and undergoing a sinusoidal rotating motion about its axis, or pitching motion. This application is selected because when the pitching amplitude is sufficiently high, the flow resembles that of the VAWT: a large and coherent vortex forms and causes dynamic stall, with a clear effect on the aerodynamic loads. With symmetric pitching, the dynamic stall occurs on each half. The blade is given a baseline sinusoidal motion with a maximum amplitude of +-25°, which creates the necessary conditions for dynamic stall. The control can operate within a range of +-10° around this baseline motion to achieve its objective. The control objective is to improve the lift-to-drag ratio, a metric commonly used to quantify the performance of a blade in the field of wind turbines. The experiment was performed in an open-section wind tunnel, at a wind speed of 16m / s.

[0113] The pressure sensors are distributed along the blade surface with pressure taps, connected to the sensors via tubes. The sensors are mounted on a PCB, which collects the pressure signals and sends them to the controller. A total of 29 sensors were mounted on the blade to measure the full development of the leading edge footprint on the blade surface, but only 5 sensors were used to perform the optimization and control. We have employed flush-mounted differential pressure sensors, but note that the control can work with any technology able to measure the pressure.

[0114] The blade, which faces the wind tunnel outlet, is connected to the pitching motor via a shaft. The pressure is acquired by the sensors, and sent via the PCB, to the controller, as input signal for the cluster-based feedback control algorithm. The algorithm runs on a raspberry pi, a board computer which can interface with a range of hardware. At each instant, given the pressure signal, the algorithm determines the optimal angle and sends the corresponding command to the pitching motor.

[0115] The results from the optimized real-time control are presented in figure 7. The actuation objective (lift-to-drag ratio L / D) is significantly improved by the actuation. The lift is increased by 23%, the drag reduced by 45%, and the overall L / D ratio is increased by 124%. The plots show the phase- averaged lift and drag coefficients, Cl and Cd, during a pitching cycle of period T=1 second. The black and red curves indicate the unactuated and actuated forces, respectively. The green and red shaded areas highlight a performance increase and decrease, respectively. The slight decrease of the P3049PC00 lift peaks is followed by a large increase during the rest of the pitching motion. The drag is decreased throughout the pitching cycle.

[0116] Example of blade sensors implementation

[0117] In an example of implementation of blade sensors according to an embodiment of the invention, the blade profile pressure sensors and blade deformation sensors may be integrated in a sensor set, for instance comprising a flexible PCB equipped with pressure sensors and inertial measurement sensors (also commonly known as inertial measurement unit “IMU). The pressure sensors may for instance be in the form of MEMS pressure sensors, per se known in the field of pressure measurement.

[0118] The blade deformation sensors, for instance in the form of per se known IMUs, enable computation of blade structural vibrations, whereby IMUs can measure both acceleration and angular rate. Measuring the angular rate allows to isolate blade pitching from actual blade structural vibration.

[0119] Each sensor set may for instance comprise at least three IMUs to increase redundancy and improve the signal quality with averaging.

[0120] An example of a sensor set is schematically illustrated in figure 4b. In this example, there are a plurality of pressure sensors, e.g. 15 sensors per blade face, whereby the preferred number of pressure sensors depends on the desired accuracy when computing the surface pressure integral (described in more detail below in relation to figure 4a) and determining the chordwise location of the suction peak. For control applications, large relative errors on the force estimation are permissible. The goal is to obtain an indication of the total aerodynamic force’s magnitude and orientation.

[0121] The blade chord c is split into 3 regions, whereby in this example about one third of the blade profile pressure sensors are disposed. The pressure sensors are equidistant within a region. In this sensor set example illustrated in figure 4b, the regions and the corresponding number of sensors are for instance : o 0% < x / c < 10%: 5 sensors o 10% < x / c < 30%: 5 sensors o 30% < x / c < 100%: 5 sensors

[0122] The exact location of the pressure sensors 9 can vary and be optimized for specific applications. An important goal is to best discretize the chordwise variation of the surface pressure. Typically, the greatest variations of pressure on an airfoil will occur at the leading edge. It is advised to have the P3049PC00 highest density of pressure sensors there. Closer to the trailing edge 24b, the pressure profile is close to linear both on the pressure and suction sides.

[0123] The blade deformation sensors (e.g. IMUs) 10 may for instance be positioned equidistant within a chordwise extending band, although this is not a critical condition and the distance between IMUs may not be equidistant.

[0124] Example of real-time control flow

[0125] The control flow is performed in parallel for each individual blade by each blade controller 20.

[0126] The actuation frequency may be based on the following three parameters: the need for signal filtering, the sensors sampling frequencies, and the need to capture all relevant physics. This frequency is typically higher than that of the turbine controller 18.

[0127] Step 1: Acquire data at given time instance

[0128] Input the measurement signals from the sensor system 3, in particular the blade deformation sensors 10 and blade profile pressure sensors 9 into the control system 4

[0129] Step 2: Select data from the sensor belt experiencing highest pressure difference

[0130] The real-time control calculation is based on the flow state measured by the blade sensors 8. The measurements will differ between the blade sensors at different positions on the blade 6, for instance between the upper and lower sensor sets 9, 10 of the embodiment illustrated in figure 1. Most of the time, the upper sensor set 9, 10 will experience a larger wind speed than the lower sensor set 9, 10. To ensure safe operation of the turbine, the control framework relies on the sensor set that measures the highest pressure difference between the blade’s pressure side and suction side. Higher pressure differences translate to greater structural loads, which are critical for the turbine’s structural integrity. Therefore, step 2 includes calculating the pressure difference AP = Pmax— Pminfor both upper and lower sensor sets 9,10 and discarding the measurements from the sensor set with the smallest pressure difference AP.

[0131] Step 3: Compute relevant physical quantities:

[0132] The goal of combining sensor signals and computing physical quantities is to minimize the number of signals while effectively distinguishing between different flow conditions. Each flow condition is described by a certain state vector s(t), and different flow conditions should not have similar state vectors. Two different flow conditions require different optimal pitching. If they are described by the same state space, the control law will generate a sub-optimal blade angle for each flow condition. Here, the state vector is optimized to yield unique state vectors for all wind conditions by optimally combining and processing various sensory inputs. P3049PC00

[0133] The total force Faero acting on the blade is obtained by integrating the pressure distribution where

[0134] • Pi is the pressure of sensor i

[0135] • is the unit vector normal to the blade surface

[0136] • As; is the blade surface of sensor i

[0137] One may then obtain the radial and tangential forces

[0138] Faero FR6R+ Ff-Pf- where

[0139] • eRis the unit vector in the radial direction

[0140] • etis the unit vector in the tangential direction

[0141] The estimation of FRand Ftfrom the pressure distribution allows to aggregate many pressure signals into two representative values. The radial force contains the dynamic stall vortex footprint, and indicates which side is the suction side. The tangential force is linearly related to the torque generated by the blade.

[0142] The pressure pSide 1 and Pside 2 fromtwo pressure sensors 9a placed on each side of the blade within a distance proximate the leading edge 24a, in particular within a distance of 10% of the chord length X / C<10%, are processed in the control system. The pressure sensor signals pSide 1 and pSide 2 allow to differentiate the state vectors when there is an ambiguity with only radial force FRand tangential force Ft, where two different flow fields would produce similar loads. Such a scenario can happen post stall, when forces are low and could resemble a different flow state in attached flow conditions with a small angle of attack. Probing pressure measurement on either side of the blade 6 offers additional information on whether or not a suction peak exists at the blade’s leading edge 24a. Two very different flow fields might produce similar FRand Ft, but they require different actuations. The above described pressure measurements allow to distinguish such cases and exclude false neighbors.

[0143] When the flow is fully separated, the vortex shedding induces large detrimental structural vibrations. Pressure measurements experience large fluctuation and an increase in the signal-to-noise ratio, making their use for control cumbersome. In that case, the flow state s(t) is primarily characterized P3049PC00 by the amplitude vortex shedding frequency in the blade deformation sensor power spectrum. In the case of stalls, large amplitude vibrations at the vortex shedding frequency will inevitably occur. The expected vortex shedding frequency is known from airfoil’s geometry and the effective wind speed obtained from the anemometer 13 and turbine rotation frequency obtained from the turbine angular position sensor 14. The amplitude of these vibrations is a great indicator of the flow state.

[0144] The magnitude of vortex related frequencies may be obtained by computing an average of the blade deformation sensor signals and then computing the power spectrum of the averaged signal. An example of an amplitude - frequency power spectrum is illustrated in figure 3h. The maximum amplitude Amaxis selected within a window around the shedding frequency from the blade deformation sensor outputting the largest measurement signal (typically from the sensor set on the upper blade 6). Structural wear and failure are caused primarily by shedding, where Amaxquantifies the shedding intensity.

[0145] The synergy between the different sensors and the information they provide about specific physical phenomena enable efficient real-time control in all wind conditions.

[0146] Step 4: Generate state vector s(t)

[0147] A state vector s t) that incorporates the synergy between sensor inputs and an optimal differentiation of the flow states may be generated in the control system.

[0148] Step5: Select appropriate control law

[0149] Depending on the current wind reading from the anemometer 13 input into the control system 4, the control law to best achieve the corresponding objective is determined. An example of objectives depending on wind speed Um[m / s] measured by the anemometer 13 is illustrated in the table below: P3049PC00

[0150] Step 6: Determine optimal pitch angle

[0151] The optimal pitch angle may be computed in the control system 4 using the state vector s(t) and the appropriate control law determined in the previous step.

[0152] The optimal control law can be simplified as a continuous function f:

[0153] This function is schematically represented on the simplified 2D representation of the feature space in figure 3h. The full feature space is spanned by all dimensions of the state vector (5 dimensions). The contour indicates the optimal pitch angle corresponding to the coordinate of the state vector s t).

[0154] Selection of the optimal pitch angle:

[0155] 1. Place the state vector s(t) in the feature space

[0156] 2. Get the optimal pitch angle at these coordinates.

[0157] 3. Send the pitch angle to pitching actuators 7.

[0158] Note: Each wind regime has a different optimal control law (different pitch angle distribution), and the appropriate control law is selected at step 5. These optimal control laws may be obtained from a data-driven optimization procedure during a calibration phase.

[0159] The optimal pitch angle is transmitted to the blade actuator to adjust the blade pitch angle.

[0160] This procedure describes the general functioning of the method, and more details about the specific control modules are given below.

[0161] Control software modules

[0162] The control software modules according to embodiments of the invention are illustrated in figures 8 and 9a to 9e and described in more detail below. Algorithms 1, 2 and three describe the initialization module, calibration module, and real-time control module, respectively. Algorithms 4 and 5 provide more in-depth information about specific steps of the control procedure.

[0163] Algorithm 1 : Initialization module P3049PC00

[0164] Algorithm 1 outlines the design of the controller feature space and control points, which are essential components of the real-time control module detailed in Algorithm 3. Real-time blade control relies on mapping the blade State into this feature space, effectively capturing relevant flow events. The feature space is populated with control points that re-route the trajectory toward regions of better performance. The feature space is spanned by the metrics (see Algorithm 4), and the control points are generated using baseline data from the unactuated wind turbine. Algorithm 1 is performed only once for a given turbine at a given location and can be re-used for different objectives (see Algorithm 5).

[0165] Some steps of the Algorithm 1 flow chart illustrated in figure 9a are indicated with a number, which corresponds to the explanation number given below. Convergence: The unactuated dataset serves as input for determining control points in the feature space (see Clustering below). These control points must correspond to physically relevant coordinates in the feature space to effectively represent blade States expected during wind turbine operations. This is ensured by collecting unactuated data for a time long enough as to capture the statistically most relevant events. The convergence is reached when the hypervolume containing the trajectory does not increase anymore. Clustering: The control points in the feature space are the centroids of the clusters obtained by clustering the unactuated trajectory. Among various clustering algorithms, spectral clustering advantageously yields clusters that better follow the temporal evolution of the data. The clusters identify the different trajectories in case of crossings, or neighboring trajectories evolving in opposite directions, where other clustering algorithms would group these trajectories in the same cluster. Dimensionality reduction: The location of the unactuated trajectory and the centroids / control points in the feature space is determined visually. Such inspection cannot be done if the feature space is spanned by more than three metrics, in which case a dimensionality reduction step is performed. Dimensionality reduction is the transformation of data from a high-dimensional space into a lowdimensional space so that the low-dimensional representation retains the key feature of the original data.

[0166] Algorithm 2: Calibration module

[0167] Algorithm 2 presents the control law optimization methodology. A control law is a set of control values attributed to all centroids (see Algorithm 3). The control law must be optimized for each objective, which may be power increase, power regulation, lifespan extension, etc (see Algorithm 5). Performing an optimization task is tedious with an experimental setup because of measurement noise and the inherent randomness of the system. Performing twice the same experiment will never give exactly the same result, and there might be random outliers, which will significantly change the output. Most numerical simulations are deterministic by nature and not prone to randomness. The optimization method using the original cluster-based feedback control may not converge as well as Bayesian P3049PC00 optimization. Bayesian optimization relies on Gaussian processes, which can naturally incorporate noise into its framework, and it is highly sample-efficient, allowing to reduce the number of experimental iterations.

[0168] Some steps of the Algorithm 2 flow chart illustrated in figure 9b include:

[0169] Step 1 - Performance monitoring: The performance is not necessarily computed from the feedback signal, but rather from additional sensors mounted on the wind turbine. These sensors include, but are not limited to, a torque meter for measuring power production, accelerometers for monitoring structural vibrations, and strain gauges on the tower for measuring turbine drag.

[0170] Step 2 - Performance computation: The performance is computed as the time average of the monitored performance metric.

[0171] Algorithm 3: Real-time control module

[0172] Algorithm 3 presents the main steps of the cluster-based feedback control, based on the method described in the original paper (Nair, 2019), and adapted for the experimental application. In this method, each centroid is attributed to a constant blade angle value, or control value. The vector grouping these control values is the control law. This control law may be optimized for the control to achieve best performance (see Algorithm 2 - Calibration module). The main feature is to use sensory feedback to track the trajectory of the blade State in the feature space. At each iteration, the actuation blade angle is determined by the average of the control values, weighted by the respective distance of the control points to the metric coordinates.

[0173] Steps of the Algorithm 3 flow chart illustrated in figure 9c include:

[0174] Step 1 - Distance computation: For low feature space dimensions (<10 dimensions), the Euclidean distance works well to compute the distance in the feature space and is therefore well suited for a setup with a few strain gauges. In case of pressure sensors, it’s easy to place a larger number of sensors, and if this number exceeds 10, it is recommended to use more appropriate metrics such as the cosine similarity or the Manhattan distance.

[0175] Step 2 -Control frequency: The control frequency is a critical enabler for efficient closed-loop control, and depends on the filtering, the sensor sampling frequency and the ability to capture the relevant physics. The minimal control frequency is the minimal value of these three values and is application dependent. In case of a small-scale wind turbine with high rotational frequency, the threshold might be the signal filtering. For a larger turbine, the signal filtering is less critical as the rotor rotates slower, and the threshold can become the need to capture the relevant physics. More details about each parameter are given below: Filtering: The frequency must be high enough to enable robust filtering of the sensors signal. Filtering induces a delay (see Algorithm 1), which can be reduced by increasing the control frequency. The minimum sampling frequency is application dependent and depends on the signal noise and the required control reaction time. P3049PC00 Sensor sampling frequency: The sensors have a sampling frequency which can affect the control frequency. Better quality hardware can usually achieve higher sampling frequency. The minimal achievable frequency is set by the hardware. Measuring relevant physics: The control frequency must be high enough to capture the relevant physics of the problem. It does not need to capture the high-frequency turbulent fluctuation, as their influence is magnitudes smaller than that of the leading-edge vortex, which is a large and low frequency coherent structure. Research by the inventors has shown that the leading-edge vortex life cycle follows a universal growth and reaches maturity after 4 convective times, time after which it separated from the blade and sheds in the turbine wake. The control preferably should be able to pick up the vortex growth to react on time, which can be achieved with an actuation frequency corresponding to at least l / 10thof the leading-edge full growth, or 0.4 convective times. The corresponding numerical can be scaled to each application based on the anticipated characteristic blade velocity, wind velocity and turbine geometry.

[0176] Algorithm 4: Acquire sensory feedback

[0177] Algorithm 4 describes the process of collecting feedback signal from the sensors measuring the flow state and blade vibration. This algorithm is fundamental and is used in all other algorithms. The algorithm includes the process from the sensor sending a signal, its pre-processing, up to the point where it can be used by the closed-loop control algorithm. VAWTs may operate in a mechanical / electronic environment where signals are polluted by electromagnetic noise and structural vibrations, unlike numerical simulations, which produce much cleaner signals. It is therefore advantageous to perform several cleaning and fdtering steps, to ensure the reliability of the input signal to the closed-loop control algorithm and a robust control.

[0178] Some steps of the Algorithm 4, illustrated in figure 9d, include:

[0179] Step 1 - receive sensor measurement data.

[0180] Step 2 - Signal processing: Cleaning and filtering the sensor signal is a critical step for experimental closed-loop on a mechanical system. The closed-loop controller operates at high frequency, so any disturbance in the feedback signal can result in a sudden high amplitude actuation, which might be detrimental for the system’s integrity. More specifically, the signal must be processed to address: Electromagnetic noise: pollutes all signals in experimental measurements. Even in a completely still environment, the sensors produce a noisy signal with a mean and a standard deviation. The standard deviation depends on numerous parameters, such as electronic components quality, cables shielding, and electric grounding of the components. Two strategies have been successfully applied to remove the electromagnetic noise. First, a moving average, which replaces the sensor value by P3049PC00 the averaged value from the N previous measurement, where N describes the window size. Second, a low-pass filter with a cut-off frequency beyond the relevant physical frequencies but below the frequencies of electromagnetic noise. The cut-off frequency is problem dependent, and may be set to 30Hz for our laboratory prototype. Both methods require a window size, a larger window size generates a stronger smoothing, but also induces a lag in the signal, which is detrimental for closed- loop control. The best window size is application dependent and may be experimentally optimized in each case. Outliers: The feedback signal is commonly polluted by outliers. Outliers are spurious data points with a magnitude significantly different from the local median value . The corresponding blade angle will significantly differ from the previous ones, resulting in possibly dangerous actuation commands. If the value of a data point is three times beyond the local median, it is considered outlier and replaced by the median value. This median is computed with the M last measured value, M being optimized for each system, and usually smaller than the filtering window size N. Structural resonance: Any mechanical system has resonant frequencies, which can be triggered when the actuation frequency is closed to a resonant frequency and can lead to structural failure. This is critical in closed-loop control application, as the structural resonance can be picked-up by sensors such as strain gauges and fed back to the control algorithm. The vibrations are then amplified and picked up again, in a vicious circle. Such behavior may be prevented by applying a band-pass filter centered around the resonance frequencies of the system.

[0181] Step 3 - State computation: The blade State is a vector containing all relevant physical quantities that characterize the flow acting on the wind turbine blade. These state vector componetns also define the axes of the feature space (see Algorithm 1). The quantities are derived either from processed sensor signals or computed by combining these signals to emphasize specific physical features (see the description "Example of a real-time control flow ’ above). In the present example, the state vector components are obtained using the pressure sensors and deformation sensors. The pressure field around the blade informs about state of the boundary layer (attached or detached), the suction and pressure side position, the velocity magnitude at the leading edge and the position of the stagnation point. These features are fundamental for a vertical-axis wind turbine blade and may be measured by the pressure sensors on the blade surface.

[0182] Algorithm 5: Switch between control objectives

[0183] Algorithm 5 (illustrated in figure 9e) presents the transition between the control objectives. The current regime is determined with the wind speed measured by an anemometer. Depending on the regime, the proper objective is selected. Examples of flow regimes and the corresponding objectives are illustrated in figure 10. P3049PC00

[0184] All objectives use the same feature space, so the feature space design (Algorithm 1: Initialization module) must be performed only once. Each optimized control law is stored and re-used whenever the wind turbine operates in this flow regime again.

[0185] The VAWT according to embodiments of the invention comprises an adaptive control which enables optimization of the actuation of the turbine geometry taking into account the wind profdes on site and various other local conditions such as pressure variations, humidity, temperature, and ice accretion on the blades. The sensor and control system according to embodiments also provides health monitoring with the acquired sensor data for predictive maintenance. The VAWT according to embodiments is also robust against partial sensor failure because the sensor and control system creates a model for the pressure sensor distribution from the collected data, such that if one a more pressure sensors fail, the control can model their values using the signals from the healthy sensors. The VAWT according to embodiments of the invention may thus be operated optimally to achieve a range of objectives including:

[0186] ■ regulating power production (increase or decrease depending on the current regime),

[0187] ■ reducing structural wear and extend lifetime,

[0188] ■ reducing start-up velocity,

[0189] ■ improving safety in survival wind conditions.

[0190] P3049PC00

[0191] List of references used

[0192] Vertical axis wind turbine 1

[0193] Generator 101

[0194] Tower 103

[0195] Turbine 102

[0196] Hub 104

[0197] Turbine rotor arm 105

[0198] Blade device 2

[0199] Blade axis Ab

[0200] Blade rotor 5

[0201] Blade 6

[0202] Leading edge 24a

[0203] Trailing edge 24b

[0204] Chord 26

[0205] First extremity 28a

[0206] Second extremity 28b

[0207] Blade actuator 7

[0208] Sensor System 3

[0209] Blade sensors 8

[0210] Blade profile pressure sensors 9

[0211] Blade deformation sensor 10

[0212] Inertial measurement sensor (IMU)

[0213] Strain gauge

[0214] Blade angular position sensor 11

[0215] Blade torque sensor 12

[0216] Wind speed sensor (e.g. Anemometer) 13

[0217] Turbine angular position or rotation speed sensor 14

[0218] Turbine torque sensori 6

[0219] Control system 4

[0220] Turbine Controller 18

[0221] Blade controllers 20

[0222] Control software modules 22

[0223] Initialization module 22a

[0224] Calibration module 22b

[0225] Realtime blade control module 22c

Claims

28P3049PC00Claims1. A vertical axis wind turbine (VAWT) ( 1 ) comprising a turbine ( 102) for coupling to a generator (101), the turbine (102) rotatable about a vertical turbine rotation axis and comprising a plurality of blade devices (2) disposed around the vertical turbine rotation axis coupled via at least one turbine rotor arm (105) to a rotor of the VAWT, each blade device (2) comprising at least one blade (6) and at least one blade actuator (7) coupled to the blade (6) configured to vary a pitch angle of the at least one blade, the turbine further comprising : a sensor system (3) including a blade sensor system, and a control system (4) connected to the sensor system and to the blade actuator (7) in a feedback loop for real-time control of the blade pitch angle, the control system configured to receive measurement data from the sensor system and to output a target blade pitch angle to the blade actuator (7), wherein the blade sensor system (8) further comprises at least one blade deformation sensor (10) arranged on the blade and a plurality of pressure sensors (9) arranged on a surface of the blade, on opposite sides of the blade relative to a chord of the blade, the at least one deformation sensor and plurality of pressure sensors connected to the control system and outputting at least a portion of said measurement data, the control system comprising software configured to compute a fluid flow state around the blade based on the measurement data output by the pressure sensors, the control system further configured to compute said target blade pitch angle based at least on said fluid flow state and an output of said blade deformation sensor as a function of a control objective, said control objective stored in the control system, input into the control system, or computed in the control system as a function of at least a wind condition impinging upon the VAWT.

2. The VAWT of the preceding claim wherein the pressure sensors (9) include at least one pressure sensor on each side of the blade within a distance from a leading edge (22a) of the blade ranging from 5% to 30% of a chord length of the blade.

3. The VAWT of any preceding claim wherein there are a plurality of pressure sensors arranged in a distributed manner chordwise between the leading edge (24a) and a trailing edge (24b) of the blade.

4. The VAWT of the preceding claim wherein there are at least three pressure sensors arranged in said distributed manner chordwise between the leading edge (24a) and a trailing edge (24b) of the blade.

5. The VAWT of any preceding claim wherein there are a plurality of pressure sensors arranged in a distributed manner spanwise between a first extremity (28a) and a second extremity (28b) of the blade.P3049PC006. The VAWT of the preceding claim wherein there are at least three pressure sensors arranged in said distributed manner spanwise between the first extremity (28a) and the second extremity (28b) of the blade.

7. The VAWT of any preceding claim wherein the at least one blade deformation sensor comprises an inertial measurement sensor (IMU).

8. The VAWT of any preceding claim wherein there are a plurality of said blade deformation sensors mounted on each blade.

9. The VAWT of any preceding claim wherein the at least one blade deformation sensor is positioned in proximity to said plurality of pressure sensors and forms therewith a sensor set extending in a band chordwise across the blade surface.

10. The VAWT of the preceding claim wherein the sensor set comprises a flexible circuit with at least one blade deformation sensor and plurality of pressure sensors connected thereon.

11. The VAWT of any preceding claim wherein the pressure sensors are MEMs pressure sensors.

12. The VAWT of any preceding claim wherein the blade device comprises a pair of said blades coupled to a centrally positioned said blade actuator, each blade extending from the blade actuator to a free end extremity (28b.

13. The VAWT of the preceding claim wherein the at least one blade deformation sensor (10) and at least some of said plurality of pressure sensors (9) are mounted proximate said free end extremity of each blade, distal from the blade actuator.

14. The VAWT of any preceding claim wherein the sensor system further comprises a wind speed sensor (13) connected to the control system, a measurement output of the wind speed sensor wind constituting or partially forming said wind condition impinging upon the VAWT.

15. The VAWT of any preceding claim wherein the blade actuator comprises a rotary electrical motor mounted on said turbine rotor arm and coupled to a rotor of the blade device.

16. The VAWT of any preceding claim wherein the control system comprises an individual blade controller (20) for each blade device (2).

17. The VAWT of the preceding claim wherein the blade controller is mounted proximate, on, or in, the blade actuator (7).P3049PC0018. The VAWT of any preceding claim wherein the control system comprises control software (22) including an initialization module (22a), a calibration module (22b) and a real-time blade control module (22c).

19. The VAWT of the preceding claim wherein the initialization module (22a) is configured to characterise and discretise the state space in which the VAWT typically operates, the calibration module (22b) is configured to optimise the blade pitch angle value associated the each discrete node obtained from the initialization modules (22a) and the real-time blade control module (22c) is configured to compute the optimal blade pitch angle at any instant by performing a weighted average of the optimal blade pitch angle values obtained in the calibration module (22b), with weights proportional to the distance between the current flow state and the discrete nodes in the state space.

20. A method of controlling a VAWT according to any preceding claim comprising the steps of :- inputting measurement signals from the sensor system (3), in particular the blade deformation sensors (10) and blade profile pressure sensors (9) into the control system (4),- computing a flow state from the pressure sensor measurement signals,- computing a power spectrum from the blade deformation measurement signals,- combining and processing the measurement signals to generate a state vector (s(t)).- defining a control objective based at least in part on a wind speed,- defining a control law adapted to achieve said defined objective,- computing a blade pitch angle in the control system using the state vector and the defined control law.

Images