Wind turbine blade monitoring

Abstract

The disclosure relates to monitoring blades of a wind turbine rotor shaft. Example embodiments include a computer-implemented method of monitoring blades of a wind5 turbine (100), the wind turbine (100) having a plurality of blades (103a, 103b) mounted to a hub (102), each blade (103a) comprising one or more strain sensors (109a1-3) mounted to the blade (103a) at corresponding respective distances from the hub (102, 201) along the blade (103a), the method comprising: i) recording (1301) strain measurements from the strain sensors (109a1-3) on each of the plurality of blades10 (103a); ii) recording (1302) a rotor speed of the wind turbine (100); iii) calculating (1303) a statistical measure of each of the recorded strain measurements; iv) calculating (1304) a difference in the statistical measure for strain sensors at corresponding positions on each of the plurality of blades over a range of rotor speeds; and v) identifying (1305) the presence or absence of damage in one of the plurality of blades15 based on the calculated difference.

Description

[0001] WIND TURBINE BLADE MONITORING

[0002] Field of the Invention

[0003] The invention relates to a method and system for monitoring blades of a wind turbine for identifying damage.

[0004] Background

[0005] Wind turbines are subjected to intermittent and varying loading during use, which can lead to unpredictable wear and damage of various components that are subject to such varying loads. The blades of a wind turbine are subjected to high and variable bending loads, which can in some cases lead to cracking. A crack originating in a rotor blade may over time propagate and weaken the blade. Regular monitoring may in some cases be able to determine the presence of a crack or other defect in time to prevent failure, but early signs of impending failure may be difficult to detect through regular site inspections.

[0006] Summary of the Invention

[0007] According to a first aspect of the invention there is provided a computer-implemented method for monitoring blades of a wind turbine having a plurality of blades mounted to a hub, each blade comprising one or more strain sensors mounted to the blade at corresponding respective distances from the hub along the blade, the method comprising: i) recording strain measurements from the strain sensors on each of the plurality of blades; ii) recording a rotor speed of the wind turbine; iii) calculating a statistical measure of each of the recorded strain measurements; iv) calculating a difference in the statistical measure for strain sensors at corresponding positions on each of the plurality of blades over a range of rotor speeds; and v) identifying the presence or absence of damage in one of the plurality of blades based on the calculated difference.

[0008] The statistical measure may be a variation of the recorded strain measurements.

[0009] The variation may be a peak to peak, RMS, standard deviation or variance of the recorded strain measurements. The calculated difference may be a change over time in the variation in recorded strain measurements for one blade relative to the other blades. The change may be an increase over time.

[0010] Each blade may have a plurality of strain sensors mounted to the blade at corresponding respective distances from the hub along the blade, the difference in the statistical measure being a difference in the statistical measure between corresponding pairs of strain sensors on each of the plurality of blades over the range of rotor speeds and the presence or absence of damage being identified based on the calculated difference for one of the plurality of blades and the calculated difference for the other ones of the plurality of blades.

[0011] The statistical measure may be a change in a mean or median of the recorded strain measurements over the range of rotor speeds

[0012] The statistical measure may be a variation of the recorded strain measurements.

[0013] The variation may be a standard deviation, peak to peak, RMS or variance of the recorded strain measurements.

[0014] The range of rotor speeds may be limited to below a rated rotor speed for the wind turbine.

[0015] The range of rotor speeds may be limited to above a minimum rotor speed for the wind turbine.

[0016] The range of rotor speeds may be limited to below around 90% of a rated rotor speed for the wind turbine.

[0017] The calculated difference for each of the plurality of blades may be an average over the range of rotor speeds.

[0018] Step v) may comprise determining whether the calculated difference for the one of the plurality of blades diverges from the calculated differences for the other ones of the plurality of blades by more than a predetermined threshold.

[0019] The predetermined threshold may be based on a range encompassing the calculated differences for the other ones of the plurality of blades.

[0020] The predetermined threshold may be a multiple of the range encompassing the calculated differences for the other ones of the plurality of blades.

[0021] The rotor speed may be determined from a measurement of rotor speed of a rotor of the wind turbine.

[0022] The rotor speed may be determined from the recorded strain measurements, for example from a fundamental frequency of the recorded strain measurements. Each blade may comprise three strain sensors mounted to the blade at corresponding respective distances from the hub along the blade.

[0023] Step iv) may comprise calculating a difference in the variation between first and second ones of the strain sensors and between second and third ones of the strain sensors on each of the plurality of blades over the range of rotor speeds.

[0024] According to a second aspect there is provided a system for monitoring blades of a wind turbine, the wind turbine having a plurality of blades mounted to a hub, the system comprising: one or more strain sensors mounted to each of the plurality of blades at corresponding respective distances from the hub along the blade; a computer connected to receive signals from the strain sensors, the computer configured to: i) record strain measurements from the strain sensors on each of the plurality of blades; ii) record a rotor speed of the wind turbine; iii) calculate a statistical measure of each of the recorded strain measurements; iv) calculate a difference in the statistical measure for strain sensors at corresponding positions on each of the plurality of blades over a range of rotor speeds; and v) identify the presence or absence of damage in one of the plurality of blades based on the calculated difference.

[0025] The statistical measure may be a variation of the recorded strain measurements.

[0026] The variation may be a peak to peak, RMS, standard deviation or variance of the recorded strain measurements.

[0027] The calculated difference may be a change over time in the variation in recorded strain measurements for one blade relative to the other blades.

[0028] The change may be an increase over time.

[0029] Each blade have a plurality of strain sensors mounted to the blade at corresponding respective distances from the hub along the blade, the difference in the statistical measure being a difference in the statistical measure between corresponding pairs of strain sensors on each of the plurality of blades over the range of rotor speeds and the presence or absence of damage being identified based on the calculated difference for one of the plurality of blades and the calculated difference for the other ones of the plurality of blades. The statistical measure may be a change in a mean or median of the recorded strain measurements over the range of rotor speeds

[0030] The statistical measure may be a variation of the recorded strain measurements.

[0031] The variation may be a standard deviation, peak to peak, RMS or variance of the recorded strain measurements.

[0032] The range of rotor speeds may be limited to below a rated rotor speed for the wind turbine.

[0033] The range of rotor speeds may be limited to above a minimum rotor speed for the wind turbine.

[0034] The range of rotor speeds may be limited to below around 90% of a rated rotor speed for the wind turbine.

[0035] The calculated difference for each of the plurality of blades may be an average over the range of rotor speeds.

[0036] The computer may be configured to determine whether the calculated difference for the one of the plurality of blades diverges from the calculated differences for the other ones of the plurality of blades by more than a predetermined threshold.

[0037] The predetermined threshold may be based on a range encompassing the calculated differences for the other ones of the plurality of blades.

[0038] The predetermined threshold may be a multiple of the range encompassing the calculated differences for the other ones of the plurality of blades.

[0039] The rotor speed may be determined from a measurement of rotor speed from a rotor of the wind turbine.

[0040] The rotor speed may be determined from a fundamental frequency of the recorded strain measurements.

[0041] Each blade may comprise three strain sensors mounted to the blade at corresponding respective distances from the hub along the blade.

[0042] The computer may be configured to calculate a difference in the variation between first and second ones of the strain sensors and between second and third ones of the strain sensors on each of the plurality of blades over the range of rotor speeds.

[0043] According to a third aspect there is provided a computer program comprising instructions to cause a computer to perform the method according to the first aspect. Detailed Description

[0044] The invention is described in further detail below by way of example and with reference to the accompanying drawings, in which:

[0045] Figure 1 is a schematic diagram of a wind turbine incorporating a system for monitoring blades of the wind turbine;

[0046] Figure 2 is a schematic planform diagram of an example blade of a wind turbine with a plurality of strain sensors mounted to the blade;

[0047] Figure 3 is a plot of strain gauge output as a function of time for corresponding positions on three blades of a wind turbine;

[0048] Figure 4 is a plot of strain gauge output over time, indicating mean and standard deviation measures;

[0049] Figure 5 is a plot of average strain measurements as a function of rotor speed for corresponding positions on two rotor blades;

[0050] Figure 6 is a plot of standard deviation over time for multiple strain gauges on blades of a wind turbine;

[0051] Figure 7 is a plot of calculated and measured rotor speed over time for an example wind turbine;

[0052] Figure 8 is a plot of difference in standard deviation of strain measurements as a function of rotor speed for a first pair of strain gauges at corresponding positions on the blades of a wind turbine;

[0053] Figure 9 is a plot of difference in standard deviation of strain measurements as a function of rotor speed for a second pair of strain gauges at corresponding positions on the blades of a wind turbine;

[0054] Figure 10 is a plot of mean strain as a function of rotor speed for strain gauges mounted to blades of a wind turbine;

[0055] Figure 11 is a plot of a slope of mean strain to rotor speed as a function of sensor position along each of the blades of a wind turbine;

[0056] Figure 12 is a plot of running average of peak to peak strain measurements over time for different blades of a wind turbine;

[0057] Figure 13 is a flow diagram illustrating an example method of monitoring blades of a wind turbine for damage;

[0058] Figure 14 is a schematic diagram of a system for monitoring blades of a wind turbine for damage; and

[0059] Figure 15 is a schematic diagram of an example local computer for the system of Figure 14. Figure 1 illustrates a partial schematic view of a wind turbine 100 comprising a monitoring system for monitoring the wind turbine 100. Wind turbine blades 103a, 103b are attached to the hub 102 via pitch bearings 104a, 104b, which permit the blades 103a, 103b to be rotated about their longitudinal axes 117a, 117b, driving rotation of the rotor shaft 121. The wind turbine 100 will typically have three blades, two of which are shown in Figure 1. The hub 102 is rotatably mounted via the rotor shaft 121 to a generator 105 via a gearbox 106, both of which are located in a nacelle 107 mounted on top of a tower 108. The rotor shaft 121 connects the hub 102 to the gearbox 106 and is mounted on first and second bearings 122, 123 for rotation about a rotational axis 124. The first bearing 122 may be termed an upwind bearing and the second bearing 123 may be termed a downwind bearing, based on the direction 125 of the prevailing wind.

[0060] The monitoring system may comprise various sensors mounted around or on the pitch bearings 104a, 104d, blades 103a, 103b or blade roots to enable the condition of the pitch bearings 104a, 104b, blades 103a, 103b and blade attachments to be monitored, as for example described in GB2209752.1 and GB2218937.7, the disclosures of which are incorporated herein by reference. The sensors may include one or more acoustic sensors, vibration sensors, strain sensors and displacement sensors. Further sensors may be located along the blades 104a, 104b for detecting damage events, or potential damage events, by measurement of vibration or displacement of the pitch bearings and / or the blades 104a, 104b. Displacement sensors may be mounted to measure a radial distance to an outer surface of the rotor shaft 121 for measuring runout, as disclosed in GB2302177.7, the disclosure of which is incorporated herein by reference. Sensor data from the various sensors is received by a computer 110 and recorded locally.

[0061] The computer 110 may communicate with a controller 111 of the wind turbine via a wired or wireless connection 112, for example to obtain control data regarding operation of the wind turbine 100 or to provide alert information to the controller 111. The computer 110 may communicate to a network 113 (e.g. the internet) via the controller 111 and / or via one or more wireless or wired connections 114, 115. The connections 114, 115 may for example be 4G or 5G radio communications links. Data recorded by the computer 110 may be periodically transmitted to a remote computer 116 via the network 113 for analysis. Analysis of the recorded data may also or alternatively be carried out locally by the computer 110.

[0062] The computer 110 or remote computer 116 may be further configured to analyse sensor data and provide an alert, which may be a signal transmitted from the wind turbine 100 to the remote computer 116 via the network 113. The network 113 may operate via a wireless connection, such as 4G or Wi-Fi, or may operate via a physical wired connection such as ethernet. The data recorded by the computer 110 may be periodically transmitted to the remote computer 116 via the network 113 for analysis. The time interval between recording data and the length of time spent recording each data set may be user configurable.

[0063] The computer 110 may be configured to transmit the recorded sensor data to a site server. The site server may store the sensor data from a plurality of wind turbines present on a site. For example, each wind turbine on a site may comprise a system for recording sensor data on each respective wind turbine. A plurality of computers 110 may then transmit their respective recorded sensor data to the site server independently. The site server may then periodically transmit the data recorded by the plurality of computers 110 to the remote computer 116. In some embodiments, the data may be uploaded to a cloud data storage service such as AWS, such that the data is accessible by the remote computer 116.

[0064] The computer 110, or the remote computer 116, may be configured to transmit an alert signal to the controller 111 of the wind turbine 100, depending on the recorded sensor data and analysis performed on the recorded sensor data. The controller 111 may for example be configured to cause the wind turbine 100 to cease or derate operation on receiving an alert signal indicating potential failure.

[0065] Figure 2 is a schematic planform drawing of one of the blades 103a of the wind turbine of Figure 1, on which various sensors are mounted, together with a schematic plot of edge-wise moment along the length of the blade 103a. A plurality of strain sensors 109al, 109a2, 109a3 are mounted to the blade 103a at respective distances from the hub along the blade 103a. The hub in this case corresponds to the blade root 201, which is connected to the pitch bearing 104a, connecting the blade 103a to the hub 102. A first strain sensor 109al is mounted in this example at a distance of 6m from the blade root 201, a second strain sensor 109a2 at a distance of 10m and a third strain sensor 109a3 at a distance of 14m. The strain sensors 109al-3 are in this example thereby spaced uniformly along the length of the blade 103a, although non-uniform spacing may alternatively be chosen. The strain sensors 109al-3 are in this example aligned along a common longitudinal axis 202 of the blade 103a, which enables the sensors to measure strains corresponding to bending of the blade 103a in the same sense. Other arrangements of sensors may be made, with the sensors mounted to the blade 103a at different positions along the length of the blade 103a. The sensors 109al-3 are mounted towards a trailing edge 204 of the blade 103a, in this example behind a shear web 207 extending between the pressure and suction spar-caps inside the blade 103a. In other arrangements, one or more of the sensors 109al-3 may be mounted towards a leading edge 205 of the blade 103a.

[0066] By placing the sensors 109al-3 spaced apart at different positions along the length of the blade 103a, a potential damage area 203 can be detected through measuring differences in strain measurements taken along the blade length and comparing these with corresponding measurements taken from sensors on the other blades of the turbine, as explained in further detail below.

[0067] Further sensors 206al, 206a2 may also be mounted to the blade 103a. In this example, first and second accelerometers 206al, 206a2 are mounted to the blade 103a between the strain sensors 109al-3. The accelerometers 206al, 206a2 enable vibrations in the blade 103a to be measured, which allows for measurements of resonant frequencies of the blade 103a. The accelerometers 206al, 206a2 may be three-axis accelerometers, allowing for vibration to be measured along the length of the blade 103 a and orthogonal to the length of the blade 103a. Vibrations orthogonal to the length of the blade 103a will tend to indicate flexural vibrations of the blade 103a. Typical flexural resonant frequencies of interest for wind turbine blades of the size indicated in Figure 2 are in the region of 0.5Hz to 20Hz.

[0068] Each of the sensors 109al-3, 206al-2 may be adhesively bonded to an interior surface of the blade 103a. Wind turbine blades are typically composed of a fibreglass composite material. The strain sensors 109al-3 may also be composed of a similar fibreglass composite material to allow accurate and repeatable strain measurements to be made of the underlying material. Each strain sensor and each vibration sensor (if present) is connected to a corresponding cable that extends towards the hub and which connects the sensor to the computer 110 for recording sensor data.

[0069] Sensor readings from the strain sensors 109al-3 are monitored and recorded by the computer 110 (see Figure 1) locally on the wind turbine. An example set of time series signals 301a-c from strain sensors mounted at the same position on three blades of a wind turbine is illustrated in Figure 3. The strain gauge output signals 301a-c are from strain gauges mounted at the same position on each of the three blades of the wind turbine, in this case at 6m along the length of the blade from the blade root, corresponding to the first strain gauge 109al in Figure 2. The strain sensor signals 301a- c exhibit similar periodic fluctuations corresponding to the rotational speed of the hub. Over multiple rotations, the strain signals tend to vary along with each other, while during each rotation the strain signals are around 120 degrees out of phase with each other. This corresponds to the 120 degree spacing of a three-blade wind turbine. The strain signals are predominantly driven by gravity and thus typically reach local maxima and minima when the blade is positioned around the 3 and 9 o’clock positions around the rotor, i.e. with the blade oriented horizontally.

[0070] Figure 4 illustrates an example plot of measured strain 401 for a single blade as a function of time. The measured strain 401 fluctuates between a maximum value 402 and a minimum value 403, which correspond to horizontal positions of the blade. Also indicated is the mean strain value p, which in this example is at around 1080 x 10’6. and the standard deviation a either side of this mean value, which is around 100 x IO’6. The variation over time is similar for each of the strain sensors on each blade, but with differing values for the mean strain and standard deviation.

[0071] The mean value of the strain over time is influenced by the aerodynamic thrust of the machine as the blades are more highly loaded with higher thrust. This phenomenon may be used to record the mean values of the strains against the rotational speed of the main rotor. For a typical wind turbine, the maximum thrust of the machine is near to the knee of the power curve, just before the rated speed. Figure 5 illustrates a plot of average strain as a function of rotor speed for two of the three blades of an example wind turbine. The plots 501a, 501b of average strain for the first and second blades can be seen to increase roughly linearly with increasing rotor speed between around 4 and 16 RPM below the rated speed, which in this example is at 17 RPM. The variability of average strain at the rated speed is due to the wind turbine operating at this speed under varying wind conditions, resulting in a larger variation in loading of the blades. The difference between the plots 501a, 501b in terms of the strain measurements as a function of rotor speed indicate that there may be differences in mechanical properties for the blades, which might suggest that one of the blades is defective, although could also be due to differences in the blades or in differences in how the strain gauges are mounted. Further analysis is required to determine whether any damage is indicated, which may be done by analysing strain measurements from more than one position and at corresponding positions on each of the three blades of the wind turbine. It is expected that the presence of a crack or delamination will locally weaken the mechanical structure of a blade, which should result in a change in strain behaviour of the blade. This can be measured by targeting measurement points along each blade where damage is known to present itself, enabling early detection and warning of any damage occurring or increasing. Referring again to Figure 2, the strain measurements from each of the strain gauges 109al-3 on each of the three blades on the wind turbine would be expected to be roughly similar for similar windspeed and pitch angle, i.e. with similar mean and standard deviation measures, given that the position of corresponding sensors is the same and the blades will be nominally mechanically identical. If the strain measurements for different blades is seen to be different, this could suggest that damage is present due to the load path along one of the blades being changed. The resulting change in strain profile due to blade damage does not, however, necessarily result in an increase in measured strain. The variation in edge-wise moment as a function of span length along the blade, as shown schematically in Figure 2, may result in nominally similar strain readings for positions along the length of the blade, given that the blade will tend to narrow along its length. A variation in mechanical properties, for example due to damage to the blade, will disrupt this variation in strain and may be detectable by detecting a variation in strain along the blade for different blades.

[0072] Figure 6 illustrates a plot of standard deviation of strain, 8a, over time for strain gauges mounted at positions indicated in Figure 2 for each of the three blades of a wind turbine. Measurements were taken over a 64-day period, in which the turbine was subjected to varying loading and with numerous start-up and shutdown operations. An example start-up / shutdown is indicated by a peak 604 in standard deviation, while a power production period is indicated by an extended period 602 of lower standard deviation. Periods of higher trending standard deviation are correlated to higher windspeeds and higher rotor speeds. Any damage progression is not readily evident from this type of plot.

[0073] Figure 7 illustrates variation in rotor speed over time, showing a measured rotor speed 701 derived from a wind turbine control system measurement via SCADA and a calculated rotor speed 702 derived from strain gauge measurements. The calculated rotor speed 702 is derived from a measure of fundamental frequency of the strain gauge measurements, while the measured rotor speed 701 is derived from a measure of rotation of the main rotor of the wind turbine. The rotor speed may alternatively be derived by performing a Coleman transformation to obtain multi-blade coordinates and calculate the rotor position, followed by differentiation to obtain rotational speed. The close correlation between the two measures indicates that either can be used to derive a measure of rotor speed over time. Knowing the relationship between rotor speed and strain, as shown in Figure 5, can allow useful information to be extracted that enables detection of damage over time. Figure 8 illustrates the variation in a difference between a variation in recorded strain, in this example the standard deviation of strain, from first and second strain gauges 109al, 109a2 (see Figure 2) on each of the three blades in an example wind turbine. The difference in standard deviation of strain measured for the first and second strain gauges on each blade, Aeffl2, is given by Aeffl2= £CT1— eff2, where Efflis the standard deviation of the strain measured by the first strain gauge 109al (Figure 2) and Eff2is the standard deviation of the strain measured by the second strain gauge 109a2. The difference may be positive or negative, as indicated in Figure 8. Figure 9 illustrates a similar plot for the difference between the standard deviation of strain as measured by the second and third strain gauges 109a2, 109a3 on each of the three blades, i.e. ^£<T23= £<T2— £3, using the same terminology as above.

[0074] Figure 8 shows that the strain measures Aeffl2for the first and third blades 801, 803 are close over the rotor speed range, while the strain measure for the second blade 802 diverges substantially from that for the first and third blades. The strain measurements for the first and third blades indicate that the standard deviation in strain is higher for the second position than the first position, while the strain measurements for the second blade indicates the opposite, i.e. the standard deviation in strain is lower for the second position than the first position. Figure 9, however, indicates that the standard deviation in strain between the second and third positions is roughly similar for each of the three blades, since the plots for each blade are indistinguishable. These results indicate that damage may be present in the second blade, and more specifically in the region 203 between the first and second strain gauges, which in this example indicates a region between 6 and 10m along the length of the blade, as indicated in Figure 2.

[0075] Based on the plots in Figures 8 and 9, a calculated difference in variation between strain measurements, in this example a standard deviation in strain measurements, from a pair of strain sensors on each of the plurality of blades over a range of rotor speeds can be used to identify potential damage in one of the blades. Damage in one of the blades may be identified based on the calculated difference for one of the blades and the calculated difference for the other blades. If the calculated difference between the standard deviation in strain measurements for the first and second sensors is given as Effi2= Effi— sff2, the calculated difference for each blade may be given by A£(J12a, ^£oi2c- The calculated differences may be similarly given as Eff23a, sa23b, ^Eff23cfor the second and third strain gauges. These values may be averaged over a range of rotor speeds to more accurately determine any differences between the blades. In the plots of Figures 8 and 9, a suitable range of rotor speeds may be between around 11 and 16 RPM, i.e. limiting the range to below a rated rotor speed for the wind turbine, given that the variation at the rated rotor speed for each of the blades is considerably greater, which will tend to reduce any difference indicated between the blades. The range of rotor speeds may in a general aspect be limited to below around 90% or 80% of a rated speed of the wind turbine. The range may also be limited to above a minimum rotor speed, such as around 5 or 10 RPM to avoid the increase in noise measured at low rotor speeds.

[0076] Based on the calculated differences A£CT12a. £ffi2b, ^£oi2c,acomparison between these values may be made by determining whether the calculated difference for one of the blades diverges from the calculated differences for the other blades. This may for example be determined by calculating differences between each combination of ^£ffi2a, ^£oi2b, ^£oi2cand determining if one lies outside a range of the others. In Figure 8, the calculated difference for the first blade 801 is given by A£CT12a. the calculated difference for the second blade 802 is given by £(J12Z, and the calculated difference for the third blade 803 is given by A£CT12c. It is evident from Figure 8, that £(J12Z, is the outlier, so A£CT12 / J- AfCT12a> AfCT12c- AfCT12aand A£CT12 / J- AECT12C> A£CT12c- A£CT12a. The calculated difference £(J12Z, for the second blade therefore diverges from the calculated differences A£CT12a. A£CT12cfor the other blades and the divergence for the second blade from both the first and third blades is greater than a range covering the first and third blades. A threshold may therefore be based on a divergence for one of the blades compared to a range covering the other blades. This threshold may be defined as a multiple of the range, for example if the divergence is more than twice the range for the other blades this can indicate damage in the diverging blade.

[0077] It is possible to identify damage in one of the blades using a minimum of two strain gauges for each blade but, with three or more sensors distributed along the length of each blade at corresponding distances, an identification of possible location of damage in one of the blades can be made more accurately, as indicated by the comparison in measurements in Figures 8 and 9. More than three strain gauges for each blade is also possible, although increasing the number of strain gauges further will tend to result in increased complexity and cost of installation together with a reduced marginal advantage for each additional strain gauge used.

[0078] In addition to calculated measures of variation in strain signals, a relationship between mean strain E^ and rotor speed m may be used to determine divergence in one of the blades of the wind turbine. Figure 10 is a plot of mean strain as a function of rotor speed using the same strain data recorded for Figures 8 and 9. Although the plots for the different blades are largely overlapping, the trend lines for each blade, given by Af^ / Am, differs between the blades and along each blade. Analysis of these trends results in the plots in Figure 11, which shows the variation in Af^ / Am as a function of sensor position along the length of each blade. As for the plots of variation in Figure 8, the plots 1101, 1103 for blades 1 and 3 are similar to each other while the plot 1102 for blade 2 diverges, in this case showing a higher slope at 6m and a lower slope at 10m. This is consistent with the expected damage on the blade being between the first and second sensor positions. Measurements of a change in mean strain as a function of rotor speed may therefore be used to confirm whether damage is detected on one of the blades.

[0079] Figure 12 illustrates plots of strain measurements over time for blades of an example wind turbine having three blades. Plots for a running average of peak to peak measured strain (xlO-6) over a period of 45 days are shown for each of the three blades. The measurements may alternatively be based on another measure of variation in the strain measurements such as variance, RMS or standard deviation. The second and third blades 1202, 1203 show a relatively stable measure over this time period, while the first blade 1201 shows a gradual increase in strain variation from around day 25 to day 50. This change over time indicates that damage may be developing in the first blade. In alternative configurations depending on the position of the strain gauges in each blade, the change over time may show a decrease rather than an increase.

[0080] Figure 13 is a flow diagram illustrating an example method of monitoring blades of a wind turbine. In step 1301, strain measurements £la, £2a, £lb, £2b, £lc, £2care recorded from each of a plurality of strain sensors on each of a plurality of blades of the wind turbine. The number of strain sensors on each blade may for example be one, two, three or more and the number of blades is typically three. In step 1302, a rotor speed of the wind turbine is recorded. This may be derived from the recorded strain measurements or may be obtained from a controller of the wind turbine. In step 1303, a statistical measure is calculated in each of the recorded strain measurements, the statistical measure being for example a mean or median of the strain measurements or a variation such as a standard deviation, variance or amplitude, for example RMS or peak-peak. In step 1304, a difference is calculated, which may be between strain measurements from corresponding strain gauges on different blades or in the variation between corresponding pairs of strain sensors on each of the plurality of blades, which is calculated over a range of rotor speeds. In step 1305, these differences are compared to determine if any one diverges by more than expected from the others, for example by greater than a predetermined threshold. If there is a divergence detected, damage may be identified in step 1308. Otherwise, no damage is identified in step 1306. The process end in step 1307. The process may repeat periodically, for example once a further set of strain measurements has been recorded over a period of time.

[0081] Figure 14 illustrates schematically an example system 101 for monitoring blades of a wind turbine. The system 101 comprises a plurality of strain sensors 109a 1-3, 109bl-3, 109cl-3 mounted to each blade at corresponding respective distances from the hub along the blade. A local computer 110 is located on the wind turbine is connected to record displacement signals from each of the displacement sensors 109a-g. The local computer 110 may periodically record and transmit recorded sensor data to a remote computer 116. Transmission of the recorded data may be done via a wireless connection, for example via a mobile data connection, that connects the local computer 110 to the remote computer 116 via the internet 113. Signal processing may be carried out by the remote computer 116, with the local computer 110 performing data gathering and transmission functions. The remote computer 116 may perform one of more calculation steps on the recorded data.

[0082] Operation of the system 101 is as described above, with processing of the recorded displacement signal data typically being carried out by the remote computer 116. The remote computer 116 may for example be a cloud-based computing service and may receive recorded signal data from a plurality of wind turbine monitoring systems.

[0083] Referring to Figure 15, an example computer 110 for the system described above includes a non-transitory computer-readable medium with program instructions stored thereon for performing the above-described method. In some embodiments, the computer 110 may include at least one memory 1503, at least one processor 1502, a network interface 1504 and a sensor interface 1506 for receiving signals from one or more strain sensors. Additionally or alternatively, in other embodiments the computer 110 may include a different type of computing device operable to carry out the program instructions. For example, in some embodiments, the computer 110 may include an application-specific integrated circuit (ASIC) that performs processor operations, or a field-programmable gate array (FPGA).

[0084] While the local computer 110 of the system may be included in a single unit and / or provided in a distinct housing 1401, as shown in Figure 15, in other embodiments at least some portion of the computer 110 may be separate from the housing 1501. For example, in some embodiments, one or more parts of the computer 110 may be part of a smartphone, tablet, notebook computer, or wearable device. Further, in some embodiments, the computer 110 may be a client device, i.e., a device actively operated by the user, while in other embodiments, the computer 110 may be a server device, e.g., a device that provides computational services to a client device. Moreover, other types of computational platforms are also possible in embodiments of the disclosure.

[0085] The memory 1503 is a computer-usable memory, such as random-access memory (RAM), read-only memory (ROM), non-volatile memory such as flash memory, a solid- state drive, a hard-disk drive, an optical memory device, and / or a magnetic storage device. The memory 1503 may be used to store recorded sensor data prior to being transmitted.

[0086] The processor 1502 of the computer 110 includes computer processing elements, e.g., a central processing unit (CPU), a digital signal processor (DSP), or a network processor. In some embodiments, the processor 1502 may include register memory that temporarily stores instructions being executed and corresponding data and / or cache memory that temporarily stores performed instructions. In certain embodiments, the memory 1503 stores program instructions that are executable by the processor 1502 for carrying out the methods and operations of the disclosure, as described herein.

[0087] The network interface 1504 provides a communications medium, such as, but not limited to, a digital and / or an analog communication medium, between the computer 110 and other computing systems or devices. In some embodiments, the network interface 1504 may operate via a wireless connection, such as IEEE 802.11 or BLUETOOTH, using an antenna 1505 to send and receive signals, while in other embodiments the network interface 1504 may operate via a physical wired connection, such as an Ethernet connection. Still in other embodiments, the network interface 1404 may communicate using another convention. The network interface 1504 may also or alternatively operate according to a wireless telecommunications standard, for example a 3G, 4G, 5G or other standard, to transmit and receive data.

[0088] Other embodiments are with the scope of the invention, which is defined by the appended claims.

Claims

CLAIMS1. A computer-implemented method of monitoring blades of a wind turbine (100), the wind turbine (100) having a plurality of blades (103a, 103b) mounted to a hub (102), each blade (103a) comprising one or more strain sensors (109al-3) mounted to the blade (103a) at corresponding respective distances from the hub (102, 201) along the blade (103a), the method comprising: i) recording (1301) strain measurements from the strain sensors (109al-3) on each of the plurality of blades (103a); ii) recording (1302) a rotor speed of the wind turbine (100); iii) calculating (1303) a statistical measure of each of the recorded strain measurements; iv) calculating (1304) a difference in the statistical measure for strain sensors at corresponding positions on each of the plurality of blades over a range of rotor speeds; and v) identifying (1305) the presence or absence of damage in one of the plurality of blades based on the calculated difference.

2. The method of claim 1, wherein the statistical measure is a variation of the recorded strain measurements.

3. The method of claim 2, wherein the variation is a peak to peak, RMS, standard deviation or variance of the recorded strain measurements.

4. The method of claim 2 or claim 3, wherein the calculated difference is a change over time in the variation in recorded strain measurements for one blade relative to the other blades.

5. The method of claim 5, wherein the change is an increase over time.

6. The method of claim 1, wherein each blade (103a) comprises a plurality of strain sensors (109al-3) mounted to the blade (103a) at corresponding respective distances from the hub (102, 201) along the blade (103a), the difference in the statistical measure is a difference in the statistical measure between corresponding pairs of strain sensors on each of the plurality of blades over the range of rotor speeds and the presence orabsence of damage is identified based on the calculated difference for one of the plurality of blades and the calculated difference for the other ones of the plurality of blades.

7. The method of claim 6, wherein the statistical measure is a change in a mean or median of the recorded strain measurements over the range of rotor speeds8. The method of claim 6, wherein the statistical measure is a variation of the recorded strain measurements.

9. The method of claim 8, wherein the variation is a standard deviation, peak to peak, RMS or variance of the recorded strain measurements.

10. The method of any preceding claim, wherein the range of rotor speeds is limited to below a rated rotor speed for the wind turbine (100).

11. The method of claim 10, wherein the range of rotor speeds is limited to above a minimum rotor speed for the wind turbine (100).

12. The method of claim 10 or claim 11, wherein the range of rotor speeds is limited to below around 90% of a rated rotor speed for the wind turbine (100).

13. The method of any preceding claim, wherein the calculated difference for each of the plurality of blades (103a, 103b) is an average over the range of rotor speeds.

14. The method of claim 6, wherein step v) comprises determining whether the calculated difference for the one of the plurality of blades (103a, 103b) diverges from the calculated differences for the other ones of the plurality of blades by more than a predetermined threshold.

15. The method of claim 14, wherein the predetermined threshold is based on a range encompassing the calculated differences for the other ones of the plurality of blades.

16. The method of claim 15, wherein the predetermined threshold is a multiple of the range encompassing the calculated differences for the other ones of the plurality of blades (103a, 103b).

17. The method of any preceding claim, wherein the rotor speed is determined from a measurement of rotor speed of a rotor (121) of the wind turbine (100).

18. The method of any one of claims 1 to 16, wherein the rotor speed is determined from the recorded strain measurements.

19. The method of claim 18, wherein the rotor speed is determined from a fundamental frequency of the recorded strain measurements.

20. The method of any preceding claim, wherein each blade (103a) comprises three strain sensors (109al-3) mounted to the blade (103a) at corresponding respective distances from the hub (102, 201) along the blade (103a).

21. The method of claim 20, wherein step iv) comprises calculating (1204) a difference in the variation between first and second ones of the strain sensors and between second and third ones of the strain sensors on each of the plurality of blades over the range of rotor speeds.

22. A system (101) for monitoring blades of a wind turbine (100), the wind turbine (100) having a plurality of blades (103a, 103b) mounted to a hub (102), the system (101) comprising: one or more strain sensors (109al-3, 109bl-3, 109cl-3) mounted to each of the plurality of blades (103a) at corresponding respective distances from the hub (102, 201) along the blade (103a); a computer (110) connected to receive signals from the strain sensors (109al-3, 109bl-3, 109cl-3), the computer (110) configured to: i) record (1301) strain measurements from the strain sensors (109al-3) on each of the plurality of blades (103a); ii) record (1302) a rotor speed of the wind turbine (100); iii) calculate (1303) a statistical measure of each of the recorded strain measurements;iv) calculate (1304) a difference in the statistical measure for strain sensors at corresponding positions on each of the plurality of blades over a range of rotor speeds; and v) identify (1305) the presence or absence of damage in one of the plurality of blades based on the calculated difference.

23. The system of claim 22, wherein the statistical measure is a variation of the recorded strain measurements.

24. The system of claim 23, wherein the variation is a peak to peak, RMS, standard deviation or variance of the recorded strain measurements.

25. The system of claim 23 or claim 24, wherein the calculated difference is a change over time in the variation in recorded strain measurements for one blade relative to the other blades.

26. The system of claim 25, wherein the change is an increase over time.

27. The system of claim 22, wherein each blade (103a) comprises a plurality of strain sensors (109al-3) mounted to the blade (103a) at corresponding respective distances from the hub (102, 201) along the blade (103a), the difference in the statistical measure is a difference in the statistical measure between corresponding pairs of strain sensors on each of the plurality of blades over the range of rotor speeds and the presence or absence of damage is identified based on the calculated difference for one of the plurality of blades and the calculated difference for the other ones of the plurality of blades.

28. The system of claim 27, wherein the statistical measure is a change in a mean or median of the recorded strain measurements over the range of rotor speeds29. The system of claim 28, wherein the statistical measure is a variation of the recorded strain measurements.

30. The system (101) of claim 29, wherein the variation is a standard deviation, peak to peak, RMS or variance of the recorded strain measurements.

31. The system (101) of any one of claims 22 to 30, wherein the range of rotor speeds is limited to below a rated rotor speed for the wind turbine (100).

32. The system (101) of claim 31, wherein the range of rotor speeds is limited to above a minimum rotor speed for the wind turbine (100).

33. The system (101) of claim 31 or claim 32, wherein the range of rotor speeds is limited to below around 90% of a rated rotor speed for the wind turbine (100).

34. The system (101) of any one of claims 22 to 33, wherein the calculated difference for each of the plurality of blades (103a, 103b) is an average over the range of rotor speeds.

35. The system (101) of any one of claims 22 to 34, wherein the computer is configured to determine whether the calculated difference for the one of the plurality of blades (103a, 103b) diverges from the calculated differences for the other ones of the plurality of blades by more than a predetermined threshold.

36. The system of claim 35, wherein the predetermined threshold is based on a range encompassing the calculated differences for the other ones of the plurality of blades.

37. The system of claim 36, wherein the predetermined threshold is a multiple of the range encompassing the calculated differences for the other ones of the plurality of blades (103a, 103b).

38. The system (101) of any one of claims 22 to 37, wherein the rotor speed is determined from a measurement of rotor speed from a rotor (121) of the wind turbine (100).

39. The system (101) of any one of claims 22 to 37, wherein the rotor speed is determined from a fundamental frequency of the recorded strain measurements.

40. The system (101) of any one of claims 22 to 39, wherein each blade (103a) comprises three strain sensors (109al-3) mounted to the blade (103a) at corresponding respective distances from the hub (102, 201) along the blade (103a).

41. The system (101) of claim 40, wherein the computer (110) is configured to calculate (1304) a difference in the variation between first and second ones of the strain sensors and between second and third ones of the strain sensors on each of the plurality of blades over the range of rotor speeds.

42. A computer program comprising instructions for causing a computer to perform the method according to any one of claims 1 to 21.

Images